Science of Ask Strings.pdf.pdf

The Science of Ask Strings

A data-driven look at how much to ask for online and offline

Nick Ellinger

Prologue

I once got a mail piece with a donation ask from the local Boy Scout troop. This in itself was a mite odd, as I

was once a Scout, albeit briefly. Boy Scouts helped make me the man I am today: one who does not

voluntarily sleep outside, eat things cooked over an open flame, or whittle

What stood out to me was the ask string. This acquisition piece, coming a full decade or so after my

Tenderfoot days, asked for $250, $500, or $1000. With no return envelope. That day, as you might guess, I

was not helpful, friendly, courteous, kind, or obedient. I was thrifty, though, so that’s something.

The ask string is one of the more underappreciated parts of a mail piece. In the continuing war between

analytical marketers and rule-of-cool marketers (the latter of whose only victories come when their

PowerPoints are prettier), the ask string is a safe stronghold of the analytical side. No brand guidelines to my

knowledge have ever specified the Pantone color in which the ask string must appear or opined over whether

ascending or descending is more aesthetically pleasing.

And thus one would think we data geeks would wield this tool with impunity. And yet the standard string of

highest previous contribution (HPC), HPC x 1.5, HPC x 2, and a blank for a donor-entered amount rules many

a day without a challenge to its throne. Sometimes a fourth option will pop in. Sometimes the ask string will

be accelerated to HPC, HPC x 2, HPC x 4, other. But all are still distinct echoes of the first customized strings.

We can do better. We can challenge these assumptions. And while they may still stand, they will be stronger

for the challenge.

This is a first attempt to do so. Using the available scientific literature, we’ll look at why ask strings affect us

psychologically and culturally. We will then try to establish our goals for the ask string. Then, we will get into

the various debates over:

 Fixed versus variable ask strings

 Round numbers versus true multiples

 At what anchor to put your initial amount

 How quickly to escalate

 Where to set a default (if you wish to set one)

 And so on

But why would you, presumably a direct marketing practitioner with limited time and money, waste time

with academic studies? After all, you have targets to hit. You get it done in the real world and sometimes

you (like me) find the commentary on nonprofit communications a bit idealistic, telling you all of the things

you should do, but not the time to do all of them. And it seems worst of all when it comes from someone

who hasn’t done your job, seen your demands, and faced your donors.

Let me reassure you that I’ve been there. I ran MADD’s direct mail program for nine years as the only staff

member on direct mail (while doing other duties as well). I know your time and resources are precious and

your need to hit your goal is great -- not just for you, but for the people you serve.

I am not sure what involuntary whittling is, but it cannot be pleasant.

But it is because time and resources are precious that I’ve put this together. There is a lot of punditry out

there with no basis in research. And often the tests that we do give us erroneous conclusions. It’s time to

put academic rigor into our discipline.

I once did a 15-segment A/B test (so, technically an A/B/C/D/E/F/G/H/I/J/K/L/M/N/O test). Each was made

with similar amounts of similar RFM segments and we send out the mailing. When the results came back,

our agency said we have a winner -- segment J (or whatever) did 10% better on income per piece than the

worst segment (M?). It was statistically the treatment we should roll out with and we could expect seven

cents more per piece in the future.

The trick was the segmentation was false. We had the same treatment for every one of those segments. I

had messed up the data segment and used the logic from the previous year (where we did actually have 15

test segments).

That’s where the rigor comes in. While I’ve included some more speculative pieces in here, most of these

results have been controlled well, studied, and peer-reviewed. They might have problems, but lack of rigor

isn’t going to be one of them.

Yes, an ask string is, in theory, easy to A/B test. In practice, testing one ask string strategy on one piece as a

one-size-fits-all strategy may bring in short-term gains. However, it won’t get you to a deeper understanding

of your donors’ behaviors. And it keeps you thinking about how to maximize the value of a communication,

rather than the value of a donor. It’s when you are thinking about how to treat donors that you can engage

with strategy, build donor loyalty, and maybe, just maybe, make your donors happier with their experiences.

Additionally, an A/B test can mask situations that where some audiences are better with A and some with B.

You’ll find a few of these within, where lifecycle or other factors cause ask strings to work differently among

groups.

So rigor is reason number one. Number two is it makes testing easier and more robust. So often, we create

tests because we need to have tests. In my opinion

, each idea you test should have a reason to believe it

before you start. You want to have a hypothesis that says “I believe this test should [increase response

rate/increase average gift/increase donor lifetime value/some combination]” so that you know what you are

going to measure against.

It also helps you create the test correctly. I’ve looked at some tests retrospectively and realized that I messed

it up in the test design. If you have a strong hypothesis, you can make sure you get the details right to support

or reject it.

Also (this is where I deviate only briefly from “pure” science), I like having a rooting interest in the test. You

should set up your parameters to make sure that you are giving every possibility equal chance to succeed

and fail.

But a successful test means that you’ve learned something that will move your program forward

and allow you to move your mission forward. There’s nothing wrong in wanting that.

The comment Internet parlance is IMHO - in my humble opinion. I am self-aware enough that I don’t think I can legitimately

use the H.

This includes actual testing like scientists do, not pulling out 25K records because it’s convenient or a random Nth of donors.

But that’s a topic for another book.

The purpose of a test is to beat the control. So much of our incrementalist testing returns no results. Here,

there are ideas that did beat the control and can help you keep your testing strategic and not just something

you do by habit.

Thus, someone else’s research can help you set your hypotheses. And a good hypothesis is half the battle.

My sincere hope is that if nothing else you will leave this with three things:

 Ask strings can make a significant difference in your solicitations

 Conventional wisdom may be the former, but isn’t the latter

 It’s folly to have just one ask string strategy – you must suit your tactics to the donor.

So without further ado...

The psychology of ask strings

Right now, if you wanted, you could use a magnet strip (or a series of numbers) on a plastic card to buy

passage on a giant metal bird. That bird could take you across an ocean in hours, as long as you turn off the

thing the size of your palm that connects you to all human knowledge for take-off and landing.

And yet you and I have the same mental equipment that supported our deep ancestors to decide only the

four f’s: fight, flee, feed, and, um, well, when two cave-people love each other very much (or are just

anatomically compatible)…

We may stand straighter with less hair and more clothing; mentally, we haven’t changed as much as we’d

like.

We deal with this by taking mental shortcuts, or heuristics, constantly. There’s a good, bad, and ugly to these

biases. They allow us to function in a complex world and many of them (e.g., trial and error) are pretty good

rules of thumb. However, many of our worst tendencies are embedded in this primitive coding. They poison

our unconscious mind. For our ancestors, it was useful to use the heuristic “the more the thing looks like me,

the more likely it is a friend.” For us, that’s called racism, sexism, and many other unpleasant -isms.

Heuristics lead to cognitive biases, where we skip over a number of steps in the thought process to arrive at

conclusions.

To summarize, we are not Spock.

Half the battle of knowing, that is. And since knowing is half the battle, then good hypothesis creation is 25% of the battle (plus

or minus 3%).

Technically speaking, it probably should be you are not Vulcan, rather than you are not Spock. Vulcans are possessed of pure

logic, whereas Spock is half-Vulcan, half-human, and thus capable of emotional displays, especially when J.J. Abrams is at the

helm. However, since “Spock” has become synecdoche for hyperrationality, I use it here.

You may believe the above footnote is not necessary to help protect myself against ubergeeks. May I remind you you are

reading a paper on the science of ask strings? A certain level of geekiness is presumed. Don’t feel bad about this; at least you

didn’t write the paper…

And we are not Spock in very predictable ways. For really good summaries of these, I strongly recommend

Thinking Fast and Slow by Daniel Kahneman and Predictably Irrational by Dan Ariely. I’ll cover only the ones

that relate to ask strings, but these works touch on what it is to be human in a way that I won’t be offended

if you put this down and read those first. Don’t worry; I’ll be waiting for you right here when you get back.

Anchoring

In business school, I took a wonderful class on negotiation. Our first assignment was to read a chapter of a

book, then take on a negotiation of cost with a classmate. The case was set up so that any price between

$10 and $20 would be acceptable to both parties, but I was shooting for a higher price and my classmate-

cum-opponent was shooting for a lower one.

The textbook had some of the basic rules of negotiation: don’t make the first offer, don’t concede twice in a

row, etc. I immediately saw the flaw in this: taking two competitive business school students, putting them

into a negotiation, and telling them both not to concede and not to make the first offer is a recipe for disaster.

So I resolved to break all of the written rules and some of the unwritten ones. I would make the first offer. I

would concede at the drop of a hat. I would appear to be disorganized and ill-prepared in thought, word,

and deed.

