Demographics: The Second-Best Way to Segment Your File
Yes, demographics are the second-best way to segment.
The best way, however, is literally almost any other way.
Take, for example, the experience of Todd Yellin, Netflix’s VP of Product Innovation. Netflix has one of the great treasure troves of data out there. What does he use? Quote: “There’s a mountain of data that we have at our disposal. That mountain is composed of two things. Garbage is 99 percent of that mountain. Gold is one percent… . Geography, age, and gender? We put that in the garbage heap. Where you live is not that important.” (Hence inspiring an earlier DonorVoice post Demographics are Garbage.)
Consider romantic comedies. Demographics and stereotyping would say to market these exclusively to women. This misses men who like romantic comedies (folks like Josh Whichard over at DonorVoice) and women who hate them (my wife). Netflix, wisely, markets romantic comedies to people who watch romantic comedies.
As should you. There’s little difference between this and the broad stroke segmentation practiced at some nonprofits where men get address labels with anchors and women get labels adorned with flowers.
The reason these segments don’t work out is surprisingly simple: there’s more difference within these demographic groups than between/among them. Consequently, they are not predictive.
As marketing professor Mark Ritson puts it in Marketing Week: “If your segment is populated by different people who want different things, it is not a segment. It’s a joke and so are your skills as a marketer.”
The reason is because there’s no way you can effectively change your messaging based on that type of segmentation, which is the point of doing segmentation in the first place.
AND…the chaos within demographic clusters isn’t solved by getting ever more granular. You could make your head hurt by trying to figure out the difference among “Red, White, and Blues,” “Heartlanders,” and “Blue Highways” (three real segments from PRIZM social groups). The truth is that even if I’m the same age, sex, gender, sexual orientation, etc., as my neighbor, if I’m caring for a loved one with Alzheimer’s and he isn’t, the way we react to an Alzheimer’s charity will be entirely different.
This most vital of differences will also be entirely invisible to demographics.
Even broad stroke attitudinal segments outperform demographics ones. NonProfit Tech for Good’s Data for Good looked at differences in giving behavior among groups segmented by gender, age, and ideology. The greatest difference between men and women in giving behavior was 20%. The greatest difference between Millennials and Boomers in giving behaviors was 32%. But the difference between liberals and conservatives was 300%.
Ideology isn’t even a great predictor as to which organizations a person will give. But it’s still an order of magnitude more predictive than demographics.
Thus, there are two keys here as we work to avoid a segmentation that is dependent on the undependable.
1. Don’t start with the person – start with the person’s need. A sure sign that an organization knows what they are doing with their segmentations is that their segments have names that describe an identity and/or a behavior. For a nature organization, it could be “outdoorspeople” versus “armchair enthusiasts.” Quick! To whom are you marketing the nature hike, field glasses, and opportunity to tour a nature preserve? Right! And you also know to whom to send the email about the documentary that features your work on the NatGeo channel or the book about environmental issues.
Personally, I’m in the latter group. I like nature – I just don’t want get any of it on me.
We’ll talk about how to brainstorm potential identities later, because they are all organization-specific. For example, a relief organization may have a segment of their file who are living vicariously through the heroic volunteers rather than reacting to usual messaging of the plight of those in need. That’s a segment that meets our earlier criteria – it requires different messaging from the rest.
Demographic segments don’t do this. The ways that you would differentiate male and female messaging are minor copy tweaks at best, and stereotypical self-parody at worst. Same for age, as we’ll discuss in tomorrow’s post on Millennials.
Segments should have global meaning. Basing them on donor needs accomplishes that. Basing them on demographics doesn’t. (See Roger’s post Milkshake Mistakes for more on this.)
2. Start with the end in mind. I recently saw the following statement from a nonprofit professional: “we have done our segmentation. We’re now working on how we use our donor knowledge.” Huh????
In the course of your nonprofit career, impressive firms with marble floors in their HQ will offer you the “opportunity” to do an expensive segmentation with no clear plan or understanding of how it is going to be applied. What they are really offering is the opportunity to pay for the marble floor in their new satellite office.
