The Red Flag Register: 40 warning signs in a PIM selection

Nobody can say who decides which attributes exist, who owns each attribute group, or who approves changes to the model.

Ask Who in our organisation owns the product data model, and who signs off attribute ownership before the RFP goes out?

Without model ownership, requirements stay vague and vendors cannot be compared on the work that matters.

The requirements list reflects e-commerce or IT only, with no input from product management, supply chain, compliance or sales.

Ask Which departments reviewed and validated the requirements, and where are their sign-offs recorded?

Missing stakeholders surface late with requirements that reopen the shortlist or delay the decision.

The timeline was set backwards from a budget deadline, and the requirements and data model weeks were cut to fit.

Ask Which steps did we shorten to meet the date, and what risk does each shortcut add to the decision?

Compressing the early weeks removes the foundation that makes demos and scoring meaningful.

By the start of scenario design, the team has not assembled representative products, assets and supplier files for vendors to use.

Ask Who owns the sample data pack, and which product families, edge cases and files will it contain by the end of week 6?

Without your own data in the demos, vendors fall back on showcase data and the scenario method loses its value.

A senior stakeholder already prefers a platform from a previous employer or a conference, before any requirements exist.

Ask Can we agree that every vendor, including the favoured one, goes through the same fit criteria and scenarios?

A favourite chosen early bends requirements and scoring towards it, and other stakeholders stop trusting the outcome.

The response marks almost every line as fully supported out of the box, with no configuration notes, no partial answers and no exceptions.

Ask For each requirement, can you state whether it is standard, configured, scripted or delivered through a partner extension?

A response without nuance means the vendor either did not read your requirements or prefers you not to see the gaps until implementation.

Answers mention industries, channels or modules you never raised, or the same paragraph appears under unrelated requirements.

Ask Which named person on your side read our requirements, and can they walk us through three answers of our choosing?

Library answers hide whether the platform fits your data model, so you lose the comparison the RFP was meant to create.

Capabilities are described in the present tense, yet you cannot find them in release notes, documentation or the demo environment.

Ask Which of these capabilities are generally available today, in which release, and where is the documentation?

Roadmap features slip, and a contract signed on future functionality leaves you paying for software you cannot use.

Questions on attribute counts, variant depth, locales or catalogue size receive the answer unlimited, or are skipped entirely.

Ask What attribute, variant, locale and product counts does your largest comparable customer run in production today?

Every platform has practical limits, and a vendor that will not name them has not tested them against a model like yours.

The commercial section is left blank or marked to be discussed, while the functional answers are complete.

Ask Can you price the scope in our RFP now, using our stated SKU count, users, locales and channels as the volume basis?

Late pricing lets a vendor anchor on enthusiasm after the demo, and it blocks an early total cost comparison.

Requirement IDs are regrouped, merged or renumbered, so you cannot compare responses line by line.

Ask Can you resubmit the response in our template, with one answer per requirement ID and no merged lines?

A response you cannot compare line by line cannot be scored fairly against the other shortlisted vendors.

The demo environment shows polished sample products from a fictional brand, and your sample data pack is absent or only partly loaded.

Ask When will our sample data pack be loaded, and which scenarios will you run on it during the session?

Showcase data hides the modelling compromises your own products would force, so the demo predicts nothing about fit.

The presenter starts with dashboards or AI features, runs out of time and skips the scenarios you weighted highest.

Ask Can we run the scenarios in the order of our script, and stop the session if time runs short instead of skipping any?

Skipped scenarios are usually the ones the platform handles worst, and they carry the most weight in your scorecard.

Every screen is pre-built and the presenter declines to add an attribute, change a validation rule or alter a workflow step live.

Ask Can you add a new attribute with a validation rule now, and show the effect on completeness for our sample products?

Live changes show how much effort configuration takes after go-live, which drives your running cost for years.

AI descriptions, extraction or classification are shown on products prepared in advance, never on items from your sample pack.

Ask Can you run the same AI function on ten products we select from our sample pack, during this session?

AI output quality depends on the input data, so results on curated items say little about results on yours.

Bulk edits, imports or exports are shown on 10 or 20 items while your catalogue holds tens of thousands.

Ask Can you run the bulk edit on 10,000 products from our pack and show the elapsed time and job log?

Performance problems appear at volume, and they become user adoption problems within weeks of go-live.

The presenter answers questions on permissions, audit or integration with promises to send documentation later.

Ask Who on your team can answer these questions live, and can that person join the second demo session?

Deferred answers rarely return with the same precision, and the gaps tend to sit in the areas that cost most to fix.

ERP, DAM or webshop integration appears as a diagram, with no running connector, API call or job monitor on screen.

Ask Can you show a running integration job, its error handling and its log, using a connector a customer runs in production?

Integration is one of the largest cost lines in a PIM project, and slides cannot show how it behaves when data fails.

Every reference went live within the last year and was proposed by the vendor's sales team.

Ask Can you add two customers live for more than three years, including one that went through a major version upgrade?

Recent go-lives still feel the launch energy, while long-running customers know the real cost of change and support.

References come from other industries, or run catalogues and channel counts far smaller than yours.

Ask Which customer runs a similar product model, catalogue size and number of channels, and can we speak to them?