In fact, the only thing going for me was that I threw out an opening price of something like $74. And then

immediately conceded to say I could go as low as $72.

My classmate later said that the opening offer completely threw him off his game. He wondered if the cases

were set up so as not to reach agreement or give false information. And his initial negotiating point of $7

was so low in comparison that he felt he couldn’t reasonably make it.

When we ended up at a price of something like $19.30, it was all because of what’s called anchoring.

Anchoring means we rely on the first piece of information we get about something more than the last.

Researchers asked one group of people whether the population of Turkey was great than five million, then

asked them to guess the population was. These subjects averaged 17 million as guess. They asked another

group of people whether the population of Turkey was less than 65 million, then asked them to guess the

population was. These subjects averaged 35 million as a guess.

You can see the pull of the reference point.

This piece of information doesn’t have to be at all relevant. People were asked to recall the last two digits of

their Social Security number, then tell how much they would pay for an item. Those with higher numbers

gave higher prices by 57-117 percent.

(This is why if you have a focal point number in your piece, it’s good

to make it higher than your average gift. If you usually say four people die every hour, move it to 96 people

die every day; that 96, if highlighted, will ask as an unconscious anchor on giving.)

This can happen in non-numerical cases as well through the primacy effect: the first things we see or hear

have undue weight. When you hear six things about a person – three good, then three bad – you will

Tversky, A., & Kahneman, D. (1974). Judgment under uncertainty: Heuristics and biases. Science, 185, 1124-1131.

The population of Turkey is about 75 million. At the time of the study, it was 38.3 million.

Ariely, Loewenstein, Prelec (2003) Coherent Arbitrariness: Stable Demand Curves Without Stable Preferences. The Quarterly

Journal of Economics, pp. 73-105.

generally think they are a good person with some forgivable quirks. When the order is reversed to three bad,

then three good (even when they are the same attributes), you will think they are a bad person with some

negligible good qualities.

These types of things don’t matter to Spock. But to humans, they weave a web in which we all get stuck.

You can do so much with this, but you can also only do so much with this. Your donor will have reference

points you don’t know about in mind – what their income is, what they give to other charities, whether they

are going to get reimbursed for their grandson’s $1832 in in-app purchases in Kwazy Cupcakes. And what

their giving history is with you and to other causes. And how much they like you.

These mute the effect of your externally imposed anchors. So don’t think you can throw out a number of

“eleventy jillion dollars” and simply hope to get one jillion dollars.

We’ll talk more about anchoring when we talk about setting the first number in your ask string, as well as

when we talk about setting a focal point for your ask string.

Social Proof and Norms

Picture yourself at the door when your neighborhood Girl Scout comes selling cookies. You take the order

form from the girl to write down your name and you glance at…

Ding ding ding!

You are correct! You glance at how many people ordered before you and how many boxes of cookies they

ordered.

The “how many people ordered” portion is called social proof. The more people on that list, the more

conditioned you are to buy also. Our monkey brains found it to be convenient to, when it doubt, do what

everyone else is doing. And that’s now hard wired.

Anchoring ties very deeply to social proof. If you give people the impression that most people are doing a

thing, that’s an anchor. If you give the impression that most people don’t give, that also is an anchor for not

giving.

And while going deep into social proof isn’t the point of this white paper, please don’t do what Charity

Navigator did and put negative social proof in your copy. Here’s their online holiday appeal:

If even a third of the 5.5 million people who visited our site last year had

donated to Charity Navigator, we would comfortably meet our annual budget

and be able to do what so many of you ask us to do every day -- rate more

charities!

But, less than 1% of our users make a donation to Charity Navigator. Those

who do would have to give at least $50 each for us to fund our operations.

Asch, S. (1946). Forming impressions of personality. Journal of Abnormal and Social Psychology,41, 258–290.

With your help, we’ve been steadily moving the needle on this and hope more

of you will join our Charity Navigator donor family.

I think they think they trying to anchor people to give $50 or more and they may be increasing their average

gift with this because of this. However, when the first thing I hear is that less than one percent of people

give to Charity Navigator, I’m less likely to give. Or I would be if my personal likelihood were not already a

negative number.

Going back to our Girl Scout cookies, if the sheet is mostly full, you can see there’s a social expectation that

you order. You also glanced up to see what everyone else was ordering. This also ties into social proof that

functions as an anchor. The direct marketing manifestation of this is letting people know what the average

person like them donates. As you might guess, people who have their anchor set by social proof higher give

more.

Social proof for giving and amount come into play in to major gift fundraising, where funds are raised in a

quiet period to get high value and substantial gifts, so that when it is opened up, people will have high

anchors driven by social proof.

And it comes into play in peer-to-peer fundraising. It’s vital to educate people fundraising on behalf of an

organization that the most important gift they get will be their first one (ideally, the one they give to

themselves). If that first gift is $100, they will almost certainly raise more than a person who gets a $10 initial

gift. Since peer-to-peer fundraising is more giving to a person than giving to a cause, people want to know

what a socially acceptable donation is. And having many donations helps you get more donations: people

are more likely to donate to a campaign the closer it is to goal. Fundraisers on Kiva get donations at almost

twice the rate when they are two-thirds or more to goal as when they are just starting.

The bottom line is social proof helps us determine both whether and what to donate by looking at whether

people like us tend to donate and what they tend to donate.

Fluency

People like easy things. People don’t like hard things. Easy makes us happy. Happy makes us do things people

want us to do.

We like common named candidates for jobs.

Easy stock symbols go up more.

This is why ask strings exist. It’s easy to check a box. It could be hard to figure out your own dollar amount

and write it in.

Round numbers are easy things. Non-round numbers are hard things.

See, for example, Jen Shang, Americus Reed, and Rachel Croson, “Identity Congruency Effects on Donations,” the Journal of

Marketing Research, vol. 45, no. 3, 2008, 35–361. and Jen Shang and Rachel Croson, “Field Experiments in Charitable Contribution:

The Impact of Social Influence on the Voluntary Provision of Public Goods,” the Economic Journal.

C.E. Cryder et al. Goal gradient in helping behavior. Journal of Experimental Social Psychology 49 (2013) 1078–1083.

Reber, R., Winkielman, P. & Schwarz, N. (1998). Effects of perceptual fluency on affective judgments. Psychological Science, 9,

45–48.

Simon Laham, Peter Koval, Adam L. Alter. (2012) The name-pronunciation effect: Why people like Mr. Smith more than Mr.

Colquhoun. Journal of Experimental Social Psychology, 48m 752–756.

Adam L. Alter and Daniel M. Oppenheimer. Predicting short-term stock fluctuations by using processing fluency. PNAS June

13, 2006 vol. 103 no. 24 9369 –9372.

Options of three things are easy. When it doubt, we pick the one in the middle – not too much, not too

little.

We are Goldilocks.

Generally, the easier (or more fluent) something is, the more we find it to be persuasive (whether it makes

sense or not).

People who read instructions in an easy font (Arial) thought following those instructions

would be a lot easier than those who read them in a hard script font (Brush or Mistral).

So, easy is good (all other things being equal).

Just-noticeable differences

Back in the 19th century, Ernst Weber and Gustav Fechner studied physical perceptions. What they found

was:

 Humans can’t notice all differences in physical stimuli.

 The smallest amount that you can notice is called the just-noticeable difference.

 As the stimulus increases in magnitude, so does the just-noticeable difference.

Think of this in terms of a weight. If you were lifting a ten-pound weight, perhaps you might notice if I added

a half a pound to it. If you were lifting a hundred-pound weight, you likely would not notice that same

addition; maybe it takes two pounds in this scenario to be noticed.

This also works for numbers and prices. If someone overcharged you ten dollars on your take-out order, you

would notice. If they overcharged you ten dollars on the purchase of your new house, you likely would not.

Everyone has different sensitivity levels to different things. The makers of the Louisville Slugger challenged

Ted Williams to identify the lightest bat out of four. The trick was that one was 33.5 ounces and the rest

were 34 ounces. He was consistently able to do this. Mere mortals can’t.

There are two important just-noticeable differences in ask string analysis. The first is to anchoring numbers.

Remember when I said you couldn’t throw out eleventy jillion donors and hope for it to be an effective

anchor on the person’s donation? This is because the difference between that and the person’s likely

donation is so vast, it’s a very noticeable difference that the brain is able to discard. However, since the last

two digits of your Social Security number is in the same perceptual ballpark as the price of an item, those

digits can anchor price effectively.

The second just-noticeable difference is to previous giving. What a person has given in the past is a strong

anchor. And, as you might guess, this tie to previous giving strengthens the more gifts that are given at the

same level. Conventional wisdom says that once someone gives the same gift three times, they are unlikely

Paul Rodway, Astrid Schepman and Jordana Lambert. Preferring the One in the Middle: Further Evidence for the Centre-stage

Effect. Applied Cognitive Psychology. Volume 26, Issue 2, pages 215–222, March/April 2012.

