ask strings

The science of ask strings

nickellinger, , , , Leave a comment

Today’s direct marketing paper says, in essence, the less you ask for, the more people respond and the less they give.  Duh.

But there are some great surprises in the paper that make it well worth exploration.

De Bruyn and Prokopec took a look at anchoring effects in ask strings.  Specifically, they 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 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 appropriately 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

Some of these may look to you as they looked to me — fairly aggressive.  In the higher steepest condition, you are asking your $100 donor to donate $180, $320, $580, or $1000 — not a common ask string by any means.  That’s why I’m glad there are studies like these that test this with other people’s money.

As I mentioned, they found asking for more got more in average donation but suppressed response rate.  However, there were several other elaborations on this:

  • Ask string steepness didn’t affect response rate. Only the lowest, left-most 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 potentially free money.
  • Steepness did increase average gift.  So 80% increases won in this case.
  • Multi-donors were more set in their ways. Indexing off of higher than their previous contribution was related to a big drop — 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 and don’t forget it.
  • The worst thing you could do was ask single donors for what they gave before.  This surprised me.  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.

They didn’t give the net revenue per piece charts in the study; I found them invaluable in understanding the implications.  These are 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

All this indicates something to me that I hadn’t thought of before (and maybe you have and have tested it — if so, please put it in the comments or email me at nick@directtodonor.com so we can have a report from the trenches): different ask strings for single versus multi-donors.

The hypothesis that I would form based on these results is that people who have given before are set in their ways of what they want to give and thus we should index from the previous contribution or the HPC.  Single donors are more pliable, so we can work to get more value out of them early in the relationship, elevating their support before they get set in their ways.

science-pinkman

Hope this has been as valuable for you as it has been for me.

The science of ask strings

How to structure your matching gift campaign

nickellinger, , , , , Leave a comment

Matching gift campaigns work. But are they necessary?

Whether it’s a grantor’s challenge fund, a campaign match, a fund set up by a generous donor or donors, matching gifts are a frequently used and frequently successful tactic.  Most of the time, it’s set up as a “double your impact” campaign.

Three researchers — Huck, Rasul, and Shepard — looked at whether a lead donor increased the success of a campaign and how the structure of the match impacted that success.

They did this for the Bavarian State Opera House.  (BTW, if you are a researcher and want to run a test with donors on your dime, email me at nick@directtodonor.com; I’m usually game.)

300px-mc3bcnchen_nationaltheater_interior

The Bavarian State Opera House.
Fundraising motto: hey, these inlays don’t gild themselves.

Here were the six test treatments:

  1. Control: No lead donor, no match commitment
  2. Lead donor: A generous donor has already funded part of the program for 60,000 Euros (remember, Bavarian State Opera House).  We need your help with the other part.
  3. 50% match: A generous donor will match your Euro with .5 Euro.
  4. 100% match: Euro for Euro match
  5. Non-linear matching: A generous donor will match any gift made over and above 50 Euros.  (Which is to say if you donate 120 Euros, the donor gives 70.  If you donate 70, the donor gives 20.  If you donate 40, the donor gives nothing)
  6. Fixed gift matching: A generous donor will match any positive gift made with 20 Euros of his* own.

Got your guesses of what will do what?  Good — here we go with the results**:

Response rate Average gift Revenue per piece
Control 3.7% 74.3 2.79
Lead donor 3.5% 132 4.62
50% match 4.2% 101 4.19
100% match 4.2% 92.3 3.84
Non-linear match 4.3% 97.9 4.18
Fixed gift match 4.7% 69.2 3.27

Yeah, not what I thought either.  I figured, from all of the virtual ink I spilled in social proof and authority last week, that the presence of a lead donor would help. Presumably, there was another mechanism in place — that of anchoring.  I’ll dedicate a full post or five to anchoring effects at some point, but now, suffice it to say that by throwing out the number of 60,000 Euros you can trigger the idea that a person’s gift should be closer to that number.  For some, that may turn them off (although the decline in response rate wasn’t statistically significant).

What surprised me was that the matches didn’t help revenue per piece (unless of course the match is generating marginal revenue).  The matches increase response rates, but the average gift was significantly lower in all of the matches.  The authors’ hypothesis is that the match has a bit of a crowding out effect — that is, the donor feels like their 50 Euros is actually 100 Euros, so they need not make the donation of 100 Euros to have the impact they wanted to have.  This is certainly plausible and consistent with previous research.

What to make of this? Like, I’m guessing, many of you, I’d only tested matching gift language versus control language. However, there is some evidence here that simply stating that a lead gift has been made can increase the anchoring effect and support the idea that a program is worth funding without potential negative byproducts of crowding out donations.