This is completely backwards. If you don’t know how you are going to apply what you purchase in market, don’t spend the money.
You start with hypotheses based on your knowledge of your donors (and your frequent donor surveys – you are doing those, right?). You gain additional knowledge specific to each donor and how they cluster. Then, and only then, do you segment based on that knowledge.
So if anyone says they’ve done their segmentation and now they will figure out what it’s good for, pity that person as someone whose mission far less flush with cash and no more flush with knowledge.
“Demographic solutions” are rife with this type of segmentation shamanism. In the demographic of “age” alone, there’s a Strauss-Howe generational theory — that every 20 or so years, there is a new generation. They further posit that there are four generational patterns in rotation: prophets, nomads, heroes, and artists.
So, for example, according to them, the “Silent Generation” are prophets, the idealists that helped create the post-war establishments that Baby Boomers, as nomads, rebelled against. ” Gen Xers” are the heroes, who grow up increasingly protected, but mature into self-reliance. “Millennials” are artists, who “grow up overprotected by adults preoccupied with the crisis, come of age as the sensitive young adults of a post-crisis world.”
Sounds about right, right?
Well, I lied. In Strauss-Howe generational theory, Silent Generation members are artists, Baby Boomers are prophets, Gen Xers are nomads, and Millennials are prophets.
In the end, the whole thing reads like that Chinese zodiac printed on a restaurant placemat: vague enough to apply to anything– or to nothing.
And the results, when you try to use them, are as accurate as taking action based on that slip of paper inside a fortune cookie.
Care to share a Demographics Horror Story?
Nick
4 responses to “Demographics: The Second-Best Way to Segment Your File”
Behavioral Science Q & A
Integrating an individual giving appeal with other communications from a charity can have both positive and negative effects, and the outcome largely depends on how it’s executed. Advantages of Integration Brand Consistency: Maintaining a consistent appearance and messaging across all communications can reinforce the org’s brand identity and strengthen brand recognition and trust among your […]
Read Full Answer
I’m not aware of any in-market tests specifically comparing recurring vs. gift frequency language. I suspect the answer might not be the same with all gift frequencies, nor with all people. It sounds like a great opportunity for you to test and find out what works for your audience. Based on the literature, here’s a couple […]
Read Full Answer
Based on what we know from existing data, those renewal notices can actually be pretty effective in getting people to donate. They tap into our psychology – creating a sense of urgency, reminding us of past support, and using personalization to make the message hit home. They’re playing on our natural tendencies to feel obligated […]
Read Full Answer
Interesting question. I had a quick look at the testing done on this topic. On the positive side, in all cases, over half of donors decide to cover the fee. In some cases, it goes as high as 65%. Not a negligible percentage at all. Here’s another test from iRaiser showing consistent results (see point […]
Read Full Answer
There’s just one thing to consider when designing a supporter journey: the supporter. More specifically, you need to take into account: Who the supporter is i.e. their identity, which is the reason they support this cause, and their personality, which describes the way they “see” and process the world. These will determine the kind of […]
Read Full Answer
I’m not an expert in this but a quick search surfaced this article on the effect of tax reforms on 2019’s charitable giving. The researchers didn’t find a reduction. Actually, they observed an “increase in charitable contributions in 2019, even with the lower tax rates and the dramatically smaller number of taxpayers who itemize their […]
Read Full Answer
Firstly, thank you for posting the story about Netflix from Todd. It’s clear when people use this as a hammer for anti-demographic rhetoric they’re pretty desperate. You and Kevin (Schulman from DonorVoice) should know better than to compare a movie streaming service with donor fundraising, acquisition, cultivation, and retention. Not only is it out of context, but it’s also been retired for the more popular (http://www.businessinsider.com/netflix-thumbs-rating-system-flawed-2017-9) thumbs-up/down. Not even they could put multivariate surveying to good use.
Secondly, in line with “when-you’re-a-hammer-everything-looks-like-a-nail” thinking, I agree with most of the remainder of the post with one overall theme – segmentation of any kind takes proper preparation, analysis, and execution. The main hurdles I see with demographic and survey segmentation strategies are cost and coverage.