Jacoby, Larry; Brooks, Lee (1984). "Nonanalytic cognition: Memory, perception and concept formation". In Bower, Gordon H.

The psychology of learning and motivation 18. New York: Academic Press. pp. 1–48. ISBN 978-0125433181.

Hyunjin Song and Norbert Schwarz. If It’s Hard to Read, It’s Hard to Do Processing Fluency Affects Effort Prediction and

Motivation. Psychological Science. 19 (10) 2008.

https://dornsife.usc.edu/assets/sites/780/docs/08_ps_song___schwarz_effort.pdf

There’s a long background and equations and stuff about this that is not relevant for our purposes.

This was later referred to as a limen or differential threshold to make it less understandable to outsiders.

Whether it goes up in proportion to the magnitude of the stimulus or the log of the magnitude of the stimulus is one of the

aforementioned things we are skipping over because it is boring and unnecessary.

to move from this amount. Like Ted Williams and his sensitivity to bat weight, long experience with

something makes us very attuned to changes in that thing.

In theory, you would try then to elevate someone’s giving by making increases in the amount that fly below

donors’ mental radars, e.g., asking for $26 when someone had previously given $25. But this could violate

fluency concerns.

I would posit a rule of thumb not technically related to ask strings: if you are asking someone to increase

their giving into what they perceive to be a different category, you need a compelling reason for them to

value you differently than they had. After all, they had a reason to value you as they did.

Cultural constructs

While not technically a cognitive bias, numerical biases do influence some giving patterns and are not logical

in the strictest Spock-y sense.

These go all the way back to Pythagoras (and possibly before), who believed in the representation of divinity

in numbers and some other things that will make you never look at a hypotenuse the same way again.

Christianity tends to focus on threes and sevens. For threes, some of this is for rhetorical purposes. After all,

two people (a priest and a Levite) came upon the wounded traveler and ignored him before the Samaritan

came and gave him aid. You probably recognize this as the same format as jokes where three people enter

a tavern where the first two guys do one thing, but the third guy does something else.

But you also have Jesus asking God three times not to be crucified, being crucified with two other people at

3 PM, being denied by Peter three times, and rising in three days.

As for sevens, there are seven days of Creation, sevenfold punishments for Cain’s killer, seven pairs of clean

species on the ark, and seven of everything in Revelations.

But these don’t tend to impact ask strings and gift amounts. There are three that do, however.

The first is giving chai. No, not tea; this chai is pronounced differently. Chai means living in Hebrew. You may

have encountered this in the toast l’chaim (to life) in the eponymous Fiddler on the Roof song. In gematria

(a numerology system used in some Jewish cultures), the letters of chai add up to 18.

Because of this, it is considered lucky among some Jewish people to give gifts in multiples of $18, with some

Jewish charities’ ask strings being or starting $18, $36, $54, etc. When Senator Joe Lieberman, the first Jewish

candidate ever a US major party ticket in 2000, was running for president in 2004, he appealed

E.g., Pythagoreans refused to eat beans because they resemble testicles.

This rhetorical device is technically known as a tricolon, in case you want to show off your local open mic night. However, take

it from one knows: open mic night audiences don’t like Greek rhetorical device humor.

010

to his supporters to give chai writ large. One email appeal asked to acquire 1800 new donors of $18 or more

in 18 days (by August 18th).

The second is the number eight in Chinese culture. The word for eight in Chinese sounds very similar to the

word for wealth/prosperity and 88 looks a bit like the stylized symbol called “double joy,” which is a very

lucky symbol.

As a result, eights are considered auspicious in everyday life and a string of them doubly so. This translates

to people wanting wedding dates, phone numbers, license plates, etc., with eights in them. It was no

coincidence that the Beijing Olympics started at 8/8/08 at 8:08 PM local time. And pricing in China will often

end with two 8’s, just like something might be priced at $199 in the US.

It also translates into charitable giving. Zhang Lei, founder of Hillhouse Capital Management, made (at the

time of the gift in 2010) the largest gift ever to Yale’s management school: $8,888,888. Reports of this gift in

the press don’t indicate whether there were also 88 cents on the gift, but I would guess there were for two

reasons: 1) it would be additionally lucky and 2) it appears Zhang would be able to afford it, a barely

noticeable difference if ever there was one.

The third is individualized. A donor may want to give an amount that corresponds to their age at diagnosis,

their number of years sober, a child’s birthday, or any number of other anchor points. These have symbolic

meaning for a person that the nonprofit will not likely know.

The point here is that we will have a battle between the fluency of round numbers and the significance of

non-round numbers that we will want to confront as we build ask strings.

The goals of ask strings

Before we run headlong into setting ask strings, it’s important to ask

what our goal is.

At this point, you are likely rolling your eyes and saying that the goal

is of course to maximize net revenue.

But how? Let’s say a campaign raises one million dollars. Would you

rather it be in the form of one one-million-dollar donation? Or one

million one-dollar donations?

The answer to this is actually neither. Single large gifts from a campaign tend not to be replicable when the

same campaign is repeated. Also, this could also make your campaign and organization dependent on one

person, which is fine while that person loves you but not as fine when they stop loving you.

And one dollar donations may be replicable, but will have significant costs to retain these donors. The cost

of resoliciting an offline donor per year is almost certainly going to eat up most, if not all, if not more than

all, of the revenues you would get from these donors.

Associated Press (8/18/2003). Bush makes re-election push on Internet. As in USA Today at

http://usatoday30.usatoday.com/news/politicselections/2003-08-18-bush-internet_x.htm.

"If you don't know where you

are going, any road will get

you there." - Lewis Carroll

"If you don't know where you

are going, you might wind up

someplace else." - Yogi Berra

011

Obviously, a normal direct marketing campaign will be closer to a Goldilocks zone than these extreme

examples. But it still brings up the question of what your goals are as you acquire and retain donors.

For acquisition, for example, what metric do you use to determine the success or failure of an acquisition

piece?

In a perfect world, you would use lifetime value. But we don’t live in a perfect world.

Lifetime value takes

time to manifest and you need to know what you are making a decision on tomorrow.

So, for your preliminary work, do you go toward net cost to acquire a donor, which will reward getting a

large number of smaller donors? Or do you go to something like net per piece, which will reward fewer larger

donors?

Or, do you look at the donors that a campaign is bringing in and their initial gift, then project out their average

giving as a poor-man’s model for lifetime value? This is a better solution, but still imperfect.

As you hone your ask string and default choices, you’ll find they are largely about trade-offs. Good ask string

choices can help you maximize revenues from your campaigns. However, there is a point at which you have

reached something close to Pareto efficiency (when your ask string choices can’t increase revenue any more)

and thus are making choices on whether you want higher response rates and lower average gifts or lower

response rates and higher average gifts along the efficient frontier.

Each choice has its own advantages, so we’ll review them here.

The argument for Hufflepuff: taking all comers regardless of value

You get lower costs from volume. Volume is a big predictor of

communication costs for means like direct mail; you save money on

all segments by having more people on file. While this impact is

minimized once you reach a certain size, it is still a factor.

You can create greater customization. Larger volume means that

you can do more segments and more customization. Perhaps it didn’t

make sense to split your list in two for a test when you had 25,000

names, but at 50,000 names, it becomes doable.

Smaller donors cost less to acquire. Because lower average gift donors generally have higher response rates,

you can generally acquire them at lower cost. I find that acquiring a sub-$10 donor can cost half as much as

acquiring a $15 donor.

Smaller donors can be a consistent source of revenue. There are some segments of donors who like to give

$5 at a time, but they will do it to every other or every third communication you send them. While it’s

I’m writing these particular lines during a presidential primary debate and feeling this imperfect world more deeply than

usual.

Even if you do take on all comers, never go full Hufflepuff and treat them all the same. People are different. Their desires are

different. Their commitments are different. Non-segmentation dishonors their individuality.

“I'll teach the lot and treat

them just the same.”

-- Helga Hufflepuff, according

to The Sorting Hat, Harry

Potter and the Order of the

Phoenix

012

013

not a home run, getting on base often means something. While it’s not the same as having a real

monthly/sustaining donor, it is semi-predictable.

You can find hidden gems. Take a look at your organization’s best current donors. Chances are fairly good

that they started out many moons ago as small donors. Yes, a person who comes into your organization at a

$100 level is far more likely to become a major donor than someone at a $5 level. But small donors give you

many more bites at the apple of finding someone who grows to love your organization and makes

commensurate gifts.

