How to run shopping feeds reviews without a vendor demo

What to take away

If you sell online from a unit in Leeds or a spare room in Bristol, you have probably sat through a feed demo that looked excellent. Every product matched. Every channel synced. Then you exported your own catalogue and half the variants vanished.

That gap between demo and reality is what a shopping feeds review should close. This guide sets out a comparison method you can run yourself: scoring, risk checks, a worked cost example and a renewal routine.

Why reviews fail

Most feed reviews fail for the same reason. They compare software features rather than outcomes. Feature lists are easy to produce and hard to falsify, so vendors lead with them.

A better frame asks what happens to your catalogue on a bad day. What happens when a supplier changes a size chart mid-season? What happens when a marketplace rejects 400 SKUs overnight? What happens when the person who holds the feed credentials leaves?

Reviews also fail when the reviewer has no stake. A comparison built on published price lists tells you little about whether a platform holds up in November. Treat any review without a stated method as marketing.

Two other patterns are common. Reviewers borrow criteria from a sector that is not yours, so a fashion retailer gets judged on electronics mapping. Reviewers also score the sales process, awarding marks for a polished onboarding call rather than a working error log.

Our own approach is set out in shopping feeds review methodology in England, which explains how criteria are weighted and what evidence we expect before scoring a platform.

Set criteria before demos

Write your criteria down before you speak to a single vendor. Criteria written afterwards bend towards whoever presented best.

Start with the attributes your catalogue actually has. If you sell clothing, variant handling matters more than anything else: size, colour, fit and stock per combination. If you sell electricals, category taxonomy mapping and energy label data matter more.

Then add the operational criteria that decide renewals. Time to fix a broken feed. Quality of error messages. Whether support answers in writing. Whether you can export your configuration if you leave. Whether the contract allows a price change mid-term.

For a mixed catalogue, aim for nine to twelve criteria. More than that and the weighting becomes noise. Fewer and you will miss the operational points that cause churn.

The shopping feeds selection checklist in England turns these into questions you can put to vendors in writing, which is useful because written answers are harder to walk back later.

Scoring that resists spin

Score each criterion from 1 to 5 and give each a weight. Weights should reflect your catalogue, not industry averages. A pure marketplace seller and a brand with its own store will weight things differently.

For a mixed homeware catalogue, one illustrative criteria list with weights is: variant accuracy 25, taxonomy mapping 20, error handling and messages 15, support in writing 10, export and portability 10, total cost of ownership 10, attribute update speed 5, interface polish 5. These weights are examples, not fixed rules.

Insist on evidence for any score above 3. Evidence means a screen recording, a sample export, a written support commitment or a named reference you can contact. Claims without evidence score 2 at most.

A worked weighting helps here. Give variant accuracy a weight of 25 and interface polish a weight of 5. A platform that scores 5 on polish and 2 on accuracy still loses to a platform scoring 2 on polish and 5 on accuracy.

Score independently before you compare notes with colleagues. Group scoring drifts towards the loudest voice in the room.

Keep the scoring sheet. It becomes the basis for your renewal review twelve months later, when nobody remembers why the platform was chosen.

Risk checks that take minutes

Two public registers belong in every review, and both are free.

The Advertising Standards Authority publishes a list of non-compliant online advertisers. If a platform or an agency on your shortlist appears there, ask why before you go further.

The Information Commissioner's Office publishes its enforcement actions, including fines and reprimands. Feed tools touch customer data, so a record of data protection failures is directly relevant to your decision.

Two more checks are worth the time. Companies House filing history shows whether a supplier files accounts on time and who controls it. A search for the trading name catches businesses that have been dissolved and restarted.

These checks are not a substitute for full due diligence, but they catch the obvious cases cheaply. A vendor that objects to you running them is telling you something.

Data law and feed tools

The legal backdrop changed with the Data (Use and Access) Act 2025. It reforms parts of UK data protection law and adds provisions around data sharing and digital verification.

For feed work, the practical question is where customer and product data sits. Ask vendors which sub-processors they use, where data is stored, and how deletion requests flow through the pipeline.

Ask about retention periods too, and whether your product data is used to train anything. Get the answers in writing, because a processor list can change quietly.

If a vendor cannot answer those questions, score them down. The Act does not change the underlying principle that you need to know who holds your data and why.

VAT and product categories

Feed rules are not the only thing that varies by category. VAT treatment does too, and it affects what you can advertise at what price.

Some goods and services are exempt from VAT, and partially exempt businesses face restrictions on recovering input tax. GOV.UK sets out the rules on VAT exemption and partial exemption, which matter if you sell a mixed catalogue of standard-rated and exempt items.

A feed that pushes a single price to every channel will misprice exempt lines. Check whether the platform supports tax rules per category before you sign.

Delivery charges and discounts need the same treatment. A promotion that applies a percentage across a basket containing exempt goods can create a tax position you did not plan for.

