Any business with a bit of nous will take advantage of a clean-up, migration project, or PIM system integration to take a deep dive into their product data quality. So, it generally improves during such events, but then, over time, quality degrades again. It might be missing attributes creeping back into product records. Or perhaps values start to drift. Maybe you see the return of the dreaded duplicates. Inevitably, marketplaces begin to reject your product feeds.
We’ve written this article to explain why this regression happens (essentially, a failure of enforcement and a not-fit-for-purpose operating model). We also look at what exactly it breaks, both operationally and commercially, and what you can put in place so your product data management will stay stable and not force you to fall once again into the bottomless pit of firefighting.
Failure: The consequences and impacts
Data failure: Standards exist but aren’t rigorously enforced. Key elements like attribute definitions, naming conventions, taxonomy rules, and completeness thresholds aren’t embedded in workflows or system controls.
Operational consequence: Your teams revert to the usual suspects: workarounds like using spreadsheets, importing supplier data manually, ad hoc quick fixes, and “just get it live!”-type shortcuts. What was clean data gets overwritten by syncs with ERP, supplier files, or overhasty internal updates.
Commercial and risk impact: Your channel rejection rates rise, time-to-market gets longer, search and filters don’t perform helpfully, returns start to proliferate due to incorrect product specs, and finally, exposure to non-compliance risk grows as regulated fields go missing or fall out of date.
Regression isn’t a series of random incidents. In fact, it becomes the default outcome when your PIM is simply treated as a storage facility rather than as an environment requiring a robust data governance framework.
Why quality flourishes briefly, then starts to fall apart
The driver behind short-term data enhancement is usually concentrated effort:
- A project team
- A backlog burn-down
- A fixed deadline
and
- Close scrutiny (but only temporary)
During that window of opportunity, exceptions get resolved and people are focused on quality in itself, and paying attention. Then the project ends.
If the business hasn’t changed how data is created, validated, approved, and monitored on a day-to-day basis, that hard-won quality drifts away from its moorings. The impulses behind this drift are consistent in most mid-to-large organisations:
- Uncontrolled data entry points: Suppliers, ERP exports, internal uploads, agencies, and regional teams all enter data into the catalogue in different ways.
- Manual workarounds ignore any existing controls: teams bypass controls to deal with issues outside the system, then they re-import, unsuspectingly creating version conflicts and overwrite problems.
- “Speed beats quality”: if the system allows incomplete product data sets to progress, the people responsible will push them through under the pressure of deadlines that need hitting.
- Ownership is fragmented or disputable: The road to commercial ruin is paved with good intentions – It takes little time for the stated “everyone owns data quality!” to become the unspoken ‘no one owns data quality.’
- Lack of monitoring: These issues stay under the radar until a channel fails, launches slip significantly, or customer complaints start to rise alarmingly.
‘The tooling trap’: PIM is software: It doesn’t enforce behaviour by itself
A Product Information Management (PIM) platform is a highly versatile piece of kit, but it doesn’t create a data governance framework automatically. Even with the best PIM on the market, you can still ship poor data if:
- Mandatory fields are not truly mandatory (or conditional rules are missing)
- Validation rules are weak or applied inconsistently across categories
- Teams can bypass workflow gates with no pushback
- Supplier data has no validation checks at ingestion
- Integrations overwrite enriched or approved fields without passing any controls
This is why announcing that “we’ve implemented a PIM!’ is NOT the same as “we have product data governance in place.”
The mechanics of regression: What you can actually see
Look for the following patterns because they’re the usual quality leakage points:
- New SKUs reintroduce old problems
Supplier onboarding is inherently inconsistent. Each new range you ingest brings its own missing attributes, odd units, and free-text values. That incoming (and loud) noise will almost certainly outpace your teams’ ability to correct it all.
- Clean data gets overwritten
ERP updates or refreshed supplier data may overwrite enriched attributes because field-level ownership and rules on overwrites aren’t defined and documented. In the worst cases, even one sync can undo weeks of work.
- Taxonomy drift creeps in
Categories are added informally. Attribute applicability changes but isn’t updated accordingly. Products end up misclassified, leading to broken channel-specific mappings and search filters.
- Shadow attributes appear
A situation emerges where teams create their own versions of the same field (such as weight vs shipping weight, or colour vs marketing colour). Doing this inevitably creates an internally contradictory catalogue.
- Quality only gets attention when it hurts
If the only trigger to remedial action is “The feed’s failed!” your people are permanently behind the commercial curve. You’re basically managing consequences, not quality issues (or, if you like, dealing with the symptoms, not the cause).
How to prevent data quality regression: Enforcement alongside an operating model
If you want product data quality to stay in the stars, you’re in need of a product data operating model designed to produce quality continuously.
Stabilise: protect what you cleaned
- Lock down integrations: define your rules for field-level overwrites (as in: What can overwrite what, and under what circumstances).
- Reduce entry points: stamp out ‘the spreadsheet fix fixation,’ where re-importing is the default.
- Clearly (and closely) define the standard that data must fulfil to be “ready” for syndication/publication – and this means the minimum publishable attribute set per category and/or channel.
Standardise: make quality unambiguous
- Maintain an attribute model: Definitions, formats, units, allowed values, controlled vocabularies, and conditional logic.
- Standardise supplier Provide templates and impose mapping rules (even if certain suppliers don’t comply, your ingestion process must have these rules).
- Align your taxonomy with channel requirements and keep the channel mapping documented and up to date.
Enforce: make it hard to create bad data
- Validation rules that block saving or publishing when required fields are missing or malformed
- Workflow gates with accountable approvals (especially for attributes related to compliance)
- Supplier onboarding validation at the point of entry: Reject or quarantine non-conforming files, and route exceptions explicitly
Monitor: Make drift visible early
- Use a scoring system for completeness by category and channel
- Configure dashboards to identify the most frequently recurring errors, top sources of overwrites, and the average time-to-fix.
- Schedule alerts for when quality drops below a given threshold, with a defined workflow for resolving the issue(s).
Once these controls are in place, the emphasis on data quality is no longer confined to periodic clean-up projects but becomes a natural (and welcome!) byproduct of normal operations.
Next steps: book a discovery call
If your data quality improves briefly but then starts degrading again, you’re in a bind, commercially. Reach out to us today at Start with Data and we can organise a discovery call – the quickest route to clarity. We use our experience across sectors and use cases to focus specifically on the root causes: operating model gaps: ownership, enforcement, overwrite rules, supplier onboarding, and monitoring.
Let’s get to the bottom of why the quality of your product data is letting you down and how we can remedy that pitfall.