The value of donors is not just in their monetary value. Having more donors means having more people

that support you and having more people that support you means more awareness of your mission in the

community, more volunteers, and more advocates. A traditional lifetime value calculation ignores the value

of donors as volunteers and advocates, which do have their own quasi-monetary value.

Doing the things you would need to do to get only larger donors cuts off your reach. Let’s say you start

instituting ZIP code selects and high ask strings to focus on higher value donors. You will be ignoring a lot of

districts that are just plain poor. And you are going to be ignoring them with your message, mission,

awareness, and advocacy.

Smaller donors can be effective peer-to-peer fundraisers. If you go exclusively for the people who eat with

multiple folks and pinkies out, you will more rarely know who is a deacon at the church or the worker who

can pass the hat at the the plant. And casting your net broadly gives you a greater opportunity to get those

types of donors. If Doctors Without Borders had ignored a local knitter who wanted to help them after a

disaster, they would have passed up over a million dollars in revenue.

Smaller donors can be better planned givers. There is a great case study from the ASPCA.

Because they

had focused on higher-value donors, they were not getting as many bequests. In fact, they were excluding

the 70+ year old, $10-and-under givers that were their best planned giving prospects. So they made a

conscious choice to go back and reacquire these donors, sending them (only) the best house mailings and

working to upgrade them to bequest giving. Bequest giving often comes from “tippers” on your direct

marketing file of a certain age who give to help you in their lifetime, but are saving a nest egg for donation

at the end of their lives.

File size is easy to explain to your boss. Would that this were not a part of the calculation, but it is. If pure

file size is what you are measured on, you can hack this by acquiring a lot of smaller value donors. I would

advocate not being measured on this if you can help it (increasing core donors is a far better measure of file

health), but if you aren’t the general, you go to war with the units you have.

It’s egalitarian. There is something nice about saying, and meaning, that even a dollar will help. (In fact,

especially in face-to-face interactions, it seems to help response rate without hurting average gift

; we’ll

discuss this effect more later).

https://directtodonor.com/2016/06/23/getting-other-people-to-do-your-hypercustomization/

http://dma.convio.net/site/DocServer/Direct_Marketing_for_Planned_Giving_1_21_10.pdf?docID=1382

http://directtodonor.com/2016/03/02/validating-small-gifts-to-increase-response-rate/

014

The advantages of Ravenclaw: focusing on fewer, larger donors

Many donors just don’t pay for themselves. Let’s say you have a

robust multichannel solicitation program that costs you about $5 per

person to run. If your $10 donors don’t average more than .5 gifts

per year (which may be pushing it, assuming that a healthy portion

of them are first-time donors that won’t retain well), these donors

are literally losing you money every time you communicate with

them. A $5 donor requires at least a donation per year. While you

can (and should) decrease your costs in soliciting these donors, this

is obviously only able to be done to a point.

Your fundraising efficiency will be much greater. In very strong average gift segments, you can be making

over a dollar, two dollars, five dollars, or more per communication to your strong segments, a virtual

impossibility with lower dollar segments. So your fundraising efficiency is much greater.

Donors generally do what they started out doing. As mentioned earlier, if someone gives you the same

amount three times, chances are you are going to be getting that amount for the rest of their useful donor

life. In fact, we’ll talk about this in the context of ask strings -- you have more pliability with a one-time donor

than someone who as built a pattern. Upgrading is good to try to do, but you can’t count on it for the bulk

of your audience.

Retention rates go up as average gift goes up. And since loyalty goes up as average gift goes up, you can

generally tell from average gift whether someone is more likely to become a good donor for you.

Larger donors translate better to upgrades. Larger donors are the people who are more likely to become

multichannel givers, major donors, and monthly givers. Part of this is greater means and part is increased

loyalty.

Smaller donors skew your focus. There’s only so much time and attention you can give to a direct marketing

program. Too much of it goes to the Sisyphean task of trying to get $5 donors to become profitable.

Neither of these approaches are correct or incorrect. But, as Yogi and Lewis Carroll said, you have to know

where you are going to guide your experimentation.

The science of ask strings

Because make no mistake: the goal of this exercise is not to lay down inviolate law. Rather, it is give you

fodder for your own experimentation. Ask strings can be organization specific, piece specific, even donor

specific. In my direct marketing time, I’ve run experiments that contradict some of the findings here. In other

words, your mileage may vary.

Fundraising efficiency should not be a metric or good for its own sake. You can tell it’s unimportant and misleading because

Charity Navigator measures it. However, it does allow you to make greater investments in other places, which is good.

“For were there such friends

anywhere as Slytherin and

Gryffindor? Unless it was the

second pair of Hufflepuff and

Ravenclaw.”

-- The Sorting Hat, Harry Potter

and the Order of the Phoenix

015

But this is what the best evidence I could find suggests. Thus, at the least, they bring up interesting

questions about what is possible even if they don’t provide total, unanimous, and immutable answers.

Fixed ask strings

Most of the science of ask strings is related to variable ask strings that depend on who the potential donor

is. However, when acquiring new donors, this is often not possible, since you know little to nothing about

who the person is (yet). Thus, while we’ll talk mostly about variable ask strings or topics that apply to both

fixed and variable ask strings, it’s important to discuss fixed ask strings.

Namely, don’t use them whenever possible. Yes, they are necessary for some acquisition purposes, but the

effort to customize them to even what little you know about a donor is worthwhile. Some tips:

Online donation forms are usually customizable. CDR Fundraising Group estimates that this simple step can

increase your response rate by 50% and your average gift by 40%.

In fact, they’ve posted code for how to

do this in Salsa Labs.

What if you don’t use Salsa Labs? Usually searching for “dynamic ask strings XXname

of giving platformXX” will get you some tips on how to.

But if these tips are Greek to you, you can always take a shortcut: setting up multiple donation forms with

different ask amounts and sending the links to customized segments of your audience. This isn’t ideal, but it

gets you most of the way there. Even if you take a shortcut and have over $100 versus under $100 versions

of your donation form to send, you will be customizing the experience for your online donor a little bit.

Use intelligence from your outside list selects. If you are like many organizations, your outside list selects

will feature a minimum threshold below which you won’t accept donors (often $5 or $10). Chances are you

have tested to get to these amounts -- one list is productive without a threshold, so you haven’t incurred the

cost; another had subpar performance, so you asked for a more select group of donors.

Chances are, your $10+ donors from one list will behave differently from your $5+ donors from other and

from your “anything goes” donors from list number three. Thus, you can use this threshold as a

customization point for your fixed ask, making sure to ask for more from people who give more.

Make sure your ask string testing doesn’t select just one winner. When you test an ask string in acquisition,

there’s a temptation to treat it like a traditional A/B test, where a winner is chosen and rolled out with. Here,

however, you may find that even though the majority of lists performed best with your control ask string,

there were a few lists that had demonstrably better results with your test version. Since different lists have

different donor characteristics, you may get better results by keeping an ask string that better fits those

donors. In other words, it’s best to have different ask strings for different lists when results warrant.

Use modeling to determine your ask. List cooperatives will be only too happy to create models for you.

Chances are, they can do a response model that maximizes response and another that maximizes average

gift. The folly is when both of these groups get the same ask strings when they were set up with different

goals in mind.

http://www.slideshare.net/salsalabs/the-ask-is-right-building-donation-forms-cdr-fundraising

https://github.com/cdrprojects/Salsa-Dynamic-Ask-Strings

016

However, you don’t have to use a co-op or pay a PhD to run a basic model. Simply take the average gifts

from your current donors at acquisition by ZIP code, standardize them (rounding to the nearest five or ten

for fluency), and use that as the basis for your fixed ask strings. After all, there’s no reason you have to treat

90210 as the same as 48208 in Detroit.

Make sure you are using information from multichannel giving when running a conversion program. Sadly,

walkers, event donors, volunteers, online donors, and e-newsletter subscribers are often dropped into an

offline acquisition with nary a thought as to ask string. Please don’t do this. You could be asking your $500

online donor or your gala chair to sign a $20 check, meaning that you’ve drastically downgraded that donor.

Instead, make sure all giving, not just channel-specific giving, is taken into account when formulating your

asks. Additionally, even if someone has not given, you can apply filters like ZIP code or historical data (e.g.,

last time, your volunteers’ average donation was twice that of your e-newsletter subscribers; why not ask

for twice as much?) to your ask string.

Hopefully these tips help make even your fixed ask string more customized. Now, let’s get to the good stuff.

Variable ask strings

Setting your anchor with donor lifecycle

The immediate first question you should have is “variable based on what?” Chances are, if you are like most,

you are using HPC or highest previous contribution. This has been the industry standard since punchcards

were used to calculate what an ask string should be.