The standard rate of VAT in the UK is 20 percent. Zero-rated items, such as most food, children's clothing and books, carry VAT at 0 percent. Exempt supplies, such as some education, healthcare and financial services, sit outside the scope. Registration is required once turnover passes £90,000 in any twelve months.

Partially exempt businesses can recover input tax on exempt supplies only under the de minimis rule. Exempt input tax must be no more than £625 a month on average and no more than 50 percent of total input tax. GOV.UK and your accountant should confirm the position for your catalogue.

In a feed, that means storing a net price, a VAT rate and a gross price per SKU, not one price across every channel. Ask vendors which tax rules they support per Google product category, and test a mixed basket before signing.

The augmented reality attribute

One attribute worth understanding is the 3D model link, which lets shoppers view a product in augmented reality. Google documents the 3D model link attribute in Merchant Centre Help.

It is optional and it applies only to certain categories, but it is a useful test of attribute coverage. A vendor that supports it probably tracks schema changes closely. A vendor that has never heard of it may be slow on other updates too.

Ask how new attributes are handled when Google, Amazon or Meta change their specifications. It is also worth asking how long the platform takes to update templates after a specification change, and whether you hear about it.

Worked example in pounds

Consider a fictional retailer, Harrow Lane Home, selling homeware from a warehouse in Sheffield. It lists 4,800 SKUs across its own store, Google Shopping and one marketplace.

Its current costs, all illustrative figures in pounds sterling:

The internal time lines assume a fully loaded cost of £35,000 per full-time equivalent, which makes 0.4 FTE £14,000 a year and 0.1 FTE £3,500.

Now suppose a new platform quotes £14,400 for the licence, double the current price. On the licence line alone it looks expensive.

But if it removes the agency retainer and cuts internal fixes to 0.1 FTE, the totals change:

The higher licence price produces a lower total, a saving of £21,300 a year. That is the comparison a scored review is designed to surface, and it is why licence price alone is a poor criterion.

Run the same table for each shortlisted platform using your own numbers. If SKU count doubles, licence bands may push the software line up while internal fixes stay broadly flat, so rerun the table at your growth forecast. Label every figure as an estimate until you have invoices in front of you.

Platforms named on most UK shortlists include Channable, DataFeedWatch, Feedonomics, Productsup and Simprosys. Licensing is usually quoted per month, per channel, or by SKU band.

For a 4,800-SKU catalogue, licence prices typically fall between £150 and £600 a month, which is £1,800 to £7,200 a year. Agency retainers for feed management commonly run from £500 to £2,000 a month, or £6,000 to £24,000 a year. Treat these as typical ranges, not quotes.

Agency versus in-house

Many retailers use an agency rather than managing feeds directly. Agency reviews need a slightly different method.

Ask for the named people who will work on your account, not the team in the pitch. How many other clients do they serve at your size? What happens when your account manager leaves?

Weight communication and handover higher than you would for software. Retainer reviews also need a notice period you can live with, and a clear answer on who owns the feed configuration when the contract ends.

Shopping feeds agency reviews in England covers the specific questions worth asking before you sign a retainer.

Comparison table structure

Build one table with platforms as rows and criteria as columns. Keep it to twelve criteria at most, or nobody will read it.

Include a column for evidence and a column for the score. A score without evidence is an opinion.

Add a final column for conditions: what would have to be true for this platform to win. That forces you to state your assumptions rather than hide them in a weighting.

Colour the cells only after the scores are locked. Colouring during scoring anchors the panel to an early favourite.

If you compare platforms by category fit rather than by price, shopping feeds product comparison in England sets out how to structure that comparison without the usual feature-list padding.

Renewal and re-scoring

A review is not a one-off purchase exercise. Re-score every twelve months, or sooner if your channel mix changes.

Keep the original scores and note what actually happened. Did the error handling hold up? Did support respond within the promised window?

This turns your review into a record. When a vendor asks for a longer contract, you have evidence rather than impressions.

Set the renewal date in the calendar at the point of purchase. Reviews that start three weeks before the contract expires become negotiations rather than assessments.

The wider market is shifting too. Shopping feeds: England market guide for 2027 covers how channel requirements and vendor pricing are changing, which is useful context when you set weightings for the year ahead.

Common questions

How long should a shopping feeds review take?

Allow two to four weeks for a shortlist of three platforms. Most of that time goes on gathering evidence and running your own catalogue through trial accounts, not on reading vendor material.

Should I score price or total cost?

Score total cost of ownership, including internal time and any agency fees. Licence price alone can point you at the more expensive option, as the worked example shows.

Do I need to check the ICO and ASA registers?

Yes. Both are public, free and quick. They will not catch every problem, but they catch the ones already on the record.

Can I rely on published reviews instead?

Use them for shortlisting, not for deciding. A published review cannot know your catalogue, your margins or your tolerance for feed errors, so the final score has to be yours.