That’s for the general case.  You might also take a look at a fixed gift match depending on your goal.  Generally, I prefer quality of donors to quantity.  However, if you were running a campaign like lapsed reactivation, you might legitimately want to maximize your response rate at the expense of short-term net revenue.

Based on this, I’m going to be looking at testing this against our typical matching gift campaign.  If you do likewise, please let me know at nick@directtodonor.com or in the comments below.  It would be great to see additional evidence on this.

*  The gendering is from the original — not my own.

** The rounding is in the original paper and throws off the revenue per piece variable a bit, but I chose to stick with what they had in the original paper.

How to structure your matching gift campaign

Ask strings (a.k.a. why you have $18 donations)

nickellinger, , Leave a comment

There is no platonic ideal for an ask string.  It varies organization to organization, piece to piece.  Testing is really is the only way to determine what asks work best for your organization.

Nor is it a place where asks translate well by medium.  Generally speaking, an average online gift can be 2-3 times higher than an average mail gift (why, you ask, would you then bother with mail?  Because retention rates on online giving are much, much lower than mail or telemarketing).

There are two types of ask strings – one fixed, when you don’t know who is going to give what, and one variable, when you have a history with a donor.

Tests that work for both

  • Generally, you want at least three options.  Probably no more than seven.  In there is a lot of room for testing.
  • You will more often see ask strings ascend, but testing descending can frequently add value.  Sometimes having them somewhat out of order works best – I’ve seen controls that ask for $15, $10, $25, or $50.
  • No dollar sign. There is some psychological evidence that seeing signs of money actually make people thriftier (hence why they are omitted on pricey restaurant menus). Try without
  • What do you call your other category? Some say an indicator like “Your most generous gift:” or “Your best gift:” strikes a better chord than “Other:” (How could it not?)
  • Circle the goal. Often, you will circle an amount that isn’t your highest, but isn’t the lowest, with a note along the lines of “Your $X gift today would really help those incontinent badgers you love so much!”  The idea is to try to upgrade the donor or get a higher gift amount in acquisition.  Social proof – adding a line like “most people in [your city] give this amount – can also help.  This helps a donor classify your organization and avoids awkwardness like the question of how much you should tip your Uber driver.  (Seriously, how much should you tip your Uber driver?  Please leave comments in the notes, as it’s one thing I can’t effectively test.)

Fixed ask string tests

  • Mission tie-ins and odd amounts. Why ask for $15 when $17 is what you really need to help buy art supplies for an inner city youth?  This compelling why also creates a strong tie between the letter and the response device.
  • Social/cultural tie-ins. If you are focusing on a particular audience or demographic, some numbers have important tie-ins.  Some Jewish people believe in giving in multiples of $18 (Chai in Hebrew means living; the letters of chai add up to 18, so “giving chai” is considered to be especially fortunate).  In Chinese, eight is especially lucky and 88 more so, as the number eight is pronounced similarly to prosper or wealth.  Eight is used like nine is in the US in prices (e.g., $1.88 versus $1.99 – they really just both mean $2).  Conversely, four is unlucky in both Chinese and Japanese, because it is homophonous to the word for death in both languages.
  • Where to start? A good rule of thumb is to start your ask string around where your current gifts are, but you may find that you test up from there.  Another tip is to look at the list you are renting – if you are getting donors $20+, you wouldn’t want to start your ask string any lower than $20.  If it’s $5+, you may (or may not) go that low.
  • Do you have to use these? If you can make your acquisition purchases in segments (e.g., give me all of your $10+ donors and all of your $50+ donors), you have dual benefits – you can use the $50+ donors as multis and you can start one ask string at $10 and the other at $50.

Variable ask strings

  • Highest previous contribution (HPC) v. the previous contribution. Let’s say you gave $50 to a nonprofit, then $25.  Should they next ask you for $50, $75, or $100?  Or $25, $37, and $50?
  • Multiplier: HPC, 1.5 HPC, and 2 HPC is probably most common, but I’ve seen HPC, 2 HPC, 4 HPC, and 8 HPC be effective (on a matching gift campaign in particular, where the idea of doubling is already planted)
  • To round or not to round? If someone make an odd gift of $27, should you next ask them for $27, $40.50, and $54?  Generally not; cents don’t make sense.  $27, $41, $54?  $30, $45, $60?  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, if someone is giving $18 or $88 for a reason, it also rounds that out.

As with many things, testing this online is far easier and cheaper.  But testing online is a different blog post for a different time.

Has anyone else found success with an innovative ask string?  Please post in the comments or email me at nick@directtodonor.com.

Ask strings (a.k.a. why you have $18 donations)