Ask the Customer, Then Use the Answer: Zero-Party Data from Collection to Campaign

A brand adds a quiz to its signup: skin type, concerns, routine. Eight thousand people answer in the first month. The answers land in a field on the profile. Every email after that is the same email everyone gets. Six months later, a new hire finds the field, asks what it is for, and nobody remembers.

Zero-party data, the things a customer tells you on purpose, is the highest-quality input a program can have: explicit intent, no inference. An answer isn’t consent to be marketed to, so the consent record still has to exist. We described its place among the four sources of customer data. The hard part was never collecting it. It is using the answer fast enough that the customer sees why they were asked, and then keeping the answer current.

The rule: activation before the question

Every question you ask has a campaign that reads the answer, built before the question goes live, and sending within a week of the answer. If the reader doesn’t exist, the question doesn’t either.

Most teams do it the other way around. The preference center gets designed first, with twenty fields, because the tool makes it easy. Then the team looks for uses. The fields that never get a reader stay empty, get filled once and go stale, and become the mystery attribute the new hire finds.

Start from the campaign that would be better with one declared fact. Ask for that fact, in the place and moment where answering is natural. Ship the campaign that uses it. Then ask the next question.

Three collection patterns, and the reader for each

Fig 1 · Ask, store, use · Three ways to collect zero-party data and the reader for each

The signup question. One question, at the moment of subscribing, that changes the first email. Which category do you shop most; are you buying for yourself or a gift; what size. The reader is the welcome series: email one is different for each answer, visibly. The customer sees the connection within a day.

The failure: five questions at signup. Completion drops, and the four extra answers have no reader.

The progressive question. One question inside a message or on a page the customer is already on, weeks after signup, asked because the next campaign needs it. In the post-purchase email: “was this for you or someone else?” On the category page, after the second visit: “what are you looking for today?” The reader is the campaign that prompted the question, and it sends within the week.

This is progressive profiling: the profile fills in over time, one useful field at a time, each with a reason the customer can see.

The preference center. Channel, frequency, topics, and the declared attributes that matter, in a place the customer can return to. The reader is the contact policy and every campaign that segments on topic. It comes last, once the signup and progressive questions have proven which fields earn their place. We wrote the consent side in consent management and preference centers; the declared-data side follows the same rule: fewer fields, each with a reader.

Where the answer lives

On the profile as a customer attribute, with the date it was given and the surface it came from. Not only as an event, because events are history and a segment needs the current answer in one field. And not in the survey tool, where no campaign can reach it.

Keep the date, because declared data goes stale. A skin-type answer from 2023 may be wrong. A “buying for a gift” answer is wrong by next month. Each attribute gets a validity period in the tracking plan, and a campaign that reads an expired one either re-asks or falls back to behavior.

When the answer and the behavior disagree

A customer declares an interest in running shoes and has viewed nothing but hiking boots for three months. Both are true: one is what they said, the other is what they did. The declared answer wins for the campaign it was asked for; behavior wins for everything else. And a disagreement that persists is itself a signal: re-ask, once, in the next message, and let the customer update the field.

The rule to avoid is the one where declared data overrides behavior forever. It produces the running-shoe email to the hiking-boot buyer for two years.

What to measure

Three numbers per question. Answer rate (how many of those asked answered), which says whether the question was placed well. Activation rate (how many answers were read by a campaign within seven days), which says whether the reader exists. Lift on the campaign that reads it, against a holdout that got the generic version, which says whether the question was worth asking.

A question with a high answer rate and no lift was the wrong question. A question with lift and a low answer rate was asked in the wrong place. Both are fixable once you can see them.

The takeaway

Ask one thing, in the place where answering is natural, because a campaign that sends within a week needs it. Store the answer on the profile with a date. Let behavior win everywhere the answer wasn’t asked for. Measure the answer rate, the activation rate, and the lift. Zero-party data is the best input a program has, and it only stays that way if every question keeps the promise it made.

Read next

• Where customer data comes from: zero, first, second, third party. Where declared data sits among the four sources.

• Consent management and preference centers: what good looks like. The consent half of the preference center.

• Frequency caps, quiet hours, and channel budgets. The contact policy that reads the preference center.