But the data would suggest that this is wrong, or at least wrong-ish.

As we mentioned in the anchoring section, people have their own internal anchors for what they traditionally

give your organization. One heuristic that we didn’t talk about previously is called the availability heuristic.

Availability means if you can recall an example of something happening, it must be as or more important

than something that you can’t easily recall happening.

A classic example of this from the literature is people overestimate the number of words that begin with the

letter R. They also underestimate the number of words where R is the third letter.

For readers my age, punchcards were paper cards that were used to store data in the early days of computing. You know how

computers are using binary code -- 1s and 0s -- to interpret the world? A punch card would have hole versus no hole as a way of

showing one versus zero. It’s basically an analog floppy disk.

For my younger readers, a floppy disc encoded your 1s and 0s on a magnetic medium (in the 1980s, you were warned to keep

magnets away from your computer lest you scramble your data) that came on a removable disc that was floppy (in the case of

8” or 5.25” inch models) or not floppy (in the case of 3.5” hard plastic models). Why the non-floppy disc was called a floppy disc,

I have no idea. These discs were put into a disc drive and read by the computer. It was basically a magnetic version of a CD-ROM.

For my even younger readers, a CD-ROM is an optical compact disc made of plastic and covered by aluminum to make them

reflective. Data was encoded using microscopic indentations, kind of like a record (OK, that probably doesn’t help you, unless

you are a vinyl hipster). You would put them in a drive and it would be read by the computer using what Dr. Evil would call

“lasers.”

For my even younger readers, I got nothing for you. Look it up on Snapchat or Instagram or something. And get off my lawn.

Amos Tversky; Daniel Kahneman. Uncertainty: Heuristics and Biases. Science, New Series, Vol. 185, No. 4157. (Sep. 27, 1974),

017

Or, similarly, some people will say there are more six-letter words that end in “ing” than end in “g” (which is

impossible). We can easily recall things that begin with R or end with “ing.” That’s how they are filed in our

brains; thus, we think it happens more often.

This can affect how our causes are seen by the public. Quick: how many people are killed by drunk driving

versus cell phone use while driving?

Got your answer?

In 2013, the last year for which we have data for both causes, drunk driving killed 10,110 people in the United

States.

Cell phone use and driving killed 445 people.

Chances are, if you are like most Americans, you thought these were about equivalent. You almost certainly

did not think the two numbers were more than an order of magnitude different.

Why is that? Because you can look at the car next to you at a stoplight and see the driver is texting. It is far

more difficult for you to look at the car next to you and see that the driver is drunk. And so our availability

heuristic can easily recall cell phone use and driving and that gets moved up in our mental queue.

Incidentally, both are dangerous. If you are reading this while driving, stop now.

So, what is the available anchor for donations to your organization? Chances are that it isn’t the highest

previous contribution.

Early research indicated that anchoring from the most recent contribution is most effective.

But more

recent scholarship indicates that it probably depends on the donor. One study took a look at reference points

by donor lifecycle category. They found that most recent contribution held as the best reference point to

help reacquire lapsed donors -- while it lowered average gift from 3.5 to 3.18 Euros, response rate jumped

from 3.25% to 5.01%.

Since your goal on lapsed campaigns is almost always to reactivate donors, rather

than to increase gift size from a select few, this response rate jump is critical.

And this makes intuitive sense. Let’s say you are using highest previous contribution as your anchor. A person

gave you $10, $20, $50, $25, and $10. You keep asking them for $50, $75, or $100, but they haven’t

responded. This donor is saying fairly clearly with their donations that they don’t value you at the $50 level

any more. Thus, your ask is out of their consideration set and they won’t give. Reverting to most recent

donation -- $10 -- has a greater potential for bringing them back into the organization and gives you another

opportunity to regain their previously higher giving levels.

http://www-nrd.nhtsa.dot.gov/Pubs/812231.pdf

http://www.distraction.gov/downloads/pdfs/Distracted_Driving_2013_Research_note.pdf

Schibrowsky, J. A. and Peltier, J. W. (1995), Decision frames and direct marketing offers: a field study in a fundraising context.

J. Direct Mark., 9: 8–16.

Griet Alice Verhaert , Dirk Van den Poe. Improving Campaign Success Rate by Tailoring Donation Requests along the Donor

Lifecycle. Journal of Interactive Marketing Volume 25, Issue 1, February 2011, Pages 51–63. http://wps-

feb.ugent.be/Papers/wp_10_666.pdf

018

The study also found that asking for the average of someone’s previous gifts is best to retain donors.

Specifically, average gift garnered .38 Euros per piece, versus .36 per for highest previous contribution and

.35 per for most recent contribution.

This is further verified by De Bruyn and Prokopec’s modeling of donor behavior.

They explain this as a

reversion to the mean. Specifically, they find that donors, in order to determine whether an ask is

appropriate or not, use their most recent gift. However, when they decide what to actually give, they tend

toward the average of their previous gifts.

Nowhere in these studies did highest previous contribution win in testing. As donors, our last gift is most

available to us and our average gift is a matter of habit. That highest gift we made may be the stuff of distant

memories or a one-time instance of extra cash.

Yet HPC is the anchoring point used by most nonprofits. Thus, I would try HPC v MRC v average gift as one

of your early ask string tests, with one of the greatest potential impacts on your fundraising.

Setting your anchor with even more specific donor lifecycle information.

Using most recent contribution or average contribution as the basis for your ask string doesn’t necessarily

mean using that actual number, however. There’s some evidence that moving from your chosen reference

point can have a positive impact with some donors.

De Bruyn and Prokopec worked with a large and anonymous European non-profit to mail to their donor list.

They did so with a 3 x 3 matrix of ask strings set by two criteria: 1) is the initial ask below, at, or above their

previous contribution? and 2) is the ask string acceleration steep (20% increases in levels), steeper (50%

increases in levels), or steepest (80% increase in levels). The ask strings were four items long.

This is a bit confusing, but here are initial and final asks for each condition, assuming a $100 donor. You’ll

note they are rounded:

Lower

Equal

Higher

Steep

$85 … $140

$100 … $170

$120 … $200

Steeper

$70 … $230

$100 … $350

$150 … $500

Steepest

$55 … $320

$100 … $580

$180 … $1000

They found, as you might expect, asking for more got more in average donation but suppressed response

rate. But they also found an important difference in lifecycle.

Griet Alice Verhaert , Dirk Van den Poe. Improving Campaign Success Rate by Tailoring Donation Requests along the Donor

Lifecycle. Journal of Interactive Marketing Volume 25, Issue 1, February 2011, Pages 51–63. http://wps-

feb.ugent.be/Papers/wp_10_666.pdf

De Bruyn, Arnaud, Sonja Prokopec (2010), "Optimizing Donations with Individually Tailored Donation Grids: An Econometric

Model of Compliance and Generosity," in Proceedings of the 37th EMAC Conference, Suzanne C. Beckmann and Torsten

Ringberg (Ed.), Copenhagen: European Marketing Academy.

http://www.debruyn.info/research/papers/proceedings/debruynprokopec2010_anchoring.pdf

Arnaud De Bruyn & Sonja Prokopec. Opening a donor's wallet: The influence of appeal scales on likelihood and magnitude of

donation. Journal of Consumer Psychology 23, 4 (2013) 496–502.

http://www.debruyn.info/research/papers/published/debruynprokopec2013_anchoring.pdf

019

Specifically, multi-donors were more set in their ways. When basing the ask off of a higher amount than their

previous contribution was related to a big drop in response rate — from an average of 10.5% among those

who had the ask string that started at equal to 9.1% among those who were asked for higher. It is, not

shockingly, as if the multi donors were saying that they had already told the nonprofit what they give (their

internal reference point).

However, single donors were more receptive to changing their gift amount. Response rates for the single

donors were 5.3% in the lower group, 4.1% in the equal group, and 4.3% in the higher group. Indexed average

gifts were .937 (lower), .909 (equal), and 1.162 (higher). So there was a trough in both response rate and

average gift for asking a single donor for the same thing they gave before.

Here are the gross revenue per piece figures, indexed to a $100 donor to make the math easy:

Single donors

Lower

Equal

Higher

Steep

$4.74

$3.54

$4.23

Steeper

$4.76

$3.96

$5.62

Steepest

$5.49

$3.68

$5.26

Multi-donors

Lower

Equal

Higher

Steep

$10.42

$10.16

$9.96

Steeper

$9.30

$10.44

$9.67

Steepest

$10.46

$10.53

$10.68

Your takeaways from this are largely goal dependent. If your goal is to have more, lesser value donors, you

probably want to ask your single donors to give significantly below their previous gift amount and use that

gift amount as the second number in the ask string. If your goal is to have fewer, higher value donors, you

should ask these single donors for more than their previous gift right off the bat.

