Consumer behavior research: trust the order history over the survey
Consumer behavior research works best when it starts with what people did (purchase data, tests, diaries) and uses interviews to explain why.
Consumer behavior research should start with what people did and use conversations to explain it. It asks why people buy, use and stop using products, and people's accounts of their choices often differ from what they do. So purchase records, tests and diaries come first. Survey answers about future intentions are the weakest evidence of the lot.
For interviewing shoppers about a recent purchase, see our consumer research guide. This page is about choosing and combining methods that capture behavior.
A made-up example: the vegetarian meals nobody orders
Here is a hypothetical example used throughout. You run a meal-kit company. In a customer survey, many subscribers say they want more vegetarian recipes, so you add several. Three months later, those recipes are picked far less often than the survey suggested.
Nobody lied. People may like the idea of eating less meat, forget how they choose on a busy Sunday, or pick the answer that sounds reasonable. The survey measured an attitude; your orders measured behavior.
People describe the shopper they mean to be
Jakob Nielsen of Nielsen Norman Group sums up the rule for product research: watch what people actually do, don't believe what they say they do, and definitely don't believe what they predict they will do. Nielsen gives three reasons self-reports drift (NN/g):
- People bend answers toward what they think you want to hear or what is socially acceptable.
- They report what they remember, and memory is unreliable.
- They rationalize what they did.
Attitudes still matter. NN/g calls attitudinal research (asking about thoughts and feelings) and behavioral research (observing actions) complementary, and says mismatches between the two are often a good source of insight (NN/g on attitudinal and behavioral research).
Five methods that capture behavior
| Method | What it shows | Main limit |
|---|---|---|
| Your own purchase and usage data | What customers chose, when and how often | Shows what happened, not why |
| A/B test | Which of two versions changes a behavior, such as sign-ups | Needs enough traffic; tests only what you offer |
| Diary study | Real choices logged as they happen over days or weeks | Participants must keep logging; takes planning |
| Observation or field visit | How people shop or use a product in context | Time-consuming, and people may act differently when watched |
| Public spending data | How households in general spend | Broad categories only |
NN/g lists analytics and A/B testing (comparing two versions on a measure such as conversion) as behavioral methods (NN/g).
Use a diary study for purchase decisions that unfold over time. Participants report experiences as they happen, over a few days to a month or longer, remotely and often more cheaply than field visits. NN/g suggests roughly 5 to 12 participants for a small discovery project and a broad guideline of about $40 per hour of expected time for a general U.S. sample (NN/g diary studies).
For the meal kit, a two-week diary where subscribers log each meal choice, and why, would show what actually drives the pick. (Expect at least one entry that says "it was late and the picture looked nice.")
For spending in general, the U.S. Bureau of Labor Statistics' Consumer Expenditure Surveys publish spending, income and demographic data, collected through an interview survey for major or recurring purchases and a diary survey for small, frequent ones (BLS).
Interviews explain the behavior you already found
Anchor every interview in a real event: "Open your last order. Walk me through how you picked these three meals." Their own order history reduces the memory problem. Skip "Would you buy more vegetarian meals?" because it invites a prediction.
In the example, interviews might show that subscribers choose meals in under a minute on their phone and default to dishes they can picture. The fix could be better photos or familiar names. Our guide to interviews versus surveys covers when each fits, and checking interviews against product data shows how to compare the two.
People who watch thousands of shoppers can tell you what is normal
Store and category managers, merchandisers, retail analysts and support leads watch thousands of customers every day. They can't replace your customers, but they know which patterns are common in the category and which are yours alone.
Instant Expert can reach them: you describe the experience you need, review who it finds, and it sends your invitations. You pay for the calls that get booked. Pages such as data and analytics professionals in grocery retail, marketing professionals in retail and marketing professionals in beauty and personal care show the kind of profiles you might shortlist.
A five-step plan
- Write the behavior you want to explain, such as "subscribers skip vegetarian meals."
- Pull the behavior data you already have.
- Where the data is thin, add a test, a diary or an observation.
- Interview a handful of customers about specific recent choices.
- Where words and actions disagree, treat the behavior as the fact and the explanation as the thing to investigate.
For an outside view from people who see many shoppers, start a search with a one-sentence description of the experience you need.
Combine behavior evidence with explanations
- 1
Name the behavior
Write the specific action you want to explain, such as skipped recipes.
- 2
Pull existing data
Check purchase and usage records before asking anyone anything.
- 3
Add a method
Run a test, diary or observation where your data is thin.
- 4
Interview
Ask about specific recent choices, using order history as a prompt.
- 5
Reconcile
Treat behavior as the fact and explanations as the thing to test.