Retail and Merchandising

The four questions a merchandising dispute is really about

Retail merchandising disputes come down to four things: assortment, which products were offered at all; placement, where those products appeared and in what order; promotion, what offer was attached to them; and pricing presentation, how the price and any comparison to it were displayed. Each of those is a decision, each was made in a system, and each of those systems keeps some record of when the decision changed.

The records are almost entirely party-held. Merchandising and category-management systems, content management platforms, theme and template version history, experiment platforms, promotion calendars and the order table itself all sit with the retailer or with a vendor working under contract to it. That is a different posture from a marketplace dispute, where the decisive record sits with a third party who has to be subpoenaed. Here the record is obtainable — but its completeness is itself a contested question, and often the honest answer is that the retailer never kept a version history of the thing now in dispute.

Counsel searching for a retail marketing expert witness usually arrives with a screenshot and a date. My first task is generally to explain why those two things together establish considerably less than they appear to.

Why what a shopper was shown is not a single fact

Modern retail sites are not one document served to everyone. If an A/B test was running during the period in dispute — a controlled comparison in which visitors are split between two versions of a page — then "what the site said" has no single answer, because different users saw different pages. Personalization rules, dynamic content and regional inventory produce the same effect. Any claim about page content built on a screenshot has to be checked against the experiment log covering the same period, and that is a first-order issue in false advertising and disclosure disputes, not a technicality.

A screenshot establishes almost nothing on its own. It does not fix the date, the server's response, whether personalization applied, or which experiment bucket the viewer was assigned to. A session recording is much better evidence of what one specific visitor experienced — what appeared on screen, in what order, what was clicked, where the visitor hesitated — but it is one visit on one device, it does not establish that the experience was typical, and it usually has fields masked or suppressed by design, so it is an incomplete record of even that visit.

The records that can fix a page at a date

Where a contemporaneous record of page state exists, it is usually in one of these:

• Content management revision history — post and page revisions with author and timestamp
• Theme or template version history — timestamped where the retailer used a version-controlled workflow, and frequently absent where someone edited the live theme in a browser
• Tag manager version history — a versioned, timestamped, attributed record of every container publish, showing what tracking was live, on what pages, from what date, published by which account
• Experiment configuration and assignment logs — what was changed, what the hypothesis was, how traffic was split, when it started and stopped, and which visitors saw which variant
• Form and checkout logs — including, critically, the version of the form or checkout step that was displayed at submission
• Third-party archives — useful, with known gaps

The experiment configuration is the most under-appreciated of these, because it is pre-registered. It records what the operator intended and believed before the outcome was known. In a dispute about whether a vendor did what it said it would, or about whether a change preceded a decline, that is direct evidence of intent and sequence rather than reconstruction after the fact.

Placement on a marketplace, where the retailer is not the publisher

Omnichannel retailers sell where they do not control the page. On a marketplace the placement question becomes which offer was featured, and that record is thinner than counsel expects. Featured Offer history is not exposed as a first-class report. A seller can infer its own share from business reporting, but a timestamped record of which seller held the featured offer at a given moment generally has to be reconstructed from third-party monitoring services — which sample at intervals, know nothing between samples, observe from a single vantage point, and carry no certification. The featured offer can also vary by customer, so a single-vantage observation is not a statement about every shopper.

The listing itself is a shared catalog record on some marketplaces, which is what makes hijack disputes possible: a third party edits a record the brand considers its own. The contribution or change log held by the marketplace is the evidence that matters there, and what a marketplace currently exposes, and for how long, should be confirmed against the account rather than assumed.

Query-level marketplace reporting, where the brand is enrolled and eligible, ties specific search terms to impressions, clicks and purchases for a specific product. That is the closest thing to a keyword-to-purchase record outside a paid-search platform, and it is directly useful when the allegation is that a rival's listing intercepted demand.

Promotion and pricing presentation as an evidentiary problem

When the allegation concerns how a price or an offer was presented, the analysis is a reconstruction: what was displayed, to whom, when, and what was actually charged. The records that answer it sit in different systems. The order table carries line items, discounts and timestamps. The promotion or offer configuration carries the rule and, if the platform versions it, the history of that rule. The checkout logs carry submission timestamps, validation failures, the abandonment point, and the payment gateway's request and response pairs with authorization codes and decline reasons. The experiment log says which variant the shopper was in.

The version question runs through all of it. In a dispute about whether a shopper agreed to something or was shown something, the question is never "does the page have a disclosure" — it is what this page looked like when this person transacted. Answering that takes the content version history, the experiment assignment, and any dynamic content rules. Three records, usually in three separate systems, usually with three different retention periods, and usually collected late.

The omnichannel case, where the store is not in the dataset

The recurring hard case is the one where the transaction did not happen where the marketing did. Digital systems observe digital behavior. Nothing in an analytics property, an ad platform or a tag container observes a purchase made at a counter. The connection between the two exists only where somebody built it in advance — a geographic test designed before the campaign ran, or offline transaction data the retailer itself fed back to the platform.

This also constrains what the platform models can say. Google's data-driven attribution covers Google's own surfaces; it does not see a marketplace, retail media, email, affiliate, direct mail or television, let alone a store visit. So a channel-share exhibit built from platform reporting in an omnichannel matter is describing a fraction of the commercial activity and presenting it as the whole. Saying which fraction, and how large the unobserved remainder is, is part of the work.

Whether a merchandising change actually moved sales

This is where retail cases are won and lost, and the honest starting position is that a chart with a vertical line on it establishes sequence and nothing else. Before the disputed change can be offered as the explanation, the competing ones have to be enumerated and addressed: seasonality against at least two prior comparable periods; category-wide movement; an own-price or promotion change, which moves conversion rate independently of traffic; stockouts, discontinued items, shipping cost changes or checkout changes; site and technical changes including migrations and tag breakage; competitor entry, exit or spend change; a cut in another channel entirely; and measurement changes on the retailer's own side, such as a tag change, a consent banner deployment or a filter change.

A report that does not visibly work through that list before landing on the disputed conduct is doing post-hoc reasoning, and should expect to be challenged as such. Working through it is also the most common reason a defense-side engagement is worth having: the list frequently contains something the other side's expert did not address, and a documented alternative explanation does not have to be proven to be damaging.

What the retail record does not settle

No record in this field establishes why a shopper chose one product over another. Merchandising data records what was offered and what was bought. The step in between is inference, and an expert who narrates it as observation is offering an opinion the data does not carry.

Marketplace analytics cannot establish causation between a listing or merchandising change and a sales change, because ranking, rival behavior, price, inventory status, seasonality and the marketplace's own advertising all move at once.

A featured-offer tracker cannot establish continuous state, only sampled state from one vantage point. Impressions and reach figures are not counts of people; reach in particular is a deduplicated estimate of accounts. And where the retailer edited a live template rather than working through version control, there may be no contemporaneous record of what the page looked like at all — in which case the correct answer is that the evidence does not support a finding either way, which is a legitimate conclusion and a far safer one than a reconstruction dressed up as a record.

Frequently Asked Questions

Published 2026-08-12