And for multi-donors, previous contribution or average contribution look like where the ask string should

start.

Acceleration of ask string and importance of the first ask amount

Another important feature here is that the ask string steepness didn’t affect response rate. Only the first ask

seemed to affect response rate significantly. The lesson here is that you can ask for more and get more

without hurting response. This is backed up by modeling that indicates that the first ask has three times the

impact on response rate and average gift as any other number in the ask string.

De Bruyn, Arnaud, Sonja Prokopec (2010), "Optimizing Donations with Individually Tailored Donation Grids: An Econometric

Model of Compliance and Generosity," in Proceedings of the 37th EMAC Conference, Suzanne C. Beckmann and Torsten

Ringberg (Ed.), Copenhagen: European Marketing Academy.

http://www.debruyn.info/research/papers/proceedings/debruynprokopec2010_anchoring.pdf

020

Other field studies have also validated that a higher first number suppresses response rate (in addition to

increasing average gift), but that using more aggressive multipliers did not have such a negative response

rate impact.

I doubt this can be taken to extremes lest you get to amounts that are outside what someone would consider

-- I haven’t seen anyone go over doubling each amount offline. However, the traditional base amount, base

amount x 50%, base amount x 100%, and other ask string is probably not sufficiently aggressive. There is also

an innovative idea to try to do a double ask string: one normal and one with an ask for $5 more by amount

(in this case, justified to provide one more nutritious meal to someone with HIV/AIDS). This increased

average gift with no loss in response rate in a limited, non-peer reviewed test. You can see a chart of this in

the presentation in the footnote.

This also has implications for ask order. While I could find no good studies on whether ask strings should be

ascending or descending, studies would tend to indicate that higher amounts first will decrease response

rate and increase average gift, and the reverse for the reverse. (That said, there is a test from Lautman

presented at the 2011 Bridge Conference where, in the mail, a descending ask string increased both average

gift and response rate.

However, because there wasn’t the usual tests of significance and audience size

and such displayed, I would say this is a “test” rather than a “roll-out with” idea.)

For online, there are also no good studies I could find as to whether horizontal, vertical, or hybrid ask strings

work best. My personal experience with this is that it is highly organization, ask, and form dependent, so a

quick A/B test should give you an indication of direction here.

How many options should you give?

This is an area not extensively researched. On the one hand, especially on things like Web forms where a

potential $5 and $5,000 donor may meet alike, many different options allows everyone to find an amount

that they would consider. On the other hand, too much choice can be overwhelming.

There is an interesting book on the topic by Barry Schwartz called The Paradox of Choice. (There was also a

TED talk on the topic; ironically, it is one of literally thousands of TED talks you could choose to watch.) You

would think that more choices are better -- a rational actor should welcome any additional choice as

increasing utility. But if you think we are rational actors, I suggest going back and re-reading the section on

psychology.

Sonja Prokopec and Arnaud De Bruyn (2010) ,"The Impact of Anchors on Donors’ Behavior: a Field Experiment", in NA -

Advances in Consumer Research Volume 37, eds. Margaret C. Campbell, Jeff Inman, and Rik Pieters, Duluth, MN : Association for

Consumer Research, Pages: 524-525 .

http://www.lautmandc.com/Uploads/file/Presentations/Tests-that-are-making-a-difference.pdf, slide 27.

http://www.lautmandc.com/Uploads/file/Presentations/Tests-that-are-making-a-difference.pdf, slide 25.

021

One particularly memorable study that he cites was about selling flavors of jam. From the book:

“When researchers set up a display featuring a line of exotic, high-quality jams,

customers who came by could taste samples, and they were given a coupon

for a dollar off if they bought a jar. In one condition of the study, 6 varieties of

the jam were available for tasting. In another, 24 varieties were available. In

either case, the entire set of 24 varieties was available for purchase. The large

array of jams attracted more people to the table than the small array, though

in both cases people tasted about the same number of jams on average. When

it came to buying, however, a huge difference became evident. Thirty percent

of the people exposed to the small array of jams actually bought a jar; only 3

percent of those exposed to the large array of jams did so.”

Clearly, the people who had 24 options could have found a flavor they liked. The six the others had from

which to choose were among their options, plus 18 others that may have been better. They were simply

overwhelmed by choices and fluency was degraded. The poster child for this, in my mind at least, is Jeremy

Renner in the end of The Hurt Locker, dazed by the variety of cereal on display.

But there’s a reason that I didn’t mention the paradox of choice in the fluency section: it might not exist and

almost certainly doesn’t exist in all cases.

Three researchers did a metastudy of choice overload research, looking at 62 results from 50 studies.

the end, they concluded that the average result was zero. Normally, a zero result in a metastudy comes from

many studies with zero effect. However, here, there were many studies with significant results: it’s just that

there were about equal numbers of cases where more choices were good as when more choices were bad.

And the researchers could not figure out any pattern in when choices helped or hurt. As they said in the

study:

“In summary, we could identify a number of potentially important

preconditions for choice overload to occur, but on the basis of the data on

hand, we could not reliably identify sufficient conditions that explain when and

why an increase in assortment size will decrease satisfaction, preference

strength, or the motivation to choose. This might account for why some

researchers have repeatedly failed to replicate the results of earlier studies

that reported such effects.”

So, there is not a lot of evidence as to whether more or fewer choices are better in ask strings, as most of

the studies take the 3-4 options plus an “other” category familiar to most nonprofits.

There is, however, some evidence that we may be artificially limiting our options. This is especially true in

acquisition, where you don’t know the giving propensities of the donor.

Scheibehenne, B, R Greifeneder, and P Todd. Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload.

Journal of Consumer Research (October 2010), Vol. 37.

http://www.scheibehenne.de/ScheibehenneGreifenederTodd2010.pdf

Scheibehenne, B, R Greifeneder, and P Todd. Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload.

Journal of Consumer Research (October 2010), Vol. 37.

http://www.scheibehenne.de/ScheibehenneGreifenederTodd2010.pdf

022

Fluent took at look at the ask strings of seven 2016 presidential candidates online.

Elections are the mayfly

equivalent of nonprofits. Because they live for a day (comparatively), their testing is the espresso to our

coffee -- distilled into a shorter period of time, swiftly copycatting other campaign moves, and able to draw

on consultants that are usually doing many campaigns simultaneously (or at least networking with those

who are). As a result, they can be potent indicators and guides for us on the nonprofit world.

The study found that all seven of the presidential candidates had either six or seven options in their ask

strings.

This allowed them to go both lower (the lowest ask amount for each was $25 or under, with four

candidates at $10 or under) and higher (three candidates’ strings ended at $2700 (the federal limit for one

person to donate) or higher) than most nonprofits. Since people of all means are coming to the site, this

provides options to both the well- and the barely heeled.

Going low may have its own special benefits, approximating a technique called validating small gifts. While

tested mostly in face-to-face fundraising, the researchers have found in several studies that rather than

talking about a million dollars, you can have success by talking about a penny – specifically, the phrase “even

a penny would help.”

This technique has some impressive results. One study of this in a face-to-face environment increased giving

from 28% to 50%.

One could argue that the success of March of Dimes in their original launch was in part a variant of this.

Although a dime meant much more then, it still was a way of giving permission to lower level gifts.

The evidence is that people actually give significantly more than a penny. While gift did go down on average,

the total revenue from the canvass went up 64% because of the increase in response rate.

Since revenue per communication is usually a pretty good way of measuring its success (in an ideal world,

you’d want to measure its impact on lifetime value, but on a one-year time horizon, you go with what you

have), I would call this a win.

The danger would be a negative anchoring effect, because the first number in an ask string is the most

important. However, this can be countered (for current donors) by reminding them of what their previous

gift was, according to one study.

Additionally, suggesting higher default amounts can also counteract negative average gift effects.

Setting defaults

With online ask strings, there is the opportunity to set a default -- that is, what radio button amount will be

automatically set, if the user were to make no changes? In print, this can be accomplished with social proof

(e.g., “most people give X”) or by circling a desired amount.

http://www.fluentco.com/donation-options-offer-a-glimpse-into-campaign-finances/

Carson, Clinton, Cruz, Kasich, Rubio, Sanders, and Trump.

Goswami I and O Urminsky. When Should the Ask Be a Nudge? The Effect of Default Amounts on Charitable Donations.

Journal of Marketing Research, Forthcoming. http://home.uchicago.edu/ourminsky/Charity_Default_Goswami_Urminsky.pdf

023

Do these defaults help? Evidence indicates that they do, increasing revenue. Specifically, social proof can be

a powerful default. One study used callers to a public broadcasting pledge drive. When people called in, they

were informed that a previous donor had given $75, $180, or $300. Citing a $300 donation increased

renewing members’ donations by 29% on average.

