Retail after peak 2026: what to settle before the first half of 2027

Peak 2026 will be compared with peak 2025 in almost every retail board pack in January. For many retailers the comparison will be unsound, because the rules that decide what is counted, what is shown and what is charged changed during 2026.

In the UK, the cookie rules in the Privacy and Electronic Communications Regulations were rewritten from 5 February 2026, and analytics that meets the statistical purposes exception can now run without consent, while advertising measurement still needs it. The CMA's price transparency guidance under the Digital Markets, Competition and Consumers Act 2024 sets out how mandatory delivery charges must appear in prices, and the CMA has started fining traders under the new regime. In the EU, low-value parcels from outside the Union have carried a €3 customs duty per item since 1 July 2026, and a €2 Union handling fee applies from 1 November 2026. Shoppers are also reaching product pages from AI assistants that read catalogues and store policies directly.

Each change moves a number the board watches: sessions, conversion rate, average order value, delivery cost, cross-border margin and channel mix. A retailer that records the date of each change and the metrics it affects can still compare one peak with another. One that does not will spend February explaining differences that have nothing to do with demand.

The ICO's guidance sets out the statistical purposes exception. Storage and access for analytics can proceed without consent when the sole purpose is aggregate statistical information about how the service is used, provided visitors get clear information and a simple, free way to object. Advertising is outside it. Connecting a visitor ID to purchases so that conversions can be shared with advertising partners needs consent, and so does measuring whether a user clicked an advert.

A retailer that has moved its analytics onto the exception will record more sessions in peak 2026 than in peak 2025 for the same number of visitors, while its advertising platforms still see only consenting users. Conversion rate in analytics falls, conversion rate in the ad platforms holds, and neither movement describes shoppers. The order count in the commerce platform is the series that did not change, so it should be the reference line in every peak report.

Returns cause a larger distortion. Google Ads lets an advertiser restate or retract a conversion using its order ID, but adjustments reach automated bidding only if they are made within 7 days of the conversion being recorded, and no adjustment is accepted after 54 days. Peak returns arrive in late December and January, mostly after the first window has closed. Automated bidding therefore learns from peak orders at their gross value, including the orders that come back, and carries that learning into January campaigns.

A retailer with high peak return rates should restate whatever it can within 7 days, accept that later adjustments correct reporting only, and set January targets on net revenue from the order book rather than on platform-reported return on ad spend.

Free delivery above a basket value is one of the oldest levers in online retail. Under the DMCC Act it is also a disclosure question. The CMA's final guidance treats mandatory delivery charges as part of the total price that must be shown from the first invitation to purchase, a term it reads to include product listings, baskets and marketplace search results. An offer of free delivery over £40 does not make the delivery charge uncertain. Until the basket reaches the threshold, the full delivery fee has to be included in the total price shown.

For order-level charges the guidance accepts three practical routes. A retailer can show the total price beside the base price with equal prominence, show a running total that stays visible throughout the visit, or fold delivery into the product price. Each route has a commercial consequence. A running total keeps product prices low but needs front-end work on every template that displays a price. Folding delivery into the product price is simple to implement, and it raises the price that marketplaces, comparison sites and AI shopping assistants read from the product feed. ChatGPT's shopping results often highlight the lowest-priced offer they know of.

The CMA is enforcing. It fined the AA £4.2 million in April 2026 for leaving a £3 booking fee out of the displayed price of driving lessons, fined StubHub £900,000 in June 2026 over delivery and service fees, and opened three further drip pricing investigations in August. Its guidance extends responsibility to marketplaces and other parties that present the price, as well as the seller.

A delivery threshold should now be set with three numbers in view: the delivery cost it covers, the margin on the orders it pulls up to the threshold, and the cost of the display route chosen. A review that looks only at the first two misses the third.

Marketplace fees are usually reviewed as a percentage of the selling price. The figure that matters for 2027 is the marketplace cost per order net of returns, set against the contribution the same order would earn on the retailer's own site after payment, delivery and acquisition costs. Without that comparison, a marketplace that looks cheap on its headline rate can turn out to be the most expensive channel once returns handling and fulfilment charges are counted.

Cross-border costs are rising. In 2025, 5.9 billion low-value e-commerce items entered the EU, four times the 2022 volume. Since 1 July 2026 the EU has applied a temporary €3 customs duty per item on consignments worth up to €150, so a parcel holding several items can carry several charges. The €2 Union handling fee per parcel applies from 1 November 2026. Under the EU customs reform, online platforms and non-EU sellers become responsible for duties and product compliance as importers for distance sales, and use of the EU Customs Data Hub becomes mandatory for e-commerce on 1 July 2028.

The UK is moving the same way. At the 2025 Autumn Budget the government announced that the £135 customs duty relief for low-value imports will be removed by March 2029 at the latest, and it published its consultation response in July 2026.

For a UK retailer shipping low-value orders to EU customers, the cost per parcel rose twice in four months, and the increase falls hardest on the small orders a free-delivery threshold was designed to grow. Cross-border thresholds and minimum order values set before July 2026 are out of date.

Agentic commerce has moved faster in announcements than in checkouts. OpenAI launched Instant Checkout in ChatGPT in September 2025 and has since shifted its focus away from it. In ChatGPT's shopping results a shopper chooses a product and a retailer and then completes the purchase on the retailer's website, much as with a search engine. Native checkout inside AI assistants exists, but for most retailers it is a waitlist or a pilot rather than a channel.

The larger change for most retailers is on the supply side. Shopify turns on agentic storefronts by default, and its default setting enables AI shopping channels automatically, including new ones as they appear. The catalogue those assistants read is generated from the store, and assistants may also consult the retailer's terms, privacy policy and returns policy when answering a shopper. Many Shopify merchants are therefore already present in AI channels without anyone having decided which ones, or checked what the assistants are reading.

The useful work for the first half of 2027 is unglamorous. Review which AI channels are switched on and why. Check that titles, attributes, stock and prices in the catalogue are correct, because an assistant comparing offers reads the feed. Make returns and delivery policies complete and consistent with what checkout actually charges. Track AI referrals as a separate channel, with conversion and return rates measured in the order book, and decide on native checkout integrations only once that data exists.

Before peak, in late October and November 2026, record every 2026 measurement break in the reporting calendar with its date and the metrics it affects, and make the order book the reference series in the peak dashboard. Audit how delivery charges appear on product pages, baskets, marketplace listings and advertising against the CMA's guidance, and fix the gaps before traffic peaks.

During peak, upload refund and cancellation adjustments to Google Ads within 7 days wherever an order ID allows it, and keep a daily net revenue line beside platform-reported value.

In January 2027, run the post-peak review on net revenue after returns, by channel and by product, and set January and February targets on that basis.

In February and March 2027, recalculate free-delivery thresholds for domestic and EU orders using delivery cost, margin on threshold-driven orders, the chosen price display route, and the €3 duty and €2 handling fee on EU parcels. Compare marketplace and own-site contribution per order net of returns before setting the 2027 channel plan.

From April to June 2027, review AI channel settings and catalogue quality, and decide on any native checkout integration using six months of referral data. Retailers that import low-value goods directly to UK customers should map the cost of losing the £135 relief and start the systems work it will need.

We start this kind of work from the order book and the refund ledger rather than from dashboards, because those are the records that did not change in 2026.