Let’s start with cost. I agree that most organizations can’t afford demographics, but at the same time they can’t afford surveying. At a minimum, surveying requires the following: contact information, a survey, and a way to analyze the results. This could mean, on the high end, sending a paper-based survey through the mail, having the responses keyed in, and then having a statistician analyze the results. That’s going to cost an organization about $5k and yield maybe 100 responses – hardly statistically significant. On the low end, sending an email to an online survey with automated analytics is going to cost an organization about $500 and yield 1,000 responses – pretty good right? Wrong – see coverage below. You see how neither is statistically sound. To get emails appended to solve the coverage problem sounds good, but it is very expensive and somewhat ethically questionable – what if the donor never gave you their email because they don’t want you emailing them? Somewhere in the middle is web-based, but that’s too random and hard to estimate – in my opinion. Anything telephone-based is a discussion for another time.
Next is coverage. As previously mentioned, email coverage can affect your survey results if the coverage is too low (we generally see about 35%), or if it’s biased one way or another – how many AOL/Gmail email addresses do they have? Looking at the distribution of work vs. home vs. freemail can tell you a lot about how donors use the email they’ve provided your organization. With the low input, the output is going to suffer. What’s an organization to do?
Demographics to the rescue! … before you stop reading and jump to “Leave a Reply”, let me say that you need to be careful where you get them from and how you use them.
What about the cost issue? This one’s easy – with the launch of TrueAppend.com you don’t have to spend a nickel on demographic segmentation reports anymore. So that’s done. It will still cost you to output the data, but at least you can get a high-level overview of your database in a few minutes for free.
What about coverage issue? We’ll be the first to say demographics cannot provide 100% coverage, but it’s still better than anything else out there by a longshot. We generally see 80% match rates – with 65% on the low end and 98% on the high end. What generally drives match rates is the same thing driving email coverage, young people don’t report their addresses and/or change of address, and older people have “fallen off” of most files because they are no longer the head of household or are living in a shared home. At least knowing the coverage by segment could better direct your efforts in data collection, regardless of the
Lastly, generations are not the same as age – collapsing 20 or 30 years of ages into one bucket is itself segmentation. Discussions of generation-based behaviors have historically ignored important demographic differences like wealth/class, race, religion, and geography. A teenager living in a rural area, near the poverty line has a very different life than a 30-year-old white collar professional who lives in a major city. Both of those people are “millennials”. In addition, marketing age ranges usually follow a pattern like 18-24, 25-34, 35-44, and so on. Those are not generational, they are age-based (alluding to the idea that people in their 30s are alike, as are people in the 40s, 50s, 60s, etc.) It would be much smarter to use a combination of demographic attributes to create a segment than to treat one attribute as gospel. That’s a strategic use of demographics. Lumping people together based on generations is not.
Here’s what I believe we can all agree on:
1 – Use RFM, demographics, firmographics, and surveying for segmentation – and please, please put some analytics/modeling over or across all of them.
2 – Use each segmentation to test acquisition, cultivation, reactivation, and retention strategies – no matter how small.
3 – Explore the unknown by acquiring and testing a random selection (which you can only get from a demographic data source) to see if there are new segments you haven’t identified.
Tim,
Demographics are cheap and available, which is hardly reason enough to make use of them for anything other than very broad brush targeting. Targeting is about efficiency, it has nothing to do with efficacy as it provides zero insight on motivation, needs or preferences. None.
As for “survey”, you are way off the mark and missing the point. Think census and business process to get coverage on ‘root cause’/donor understanding data – that only comes from asking folks (in right way, with right questions, validated as being causal of behavior) – for 100% of folks.
We have lots of examples of doing this, in-market. Happy to share over a beer (that you’ll be buying) next time we see each other.
Until then my friend, keep that beard long and pink..
Sorry you didn’t care for the Netflix quote. Definitely didn’t want to make it seem like nonprofits and Netflix were the same; rather I wanted to illustrate that when you have all of the possible data at your fingertips, demographics is one of the last things you’d use. (Also, the romantic comedy example is a nice simple analogy for gender-based donation approaches.)
