Survey and Quiz Segmentation: Self-Declared Preferences
Asking people to sort themselves, the response rates to expect, and the gap between declared and actual behaviour.
On this page
The short version
- Asking is the cheapest segmentation data available and the least reliable. People answer accurately about their intentions and inaccurately about their behaviour.
- One question in the welcome sequence outperforms a survey campaign by a wide margin, because it arrives at the moment of highest willingness.
- Use declared data to decide what to offer, and behavioural data to decide what is true.
The simplest way to segment a list is to ask it. No integration, no event tracking, no join between systems — just a question and a stored answer.
It works, with one important qualification: what people say about themselves is a reliable guide to what they want and an unreliable guide to what they do.
Where declared data is accurate and where it is not
Accurate: what someone is interested in, which of several topics they would prefer, how often they want to hear from you, what their role is, what problem they are currently trying to solve. These are questions about preference and current state, and people answer them honestly.
Unreliable: what they will do, how often they do something, how much they spend, how experienced they are. These invite an aspirational or flattering answer, and the gap between the declared and the actual is systematic rather than random.
The practical rule: ask about preferences and use behaviour for facts. A subscriber who said they want weekly emails and opens one a month has told you two different things, and both are useful — the first tells you what to offer, the second what is true.
Question types by reliability
| Question | Reliability | Use for |
|---|---|---|
| Which of these topics interests you most? | High | Content routing |
| How often would you like to hear from us? | High | Frequency preference |
| What are you trying to fix right now? | High | Sequence selection |
| Which best describes your role? | High, with a short option set | Version of the message |
| How experienced are you with this? | Moderate — skews confident | Tone, loosely |
| How often do you send email? | Low — self-report | Nothing important |
| What is your budget? | Low, and off-putting | Avoid |
Why a survey campaign underperforms
A dedicated survey asks for effort with no immediate return, arrives at no particular moment, and is answered by a self-selecting minority — usually your most engaged subscribers, whose answers are therefore unrepresentative of the list.
That last point is the one that matters and is usually missed. Segmenting a whole list on data supplied by the fifteen percent who answered means the other eighty-five percent are being classified by a default, and the defaults were chosen by people unlike them.
One embedded question at a natural moment avoids all three problems, and it can be repeated at different points to build up a picture over time.
Honouring what people tell you
The fastest way to make this data worthless is to ask and then ignore the answer. A subscriber who selected monthly and receives weekly emails has learned that the question was decorative, and they will not answer the next one.
It also damages more than the segmentation. Asking a preference and overriding it is a small broken promise, and those accumulate into the impression that nothing you say is load-bearing.
So only ask questions whose answers you will act on. A preference you cannot honour is a question not to ask.
Survey segmentation check
- One question at a time, in a sequence rather than a campaign
- Asked immediately after delivering something
- Preference questions, not self-reported behaviour
- Three or four options, each a single click
- Every answer changes something the subscriber will notice
- Frequency preferences are actually honoured
- Non-responders are not classified by a silent default
Frequently asked questions
What response rate should I expect?
A one-click question in a welcome email does far better than a survey campaign — often by several multiples. Absolute rates vary too much by list to quote, so measure your own before designing anything around it.
Should the question be in the email or on a page?
In the email, as links. Every step between the question and the answer costs responses, and a link that records the answer directly is one gesture.
How do I handle people who never answer?
Leave them unclassified rather than assigning a default. A default silently fills your segments with people who never expressed a preference, which makes the segment data worse than having none.