Fit in one industry rarely transfers to another, because the data model and standards differ sharply.

The vendor offers a director or sponsor, and declines to connect you with the product data team that uses the platform daily.

Ask Can we also speak with a daily user and with the person who maintains the integrations?

Sponsors defend their decision, while daily users and integration owners describe performance, workarounds and support.

The vendor insists on joining or moderating every reference call, or provides written statements instead of calls.

Ask Can we speak with the reference customer without anyone from your company on the call?

References speak openly only without the vendor present, and moderated calls rarely surface the issues you need to hear.

References run an older platform generation or an on-premise edition, while you are buying the new SaaS product.

Ask Which customers run the exact edition and version in our quote, and since when?

A new platform generation can differ from the old one in architecture and maturity, so older references prove little.

The proposal shows senior consultants during the sale, but the contract does not name who will deliver the project.

Ask Can you name the solution architect and lead consultant, and commit to them in the contract with a replacement clause?

Delivery quality depends on the people on the project, and unnamed teams often change between sale and kick-off.

The partner quotes a fixed implementation price before seeing your sample data, source systems or data quality.

Ask Which assumptions about data volume, quality and source systems sit behind this price, and what happens if they prove wrong?

A fixed price built on assumptions turns into change requests once the real data arrives.

The partner has only one or two certified consultants on the platform, or certified them during the sales cycle.

Ask How many consultants hold current certification on this platform, and how many projects on it have you delivered in the last two years?

A thin bench creates single points of failure and slows delivery whenever a key consultant is unavailable.

The project plan includes custom development for functions such as workflows, imports or exports that the platform covers through configuration.

Ask For each custom component, which standard feature did you assess first, and why does it fall short?

Custom code raises the cost of every upgrade and ties you to the partner that wrote it.

The partner advises on vendor choice while also reselling licences or receiving referral fees it has not disclosed.

Ask Do you receive any margin, referral fee or incentive from the vendors on our shortlist, and can you disclose it in writing?

Undisclosed incentives compromise the independence of advice at the moment the decision matters most.

A second offer arrives with a large discount on identical scope, often close to the vendor's quarter or year end.

Ask What changed between the two offers, and which discount will apply at renewal and for additional volume?

A deep first-year discount without scope change usually returns as a renewal uplift, so the total cost barely moves.

Price is tied to API calls, channel-SKU combinations or data volume, and the vendor cannot model it against your three-year growth plan.

Ask Can you price years one to five using our SKU, channel and market growth plan, with the overage rate for each metric?

A metric you cannot forecast turns growth into an unbudgeted cost and weakens your position at every renewal.

AI generation, translation or enrichment is bundled at no cost, with no stated credit volume, end date or post-trial price.

Ask How many AI credits are included per year, what does one credit buy, and what is the price per credit after the included volume?

AI usage grows quickly once teams adopt it, and undefined credit terms become a new cost line at renewal.

The order form shows a flat annual fee, while the master agreement allows an annual increase without a stated cap.

Ask What is the maximum annual price increase over the full term and at renewal, and can you state it on the order form?

An uncapped uplift compounds every year and can erase the first-year discount within the contract term.

Functions shown in the scenario demo, such as supplier portals, syndication or AI, turn out to sit in separate paid modules.

Ask Can you map every scenario we saw in the demo to the licence lines in your quote, and price any missing module?

Buyers compare quotes on a scope they assume matches the demo, so missing modules distort the whole comparison.

The quote includes one production environment, while development, test or staging environments appear later as separate fees.

Ask How many non-production environments are included, and what does each additional environment cost per year?

Implementation and release testing need at least one non-production environment, so the cost is unavoidable.

The agreement allows the vendor to set renewal pricing at its then-current list price, with no percentage or index cap.

Ask Will you agree to a renewal cap tied to a named price index or a fixed percentage, written into the order form?

Switching PIM systems takes months, so without a cap the vendor holds the stronger position at every renewal.

The exit clause promises data return without naming formats, or only through the vendor's own export package.

Ask Can the contract state that all product data, assets, history and configuration export in documented open formats at no extra cost?

Product data without its model, relations and history loses much of its value when you migrate to another platform.

The uptime percentage excludes planned maintenance, but the agreement sets no limit on the number, length or timing of maintenance windows.

Ask How many hours of planned maintenance per month are allowed, in which time windows, and with how much notice?

Unlimited maintenance exclusions make the availability figure meaningless for teams that publish to channels daily.

The agreement refers to online policies or service descriptions that the vendor may update at any time.

Ask Can you attach the current versions of all linked policies to the contract, and require our consent for material changes?

Terms the vendor can edit after signature can change your SLA, data rights or AI terms without negotiation.

The AI terms allow the vendor or its model providers to use customer content for training or product improvement by default.

Ask Can the contract state that our data and AI outputs are excluded from model training, by you and by any model provider you use?

Product data, pricing logic and unreleased products are commercially sensitive and should not leave your control.

The term renews automatically for another full period unless you give notice six months or more before expiry.

Ask Can we shorten the notice period to 90 days and receive a written reminder from you before it starts?

A long notice window means a missed date locks you into another term at the vendor's renewal price.