A point of clarification, though: this quote wasn’t specific to the five-star method of rating they used to use or to the thumbs-up/thumbs-down new system. Rather, it’s to what data are worth getting. Hopefully, you liked the Ritson one better that makes the same point.
I disagree with you about the costs of surveying for a couple reasons (even in the mail):
1. Not all surveys are standalone surveys. If you include a survey as part of a pre-existing mail piece, the marginal cost is minimal.
2. Survey mail pieces, even when done as a standalone, can be profitable. We’ve had the positive surprise of people including donations with their surveys to the point that pieces more than broke even several times. Granted, you will almost never raise more money with a no-ask survey than with an ask (unless the ask stinks), but it can be net positive or neutral.
And, as you point out, online surveys don’t have and will never have full coverage. But if that’s your criterion, no survey should ever be done. All polls, surveys, and such have a margin of error because you don’t survey everyone. So is it always.
Nor is it what these surveys are for. They are to get people to raise their hands and let you know who they are and how they would like to be communicated with.
Let’s say you are an animal shelter. Your research (from surveys that are weighted based on commitment to the organization, not demographics) indicates that the #1 differentiator is whether they have adopted an animal from a shelter. (And that’s fairly likely…)
That’s not data you can get any other way. Yes, you maintain a group of donors for whom you don’t have that information on, who you continue to ask about. But your efforts all drive toward getting that data point, online and offline. And when the person answers the survey that gives you that information, you aren’t looking at significance, but rather appending that to the donor record and customizing the donor journey.
On age-breakdown, I’d agree generations aren’t predictive. I’d further agree that, to bastardize Kimberle Crenshaw et al, that you have to look at the intersections of age, race, wealth, location, orientation, religion, sex, gender, etc. to understand the people at each intersection. The challenge is that even if you took seven different demographic categories and restricted them to two different values, that’s 2^7 = 128 segments. That’s not dicing an audience – that’s throwing it in a blender and hitting frappe.
But I’m guessing you are advocating a modeling solution to that and that would be a better use of the data. The challenge there is that that (theoretically) helps you with in/out segmentation — should this person get this piece. But to the Ritson point above, that’s not a true segment, because it’s populated with different people who want different things.
More and more, I’m hoping people focus less on yes/no segmentations and more on ones that allow them to change the message they deliver to a particular audience.
That’s why I can’t agree with your “what we can all agree to” statements.
They ignore any sort of user-revealed data, which turn out to be most predictive. Further, as I argued against in http://www.theagitator.net/uncategorized/rfm-segmentation-first-refuge-of-the-scoundrel/, it uses transactional data as the first way of slicing and dicing your file, which eliminates the ability to customize by interest/identity.
But as ever I could be wrong (and often am). Are there examples you have of times when demographic-based or -focused segmentation has been able to effectively change messaging with positive results that you can share?
I’m know you’re both aware of the fact that almost 70% of demographics is built using surveys (registration cards, giveaways, etc.) so I guess it’s more of a question of application. I’m advocating understanding the composition of your donors so you can broadly segment them into groups for discussion, testing, testing, testing, and then implementing whatever yields the best result. The problem is that most organizations see all approaches as unapproachable and that needs to change.
Most of our greatest successes come from customers who have been able to better engage donors by evaluating life events and incorporating that into asks: using empty-nester, new parent, first-time homeowner, home upgrader/downgrader, retirement, and other household characteristics. We also perform look-alike services to test into new geographies where they might not have been aware an affinity existed. That is the simplest form of modeling. My recommendation on modeling overall is to use it as a guide and to help predict when you might be vectoring towards defeat (zero). It can help organizations save a lot of time and money.
As far as the census is concerned, if lack of resolution and red-lining are something NPOs are interested in, they should hire Brunell – he’s doing ‘great’ things. https://www.motherjones.com/politics/2018/01/trumps-pick-to-run-2020-census-has-defended-racial-gerrymandering-and-voter-suppression-laws/
I’m 100% in favor of surveying because I believe in the use of 1st, 2nd, and 3rd-party data while respecting the privacy and understanding the limitations of each.
Great stuff, thanks for (re)starting the conversation.