
Audience research
Part of How to run shopping feeds audience research for multichannel shops
Before you write shopping feeds target audience profiles, do this
Most teams write shopping feeds target audience profiles from gut feel, then wonder why campaigns miss. Build profiles from first-party data instead.
What to take away
- Most teams describe their ideal shopper first, then hunt for data to fit. That order produces fiction. Start with the records you hold.
- Your first-party sources are order history, on-site search, support tickets, returns and consent records.
- Consent must be documented before an email address goes near a feed.
- Write the profile last. It should summarise what the data shows, not what the team hopes.
Start with the data you hold
Pull twelve months of order lines, not three. Seasonality distorts short windows, and a Christmas-only view makes gifting look like your whole business. Export SKU, price paid, postcode district and the channel behind the order, then the same period from site search and returns.
Twelve-month data export checklist
- Pull twelve months of order lines
- Export SKU and price paid
- Export postcode district
- Export order channel
- Add same-period site search
- Add same-period returns
You now have four overlapping lists. The overlap is your audience. A district that buys at full price and keeps the item behaves differently from one that waits for a sale code.
Separate buyers from browsers
Feed platforms optimise towards whatever you tell them a conversion is. If a newsletter signup counts as a conversion, your profile fills with people who never buy. Decide which events count, then check that your feed software passes the same definitions to every channel.
Do conversion definitions match?
Do all channels share one conversion definition?
Write audience profiles
Fix definitions first
If the definitions differ, fix that before writing anything about audience. The wider audience research guide for England sets out how the pieces fit together.
Check consent records first
Email feeds depend on lawful consent. Under UK GDPR and the Privacy and Electronic Communications Regulations, you generally need consent before sending direct marketing by electronic mail to individuals. Consent must be specific, recorded and easy to withdraw.
Before you segment, confirm every address has a dated, documented opt-in. If the record is missing, remove the address. One suppressed contact costs less than a complaint, and the ICO guidance on direct marketing using electronic mail explains what a valid record looks like.
Talk to customers, not personas
Data shows what happened. It rarely shows why. Six conversations with recent buyers will surface reasons no dashboard holds, such as a delivery preference or a size uncertainty that keeps someone browsing.
Run them before building the profile, not after. A short method for recruiting and scripting them sits in shopping feeds customer interviews in England.
Write the profile last
By now you have segments with counts, behaviours and stated reasons. Write one page per segment. Include the size, average order value, the channel that reaches them and why they buy. Leave out adjectives about attitude.
Where a rule affects how you may contact them, check the current position. The GOV.UK business regulation guidance and tools page is a sensible starting point for retailers checking obligations.
Worked example
A homeware retailer with 4,200 orders in twelve months splits its buyers into three groups.
| Segment | Orders | Average order value | Main channel |
|---|---|---|---|
| Repeat gifting buyers | 900 | £68 | Email feed |
| One-off clearance buyers | 2,100 | £24 | Paid search feed |
| Full-price browsers | 1,200 | £91 | Organic feed |
Marketing spend was £6,000 a month across all three groups. The team had put £4,200 behind clearance buyers, the lowest-value group. After rebuilding the profiles, it moved £1,500 a month towards the full-price group and held gifting spend steady. These figures are illustrative, not measured results.
Keep the profiles current
Review the numbers each quarter. A segment that was 40 per cent of orders in spring can shrink by autumn. Set a calendar reminder and repeat the export.
Feed errors also distort the picture. A disapproved product never reaches the audience you built the profile for, so check diagnostics alongside the data. Google's guide to common issues in Merchant Centre and how to resolve them is a quick way to rule out a feed fault before blaming the audience.
Common questions
How many customer conversations is enough?
Six to ten usually surfaces recurring reasons. Stop when new conversations repeat what you have already heard.
Should the profile include people who never buy?
Only if you can reach them lawfully and affordably. Otherwise keep them on a separate list.
How often should the profile change?
Review quarterly and after any major pricing or range change. Monthly shifts are usually noise.
Do I need consent for every email address in the feed?
Yes, for direct marketing to individuals. Keep a dated record of each opt-in and remove any address you cannot evidence.



