Before you report on product photography measurement, fix your brief

What to take away

• Most teams open a dashboard first. That is the mistake. Measurement fails when the brief, the asset naming and the baseline are not fixed before a shoot is booked.
• Agree three things in writing: what an asset is for, which number should move if it works, and who owns that number when it does not.
• Report a small set of metrics consistently. Conversion rate by product page, return rate by reason code, cost per usable asset and time from brief to live cover most catalogues.
• Score every asset or shoot against a written rubric so decisions are repeatable and suppliers face the same standard.
• Review monthly, and keep a dated change log. Without that record, every result is arguable.

Why visual measurement usually fails

The failure is rarely analytical. It is definitional. A team commissions new imagery, sees a conversion lift, and cannot say whether the lift came from the photography, a price change, a stock fix or a seasonal shift.

That happens because the asset was never described in a way the analytics can read. If a product page carries six images and nobody recorded which ones changed, the comparison is guesswork dressed as data.

Naming is the cheapest part of the fix. Give every file a stable identifier that carries the SKU, the angle and a version number, then carry that identifier into the page markup and the analytics events.

A second failure is ownership. Photography sits with brand, the product page sits with ecommerce, and the analytics sits with a third team. Nobody holds the number that links them.

A third is the tool-first habit. Teams buy a dashboard, then look for something to put in it. The dashboard ends up dictating the questions, which is backwards.

The fix is unglamorous. Write down what each asset type is for, agree the measure that proves it, and name the person who reports it. Do that before the next shoot, not after.

The supporting reading on product photography measurement mistakes sets out the recurring errors in more detail, including the ones that look like good practice.

Set the brief before you set the metric

A brief that says "refresh the category imagery" cannot be measured. A brief that says "replace the packshot on 40 best-selling SKUs so the label is legible on a 360px thumbnail" can be.

Write four lines for every commission: the asset type, the job it does, the page or placement it will sit on, and the decision it should inform. If a shoot cannot fill those lines, it is a brand exercise and should be budgeted as one.

The placement line matters more than most teams expect. One file may serve a marketplace tile, a category grid, a product page gallery and a paid social crop. Each has its own aspect ratio and its own legibility floor, and naming them in the brief prevents the rework that follows a delivery.

This discipline also protects the supplier relationship. A photographer who knows the crop and the placement can shoot for it, rather than delivering a file that gets reworked later. The wider procurement picture is covered in the product photography tools and supplier guide for 2027, which is worth reading before you renegotiate rates.

Keep the brief in the same system as the asset record. A brief in an email thread is a brief that will be lost by the time anyone asks why the images changed.

Build a baseline you can defend

You cannot claim an improvement without a before. Capture at least four weeks of the relevant metrics before any change goes live, and note anything else that happened in that window.

Four weeks is a planning convention, not a rule. If your category sells in a weekly rhythm, four weeks gives you four comparable cycles. If it sells in a monthly one, four weeks gives you a single cycle, and you should extend the window.

Record the baseline at the level you will report at. Category-level averages hide SKU-level movement, and SKU-level noise hides category trends. Pick one primary level and stay with it for the whole test.

Where you use analytics to gather that baseline, the ICO guidance on online tracking is the reference for what you may collect and how consent applies to the tags doing the collecting. Getting that wrong invalidates the dataset as well as the compliance position.

Consider the traffic you actually have. A page with a few hundred sessions a week will not produce a readable conversion signal in a fortnight, however strong the imagery. For example, a SKU group receiving 300 sessions a week and converting at 2 per cent will move by only a handful of orders over that period. Pool those products with comparable lines or run the test longer.

Store the baseline with a date and a version number. When someone challenges a result six months later, the version is your evidence.

Choose the metrics that answer a business question

Every metric should map to a decision someone is willing to make. If no decision follows from a number, it is decoration.

For most ecommerce catalogues, four families cover the ground. Commercial performance, usually conversion rate and average order value at product level. Fulfilment quality, usually return rate and return reason. Production efficiency, usually cost per usable asset and time from brief to live. And asset quality itself, which is where a rubric earns its place.

Be careful with engagement metrics. Time on page and image zoom rates can be useful diagnostics, but they are not outcomes. A shopper zooming repeatedly may be struggling, not deciding.

Return reason codes deserve more attention than they usually get. If returns cluster on a colour mismatch or a missing accessory, the imagery is a plausible cause and often a cheap one to correct. Read the free-text comments alongside the codes.

The detailed definitions, including how to calculate cost per usable asset so it survives a finance review, are in the guide to product photography key metrics. Use that as the dictionary for the rest of this process.

A scoring rubric for visual assets

A rubric turns opinion into a repeatable judgement. Score each asset or batch on a five-point scale, weight the criteria, and record the total alongside the asset record.