The study also looked at social information in both mail

and phone for renewing members. When a person was given information that someone had given a larger

donation, their gift increased by more than $12. When they received information that someone had given a

smaller donation than their previous gift, their gift decreased by $24. (Equal information increased by $5).

In point of fact, the closer the in-group is (e.g., “most Wisconsites give X” versus “most people give X”), the

more powerful the social proof is.

Looking at the $300 donation study again, the researchers replicated

this, but changed the sex of the other donor either to match or mismatch with the caller. When the gender

of the person who gave the previous gift matched the caller, there was a 34 percent increase in giving

compared with a mismatched sex for the alleged other giver.

And a test presented by M+R at the 2013

Blackbaud Conference found that adding “Most PETA supporters give at this level” to an ask string increased

revenues by 25%.

This, of course, only works for PETA -- you’d want to swap in the name of your own

organization, especially if you are the National Pork Producers Council or the like.

Take care with this tactic, though. Evidence shows that while social proof can help significantly increase

donations from lapsed donors, it also significantly lowers response rate.

Since the focus with lapsed donors

is usually to bring them back into the organization (although your goals may vary), you may not want to take

that approach.

Non-social defaults can also help. This seemed to be especially true among older donors, more frequent

donors, and previous year donors.

Specifically, researchers found that a low bar, as with the “even a penny

will help” approach, helped significantly increase donations.

It should also be noted that recent evidence suggests that if someone has a positive experience with a

default, they will tend to default more and more.

While not specific to the nonprofit context, you can see

how this would happen -- if the default on your Web page is $50, donors who donate more than once could

get stuck at that ask amount over time.

Jen Shang and Rachel Croson, “Field Experiments in Charitable Contribution: The Impact of Social Influence on the Voluntary

Provision of Public Goods,” the Economic Journal.

http://rcgd.isr.umich.edu/seminars/Fall2006/Croson/Shang%20Croson%20EJ%20(2).pdf

Dorina Hynsenbelli, Enrico Rubaltelli, and Rino Rumiati, “Other’s Opinions Count, but Not All of Them: Anchoring to Ingroup

Versus Outgroup Members’ Behavior in Charitable Giving,”Judgment and Decision Making, Vol. 8, No. 6, November 2013, pp.

678–690. http://journal.sjdm.org/12/121219/jdm121219.pdf

Jen Shang, Americus Reed, and Rachel Croson, “Identity Congruency Effects on Donations,” the Journal of Marketing Research,

vol. 45, no. 3, 2008, 35–361.

http://www.mrss.com/lab/60tests60min-bbcon/

Kristoffer Jackson, “The Effect of Social Information on Giving from Lapsed Donors: Evidence from a Field Experiment.”

VOLUNTAS: International Journal of Voluntary and Nonprofit Organizations. April 2016, Volume 27, Issue 2, pp 920-940.

Goswami I and O Urminsky. When Should the Ask Be a Nudge? The Effect of Default Amounts on Charitable Donations. Journal

of Marketing Research, Forthcoming.

http://home.uchicago.edu/ourminsky/Charity_Default_Goswami_Urminsky.pdf

Goswami I and O Urminsky. When Should the Ask Be a Nudge? The Effect of Default Amounts on Charitable Donations. Journal

of Marketing Research, Forthcoming.

http://home.uchicago.edu/ourminsky/Charity_Default_Goswami_Urminsky.pdf

Working paper at

http://blogg.nhh.no/thechoicelab/wp-content/uploads/2016/07/Good-Nudge-Lullaby-working-paper-July-2016.pdf.

024

Rounding your ask amounts

Rounding can be controversial. On the one hand, round numbers could potentially help with fluency.

Rounding out ask strings can help you get out of weird numbers that consistent upgrading can create (e.g.,

if you donate $30, then upgrade by 50% each time, that’s $45, $67.50, $101.25, then $151.875. And if you

haven’t got a ha’penny, then God bless you).

On the other hand, as we discussed in cultural constructs, if someone is giving $18 or $88 for a reason, it also

rounds that out. Additionally, there is a potential draw in using an odd amount specifically to stop someone

short (e.g., your $17 will feed X people). In fact, in face-to-face settings, panhandlers found that when they

asked for change, 44% of people contributed. When they asked for a quarter, 64% contributed. When they

asked for 17 cents or 37 cents, 75% of people made a contribution.

This is called the pique technique; the

idea is that the odd request breaks people out of their normal mental structures, forcing them to think about

what you are saying.

However, this may or may not be as applicable in non-face-to-face environments. Burger et al took a look at

the mechanism by which this worked. They found that contributions only increased among people who came

over to ask a question; there was no difference in giving between people who were given a specific answer

(e.g., “I need to buy a stamp”) versus those who were given a generic answer (e.g., “I need to buy some

stuff”).

Since there isn’t a mechanism for someone to stop and ask you a question in the mail, this

technique may not work on asynchronous platforms. And, in fact, a study of the pique technique when

applied to causes, rather than donating directly to the person face-to-face, found no significant difference

with this technique versus not using the technique.

There is, however, evidence that rounded numbers can increase giving. One study of donations found that

rounded values increase giving by seven percent. Specifically, they found that people were more likely to

choose things that were on the ask string than rounded numbers not on the ask string, but that a good

number of people wrote rounded numbers in as the other when not on the ask string. Additionally, they

found when a round number was on the ask string, there was a particularly strong pull of that number on

donations.

In addition to articulating that round numbers have a pull that is independent of the pull of a person’s

internal reference point and the ask string itself, they also helped define a conundrum. To wit: what is a

round number? As the authors put it:

Santos, M., Leve C., & Pratkanis, A. (1994). Hey buddy, can you spare seventeen cents? Mindful persuasion and the pique

technique. Journal of Applied Social Psychology, 29, 755-764.

Burger J, et al. The Pique Technique: Overcoming Mindlessness or Shifting Heuristics?. Journal of Applied Social Psychology,

2007, 37, 9, pp. 2086–2096.

https://www.scu.edu/media/college-of-arts-and-sciences/psychology/documents/Pique-Technique.pdf

Thomas Hugh Feeley, Sarah Dietrich, & Jennifer Musone. Replicating the Pique Technique of Compliance-Gaining. December

2014.

http://www.buffalo.edu/content/cas/communication/faculty/feeley/_jcr_content/par/switch/fourth/download_1/file.res/THFe

eley_Pique_Technique.pdf

Pierre Desmet &, Fred M. Feinberg. Ask and ye shall receive: The effect of the appeals scale on consumers donation behavior.

Journal of Economic Psychology 24 (2003) 349–376.

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.399.811&rep=rep1&type=pdf

025

“While the notion of a round quantity is seemingly intuitive, it is nonetheless

difficult to make precise. Round numbers are operationalized in the present

paper as the face values of commonly used French currency notes or small

integral multiples thereof. Based on the data used here, this functional

definition accounts for all but a negligible proportion of off-scale donation

amounts, in the sense that adding or removing additional rounded values does

not substantially alter any aspect of the analysis.”

These banknote values are especially strong -- $100 in particular. Reiley and Samek did a study in which they

tried two different ask strings.

The first was $35, $50, $75, $95, $250, and other. The second was the same

as the first, but with $100 swapped in for the $95.

With the $100 ask, revenue per solicitation went up 29%. Average gift went up almost 20% and response

rate went up 7.6%. It’s rare that the ask string with the higher gift amount in it also had a higher response

rate, so $100 is something of a magic number in ask strings.

But it’s not the only magic number. The effect seems to hold with other banknote numbers. Another study

looked at whether people were more likely to give $20 or a strange amount like $20.03 (if they graduated

from the college in question in 2003). People were less likely to give the strange amount (although this was

not statistically significant at the .05 level).

In summary, fluency seems to trump creativity in the case for round numbers, especially at higher levels. If

you are looking to increase average gift, it should be to something that is highly fluent like $20, $50, or $100;

if you have to be creative, do it with lower ask amounts so that they don’t have as much draw. (While it

hasn’t been studied that I know of, I would argue this fluency effect is greater offline than online -- making

a $53.74 donation online is nearly as easy as a $50 one, but when one is writing out a check, it become far

more difficult.)

There is some evidence that one oddball (aka disfluent), very high donation amount can be helpful. Make-a-

Wish Canada has an ask string in their control acquisition letter that goes $25, $50, $100, $250, $1000,

$6,518 (the average cost of one wish), or other. SOFII interviewed the writer of the piece, Harvey McKinnon;

he had this to say about this amount:

Pierre Desmet &, Fred M. Feinberg. Ask and ye shall receive: The effect of the appeals scale on consumers donation behavior.

