A retail media plan for 2027 with evidence gates, named owners and practical stop rules

A useful retail media strategy begins with one retailer-controlled advertising decision. It does not begin with a forecast, a supplier shortlist or a promise to turn customer data into sales. For an England operation, the first plan might ask whether a defined advertiser proposition should be tested in one on-site placement, using approved products, synthetic audience records and buyer-set stop conditions.

This guide is a planning method, not evidence that a campaign will work. The retailer, advertiser, agency, media owner, platform and measurement supplier may all handle different parts of the exchange. A named owner must document those boundaries before money or personal data moves. Publication also requires the qualified reviews listed above.

Fix the category before choosing a channel

IAB UK defines retail media as advertising space, retail data assets and in-store opportunities owned by a retailer or marketplace and made available to brands. Its industry definition divides delivery into on-site, off-site and in-store environments. It is useful for scoping, but it is not legislation and it does not prove that a proposed inventory, data use or claim is permitted.

Write a category record with these fields:

Field	Buyer entry
Retailer or marketplace property	Exact legal entity and controlled property
Advertiser exchange	What is supplied, by whom, for what consideration
Environment	On-site, off-site or in-store
Consumer task	The affected browse, compare, buy or visit task
Product scope	Approved products, identifiers and factual evidence
England connection	Store, audience journey or operating team, stated precisely
Exclusions	Trade promotion, ordinary merchandising and unrelated digital advertising

Do not combine environments merely because a sales deck calls them omnichannel. IAB UK's separate definition of off-site digital retail media involves retailer data used with inventory outside the retailer's online shopping properties. That introduces another publisher or platform boundary. An in-store screen has different accessibility, placement and operational questions from a product-page advert.

Start with a decision and a safe baseline

Record the decision in a sentence: "The named owner will decide whether to test, retain, change or reject this proposition by the review date, using the listed evidence." The baseline is the current way of working, including doing nothing new. Describe its cost, controls, known defects and evidence gaps without converting absence of measurement into zero effect.

The objective needs one primary outcome and several non-compensating guardrails. A primary outcome might be correct delivery of an approved placement to its eligible opportunity set. Its numerator, denominator, period, timezone, exclusions and source query stay blank until the buyer approves them. A sales event observed after exposure is not automatically an incremental result.

The 2025 edition of HM Treasury's Magenta Book is written for public-policy evaluation, not advertising. Its distinction between observing an outcome and establishing attribution is still a useful method boundary. A causal claim needs a credible comparison with what would have happened without the intervention, plus adequate data and specialist statistical review.

Guardrails should cover the consumer journey, advertising recognition, total price, accessibility, privacy choice, security, product evidence and operational correction. Failure in one gate cannot be offset by a better click or sales measure.

Map the exchange and its owners

Draw the proposed flow before selecting technology. The advertiser supplies the brief, substantiation and approved creative. The retailer decides what inventory it controls and which products may appear. An agency may act within written authority. A serving platform may receive creative and eligibility instructions. A measurement supplier may receive exposure and outcome records. These roles are not interchangeable, and a contract label does not settle every data-protection or consumer-law question.

For each hand-off, record input, permitted purpose, source, recipient, schema or document version, approval owner, retention, correction route and exit copy. Mark inferred fields separately from observed facts. If no one can explain where a field came from or how it can be corrected, exclude it from the first test.

The ICO's data protection by design and by default guidance was updated on 5 February 2026 following the Data (Use and Access) Act 2025. It puts privacy consideration at the start of the work. A named privacy reviewer must establish the parties' actual roles, purposes, lawful basis, necessity and transparency; a strategy document cannot decide those legal facts by itself.

Compare channel choices on the same job

Channel means the delivery environment here, not an acquisition source or pricing model. Compare the current baseline, one on-site placement, one off-site placement and one in-store placement against the same approved product, audience task, time window and evidence standard.

Decision field	Current route	On-site	Off-site	In-store
Inventory controller	Buyer records	Buyer records	Buyer records	Buyer records
Advert recognition evidence	Required	Required	Required	Required
Data recipients and device operations	Map	Map	Map	Map
Accessible customer task	Test	Test	Test	Test
Measurement population and unit	Same definition	Same definition	Same definition	Same definition
Full economic cost	Blank	Blank	Blank	Blank
Withdrawal and correction path	Demonstrate	Demonstrate	Demonstrate	Demonstrate

Do not award a winner while cells are unknown or units differ. IAB Europe published its Commerce Media Measurement Standards V2.1 on 22 January 2026. The industry record supplies definitions for parts of the measurement funnel. Its scope must be recorded rather than assumed to cover every England retailer or channel.