The weights below are illustrative for a general ecommerce catalogue. Adjust them to your category, then keep them fixed for at least two quarters so results stay comparable.

Criterion	What a top score looks like	Weight
Product accuracy	Colour, scale and included items match the listing exactly	20%
Thumbnail legibility	Product is identifiable at the smallest placement used	20%
Consistency with range	Lighting, angle and background match sibling SKUs	15%
Information value	Image answers a likely pre-purchase question	15%
Technical quality	Resolution, crop and file weight suit every placement	15%
Channel fit	Meets the spec of each marketplace or feed it feeds	15%

Score 1 to 5 on each criterion, multiply by the weight, and sum. A score below 3.5 should trigger a reshoot request rather than a debate.

Two rules keep a rubric honest. Score blind where you can, so the person scoring does not know which supplier produced the asset. And re-score a sample of older assets each quarter to check that your standards have not drifted.

Publish the rubric to suppliers before the shoot, not after delivery. Send it with the brief and include two worked examples, one that passes and one that does not, so the standard is visible rather than implied.

Track rework rate as a supplier measure: the proportion of delivered assets that needed a second pass. It is the clearest early signal that a brief was ambiguous or a standard was assumed.

Reporting cadence and format

Report monthly. Weekly reporting on visual assets generates noise, invites over-reaction, and consumes the time you need for the next shoot.

One page is enough. Lead with the decision you are asking for, then the metric that supports it, then the caveat. Put the detail in an appendix nobody has to read.

Separate production reporting from performance reporting. Production reporting covers throughput, cost and rework. Performance reporting covers what happened on site after the change. Mixing them produces a document that satisfies nobody.

Where you need a reusable structure, the product photography reporting dashboard in England shows how to lay out the panels so a non-specialist can read them in under a minute.

Keep a change log. Every annotation on the chart should correspond to a dated entry describing what changed and when. This single habit resolves most arguments about results.

Report the months that went against you as promptly as the ones that did not. A test where new imagery failed to move conversion is useful information. Holding it back until budget season makes the whole programme look like advocacy.

Attribution when several things change at once

Perfect attribution is not available to most ecommerce teams. Accept that and design around it.

Use holdout groups where volume allows. Leave a matched set of SKUs on the old imagery for the test window and compare like with like. Where volume is too low, use a staggered rollout so the change arrives at different times across comparable groups.

Holdouts carry a cost. Leaving a matched group on old imagery means deliberately keeping weaker assets live for a period, which someone in trading will notice. Agree the size of that group in advance, and agree who signs it off.

Be explicit about what you cannot separate. If a price promotion ran during the same fortnight, say so in the report rather than presenting the visual result as clean. A stated limitation is more persuasive than a suspiciously tidy number.

The methods comparison in product photography attribution methods in England covers holdouts, staggered rollouts and the practical limits of each approach.

If the business needs analysts who can run this properly, formal training is a reasonable investment. The CIM data and insights training courses set out the kind of skills a measurement owner needs across data handling, interpretation and reporting.

Compliance and asset records

Measurement touches compliance more often than teams expect. Adverts that make a claim must be capable of substantiation, and the imagery is part of the advert. The ASA AdviceOnline library for marketers is the practical reference for what visual advertising may show and claim.

Ecommerce listings carry their own transparency duties. The Electronic Commerce (EC Directive) Regulations 2002 set out the information a commercial communication must provide, which affects how product pages and their supporting assets are presented.

Asset records also have a security dimension. Original files, licences and model releases are business assets, and the GOV.UK guidance on business innovation, protection and cyber security is a sensible starting point for storing and protecting them.

Retention is a decision, not an accident. Decide how long you keep raw files, retouched masters and the scoring records that refer to them, and write that down. Deleting the record while keeping the file is the worst of both.

None of this is a reason to over-collect data. It is a reason to label and retain what you already have.

Common questions

How long before product photography results are meaningful?

For most catalogues, four to six weeks of consistent data is the minimum before a result is worth reporting. High-traffic categories can read faster. Low-volume SKUs may need a full quarter, and should be pooled with comparable products rather than judged alone.

Should we measure image quality or commercial performance?

Both, but keep them apart. Quality scoring tells you whether the asset meets the brief. Commercial metrics tell you whether meeting the brief mattered. A high quality score with no commercial movement is a signal to revisit the brief, not the photographer.

What is a reasonable cost per usable asset?

It depends on category, retouch depth and how many placements each file must serve. Build the figure from your own invoices: total shoot cost divided by the number of assets actually published. Treat any external benchmark as a sanity check only.

Do we need a dashboard before we start measuring?

No. A spreadsheet with a fixed column structure will carry you through the first two quarters. Build the dashboard once you know which questions you keep asking, otherwise you will rebuild it twice.