Journal of Economic Psychology 24 (2003) 349–376.

http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.399.811&rep=rep1&type=pdf

David Reiley and Anya Samek. How do Suggested Donations Affect Charitable Gifts? Evidence from a Field Experiment in

Public Broadcasting. Working Paper No.: 142- SPI, November 2015.

http://spihub.org/site/resource_files/publications/spi_wp_142_samek.pdf

James T. Edwards, John A. List.Toward an Understanding of why Suggestions Work in Charitable Fundraising: Theory and

Evidence from a Natural Field Experiment. NBER Working Paper No. 19665. Issued in November 2013.

http://www.nber.org/papers/w19665.pdf

026

“There is, however, one twist: there is an option to donate a sum of $6,518.

We put that figure in because it is the actual average cost of granting a wish.

Every now and then, when I’ve done that before, you find a donor who is

willing to donate at that level. We did this once for a hospital when the price

point for a piece of equipment was $6,942.73. Thirteen people “bought” this

device. These donors upgraded from an average of $65 to nearly $7,000. It

never hurts to ask.”

As we’ve seen, there’s only one potential inaccuracy in this: sometimes it does hurt to ask.

But the idea of providing an outlier as a potential anchor and/or moneymaker is an interesting one. While I

could find no data on this from the nonprofit world, there are studies from the for-profit world that

indicate that the existence of one very high product can increase what someone is willing to spend on a

medium-priced product.

This explains the existence of those news stories you hear from time to time

about how someone has created an even more expensive hamburger, made with wagyu beef, caviar, gold

leaf, and unicorn tears. Having this burger on a menu could also increase average spending in addition to

the free publicity.

The future of ask strings

What has been described so far is largely tweaks on a formula: base amount, base amount times something,

base amount times something larger, other. Check a box. Send your gift.

This is because this is largely what has been tested. When one delves into the research on a topic, one is

limited by the research that exists.

It need not always be so, however. There are techniques that have the potential to break out of this familiar

formula that are less tested, but have promise.

Modeling

There are some new studies that indicate that we are heading for an even-more-customized world of ask

strings. An econometric look at donation behavior has been used to individualize ask strings, raising average

gift by 22% and overall revenues by 36%.

And predictive modeling has been helpful in creation of some ask

strings in scientific settings.

It could be argued that all of the previous techniques are outdated as a result. However, any good model will

need grist for the mill and the optimizations and tests suggest previously can provide the experimentation

to get the most out of any good model.

Aradhna Krishna et al. Effects of Extreme-Priced Products on Consumer Reservation Prices. Journal of Consumer Psychology,

16(2), 176–190. http://www.bm.ust.hk/mark/staff/Rashmi/Rashmi%20-%20JCP%20-%202006.pdf

De Bruyn, Arnaud, Sonja Prokopec (2010), "Optimizing Donations with Individually Tailored Donation Grids: An Econometric

Model of Compliance and Generosity," in Proceedings of the 37th EMAC Conference, Suzanne C. Beckmann and Torsten

Ringberg (Ed.), Copenhagen: European Marketing Academy.

http://www.debruyn.info/research/papers/proceedings/debruynprokopec2010_anchoring.pdf

Lee, KY and F Feinberg. Modeling Scale Attraction Effects: An Application to Charitable Donations and Optimal Laddering. Ross

School of Business Working Paper Working Paper No. 1202, June 2013.

https://deepblue.lib.umich.edu/bitstream/handle/2027.42/100183/1202_Feinberg.pdf

027

Modeling need not be intensive to be effective. I’ve seen positive results from a simple model that groups

potential donors (that is, in an acquisition mailing) into high and low in terms of average gift. From there,

the “high” piece is customized to maximize average gift with a higher ask; the “low” piece is customized to

maximize response rate.

Online ask strings

For the most part, for online ask strings, we’ve taken the old familiar mail ask string from a mail piece, stuck

it on a Web page, and called it a day. Yes, there is usually a bit more text and the default ask is likely larger

than your default acquisition ask in the mail, but it’s largely swapping out the check box for a radio button.

This need not be the way. In the mail, defaults are often circled or have arrows. These certainly exist online,

but there are also opportunities for subtle (or unsubtle) changes in color, font size, or bolding.

Online also has a strong opportunity for upselling with the time between donation and processing. A person

has selected an amount, given their credit card information, and decided to make the donation. But the

series of tubes has not yet processed the credit card.

This could be an upgrade in amount, like the ask for $5 more to provide one more nutritious meal to someone

with HIV/AIDS. But I think this is a better opportunity to upgrade someone in kind to a monthly donor. This

is a person who wants to help and, as long as the additional impact that can be had is explained and the ask

is easily avoided, a monthly giving ask will likely convert some people. Unfortunately, it’s not well tested.

Finally, as we’ve seen, online has particular challenges because donation forms tend to be the same for

everyone, regardless of whether someone has given before or how much they are capable of and desiring

to give. However, it also holds potential solutions for these as well. Savvy for-profit marketers are already

able to customize content and price based on just the variables like browser type.

We should be able to do likewise with modeling. We should be asking people browsing on the newest shiny

tech toy from posh enclaves for more than, say, me. And modeling can turn up things that you may not think

of. Political campaigns are finding that those with AOL.com email addresses are the most valuable members

of their list

; we should be able to deploy our asks with similar knowledge.

No ask string

After all we’ve talked about for ask strings, surely I’m not recommending doing away with them?

I’m recommending you test it. And don’t call me Shirley.

NextAfter did a test of this with Harvest Ministries online. They tested a lower ask string versus a higher ask

string for their default donation page on their site. As almost an afterthought, they also tested eliminating

the ask string altogether and simply letting the potential donor choose. This then went out in an email that

coordinated with the online form (low ask, high ask, and no ask string).

http://www.fluentco.com/aol-email-addresses-worth-more-to-campaigns/

028

The results were astounding. Conversion rate went up 43% (from .1% to .14%).

As a result, the string-less

form became the new control. This may more effective online than offline, as the ask string could be an

artifact of mail systems used by those who view email as direct mail plus electrons.

That said, it may be that its time has come offline as well. Many of the impacts of an ask string can be

accomplished by copy. With a note that highlights a person’s previous gift (or the average gift that donors

are making or the like), you can set up effective anchors. Same for social proof.

This may also allow for better use of techniques that come from face-to-face fundraising, like legitimizing

small gifts (aka “even a dollar (or penny) would help”). These techniques work well in the flesh with the

phrase “even a penny would help” increasing giving from 28% to 50%.

The phrase fails Kant’s categorical imperative: if everyone did it, it actually would not help. Your penny would

be eaten up by credit card fees, postage, and acknowledgments. I have not used this technique extensively

because I’ve been worried about the anchoring effect. My concern has been that while response rate may

go up, average gift would plummet and, as a result, we’d have more lower-value donors instead of fewer

high-value donors. The former can be a strategy, but isn’t the one I’ve traditionally aimed for.

However, results from online and mail programs of this technique have been decidedly mixed. Part of this

may be a disconnect between the message in the body text, which says any gift would help, and the ask

string, which starts out at a price that definitely isn’t a penny.

Even in the modeling examples above, the goal is to guess as close to the maximum amount that a person

would be willing to give. When your guess is below that, you may leave money on the table; more and the

models show you will lower response rate. Thus, letting someone choose their own gift may help them

pick the right gift for them.

Final notes

This has been an attempt to catalog the research on ask strings and methods. The solutions range from the

tweak to the radical. But in the end, we are limited in one part by the research that has been done and is

public. In the other part, we are limited only by our tolerance for risk and our imaginations.

It’s my hope that we’ll be able hear stories from the field, reinforcing or disputing what’s written here,

pushing the envelope ever forward. I would love to add what you know (studies I’ve missed, or studies as

yet undone) to this body of knowledge in The Science of Ask Strings 2.

Until then, thank you for reading – hope it helps!

https://www.nextafter.com/research/2016/05/how-ask-amount-affects-donor-conversion-in-an-email-fundraising-appeal/

Cialdini, R. B. & Schroeder, D. A. (1976). Increasing compliance by legitimizing paltry contributions: When even a penny helps.

Journal of Personality and Social Psychology, 34, 599–604.

029

DonorVoice

The Experience and Relationship Company

US Contacts:

Kevin Schulman, Founder and Managing Partner

[email protected]

Josh Whichard, Partner

[email protected]

Nick Ellinger, VP of Marketing Strategy

[email protected]

UK Contact:

Charlie , Hulme, Managing Director

[email protected]

Phone: 202-246-9649