Keep the offer and customer journey honest

A placement can affect what a customer notices, believes and pays. CAP Code rule 2.1 says marketing communications must be obviously identifiable, while rule 2.4 addresses advertorial presentation in its section on recognition of marketing communications. A named advertising reviewer should assess the actual placement, wording, context and audience. A generic "sponsored" field in a plan is not evidence that the final presentation will be clear.

When creative contains a consumer price or buying invitation, preserve the retailer's product evidence and total-price controls. The CMA's final price transparency guidance was published on 18 November 2025 and updated on 7 January 2026. It covers mandatory fees, taxes and charges, drip pricing and partitioned pricing. The commercial owner must not optimise a placement by hiding information or preselecting an optional choice.

Create a claim register for every product statement, comparison, environmental claim, scarcity message, price and limitation. Each row needs its source, scope, owner, approval, expiry and withdrawal instruction. Creative cannot progress if substantiation is missing, even when media inventory is available.

Plan data and tracking as separate decisions

List every storage or access operation, identifier, match, audience rule, suppression, exposure record and outcome record. State whether it occurs on a retailer site, an advertiser destination, an external publisher or an in-store device. Do not call a dataset first party as a substitute for explaining source, rights, accuracy, purpose and recipients.

The ICO finalised its storage and access technologies guidance on 29 April 2026 after consultations and legal changes. It addresses cookies, pixels, fingerprinting and similar techniques under PECR and, where relevant, UK GDPR. The buyer must document whether an exception applies or valid consent is required. Refusal must be included in the test path, and it must not break a necessary shopping function.

Measurement validity and permission are separate gates. A technically useful identifier can still be impermissible. Conversely, a permitted operation may produce incomplete or biased reporting. Preserve both findings rather than turning one into approval of the other.

Use a reversible test, not a launch promise

The smallest test should use synthetic or specifically authorised records. Test the exact advert label, product claim, price and optional choice; tracking accepted and refused; missing, late and duplicate exposure events; an inaccessible journey; incorrect creative withdrawal; access revocation; incident escalation; and export or deletion.

NCSC guidance on building and operating a secure online service covers risk ownership, access management, logging, monitoring, incidents and supply-chain responsibility. Translate those topics into evidence requests for the actual service. Do not say the service is secure merely because a control appears in a document.

Each test case needs a fixture ID, expected result, observed result, evidence location, tester, date and verdict. Keep production delivery disabled until advertising, consumer, privacy, accessibility, security, IP, contract and finance owners have independently passed their gates. Any unresolved high-risk result means hold.

Give budget and measurement different ledgers

The budget records cash and economic cost, while the measurement plan records events and outcomes. Keep media charge, agency work, retailer labour, creative production, accessibility work, privacy and legal review, technology, tax treatment, refunds, incidents and exit work separate. Unknown values remain blank. An accountant must review VAT and accounting treatment for the actual parties and exchange.

For measurement, define eligible opportunities before impressions; valid exposures before clicks; and order states before any sales field. State numerator, denominator, attribution window, timezone, returns and cancellations, missing data and source version. Never subtract media cost from gross attributed sales and label the remainder return on investment. Contribution and causal incrementality need reviewed inputs and methods.

Assign authority, not decorative job titles

The accountable business owner decides whether the proposed exchange is worth investigating. The inventory owner controls the placement. Product and merchandising owners verify product facts. Advertising and consumer specialists review presentation. Privacy, security and accessibility reviewers control their own gates. Finance reconciles consideration and records. An analyst maintains definitions and uncertainty. An incident owner can pause delivery and preserve evidence.

Put a real name beside each role before approval. Record a deputy and escalation route. A senior commercial sponsor cannot waive another specialist's stop condition simply because the projected opportunity looks attractive.

Use 2027 as a recheck horizon

Nothing in this guide predicts conditions in 2027. Treat that year as the planning horizon for re-opening sources, contracts, permissions, channel definitions and test results. Recheck sooner after regulator guidance changes, supplier notices, a new data recipient, an altered consumer journey, a security incident, material complaint or measurement-definition change.

At each review, choose proceed, revise, pause or stop. Proceed only when the evidence matches the precise proposition and every independent gate passes. Revise when a bounded defect has a safe correction and retest. Pause when evidence is missing. Stop when the proposition relies on prohibited data, unsupported claims, hidden consumer choices, an inaccessible task or an unowned recovery path.

The decision record

The final page should fit on one decision sheet: named parties, bounded environment, approved product scope, customer task, current baseline, options rejected, assumptions, costs, measurement definitions, evidence links, gate verdicts, unresolved matters, test results, owner and next review date.

Do not fill a gap with a market average or supplier assertion. The defensible answer may be to keep the current route while the retailer obtains evidence. This draft remains on publication hold until named UK specialists confirm the current legal and technical position against the exact plan.