How to build shopping feeds buyer personas from customer research

What to take away

A shopping feeds buyer persona is a short written profile of the shoppers who share one buying trigger, one set of objections and one route into your catalogue.

Start with the questions your feed reports cannot answer

Feed reporting shows what changed after an edit. It rarely explains why a shopper hesitated, or which rival product they opened in another tab.

Personas close that gap. They record the trigger, the doubt and the proof someone needed before paying. Before you open a spreadsheet, list the questions your reports cannot answer.

Use order exports to compare first purchases, channel, discount use, repeat orders and returns. Where available, add Google Analytics 4 acquisition and landing-page data, on-site search logs, review text, returns or refund reasons and post-purchase survey answers; avoid putting unnecessary personal details into the persona record.

Group buyers by behaviour, not by age

Demographics describe people. Behaviour predicts orders. Sort buyers by how they arrived, what they bought first and how often they returned.

A shopper who buys once at full price behaves nothing like one who waits for a markdown. Keep them as separate personas even when they share a postcode.

In the order export, compare first purchases, repeat orders, discount use and returns by channel and product. Then compare those groups with Merchant Center queries and product performance: recurring combinations of query, product and buying behaviour are candidates for a persona. Keep groups separate only when their trigger or objection differs; otherwise combine them.

Put real voices into the profile

Interviews supply the wording that belongs in your product titles. Ask about the last purchase, the alternatives considered and the moment the basket was nearly abandoned.

Recruit recent buyers through a research invitation that follows your privacy notice and contact rules, and include different buying behaviours and channels. There is no universal interview count: set a practical recruitment target, cover each behaviour group and mark a persona as provisional if its evidence is thin. Ask what they were trying to solve, which alternatives they considered, what nearly stopped them and what information would have helped; if you offer an incentive, explain it before they agree.

Before collecting interview notes, document an appropriate lawful basis under Article 6 of the UK GDPR and follow the ICO's guidance on lawful basis; consent and legitimate interests are examples, but the right basis depends on the circumstances. The Data Protection Act 2018 also applies. Agreement to take part in an interview is not automatically the data-protection lawful basis, and the ICO's subject access request guidance covers a separate obligation if a customer asks to see their personal data.

Write the persona in the shopper's own phrases, then trim anything you cannot act on.

Use the same fields on each one-page profile: behaviour group, trigger, objection, proof needed, shopper's own phrase, evidence and date, feed action, owner and measure. For example: trigger—needs a replacement quickly; objection—delivery date is unclear; proof needed—a clear delivery promise; feed action—use accurate delivery-related wording in the title or make the feed and landing-page details agree.

Give every persona one feed action

A profile that only describes someone is a waste of a page. Each persona needs an owner, a change and a measure.

In Google Merchant Center, use product_type to organise relevant products and custom_label_0 through custom_label_4 to flag products for a persona-related feed action or performance review. A label describes products in your feed; it does not identify an individual shopper. Check the title and destination details against the actual product offer before publishing a change.

For a phone-first persona, compare Merchant Center product and landing-page activity with Google Analytics 4. Make sure the feed title, product details and landing page agree; Google's guidance on optimising a website for mobile covers page speed, tap targets and form length.

Match personas to cross-border orders

For orders to the EU from Northern Ireland, record delivery, returns and tax as possible shopper objections, then check that destination details in the feed and at checkout agree. GOV.UK explains how to report and pay VAT on distance sales of goods from Northern Ireland to the EU.

Use those cross-border questions to check whether a persona needs a destination-specific title, delivery detail or landing-page note. Do not add a tax or delivery claim that the offer cannot support.

Turn interview notes into a repeatable habit

One round of interviews is an anecdote. Personas need a method you can run again when a range changes.

Use the same opening questions and explain voluntary participation each time. Record the interview date, the buyer segment and participation information, and document the UK GDPR lawful basis separately. Group answers under the same trigger, objection and proof fields so that findings from the next round can be compared.

Checklist before you publish personas

Common questions

How many personas should a small retailer keep?

Four to six is workable for most catalogues. Fewer than four usually leaves gaps between search, marketplace and returning buyers. Beyond six, keeping them current becomes a chore.

Can I build personas without speaking to customers?

Yes, from order data and feed reports. Treat the result as a first draft, because reports show what happened but never the objection behind it.

How often should personas be reviewed?

Quarterly suits most catalogues. Bring the review forward after a range change, a price move or a new market, since those events shift buying triggers.

Do personas need to match product categories?

No. Group by how people buy, not by how the catalogue is filed. One persona can cover several categories when the trigger and the doubt are the same.