“There are so many different tentacles to all this stuff. You just need to try to get progress made and then iterate… The 80% is pretty quick, but that last 20%, you just discuss and iterate on it forever and go nowhere.”
– UpClear CPG Data Summit participant
This observation, shared by a participant at UpClear’s 2026 CPG Data Summit in New York, got a lot of heads nodding. It reflects an idea that many CPG and FMCG practitioners have found but rarely discuss: the pursuit of perfect data quality can become its own obstacle. When teams get stuck on improving their data system, it’s often because they’re waiting for a state of readiness that never quite arrives.
The question worth asking instead is: what does good enough actually look like, and how do you get there without getting lost in minutiae?
The first thing to accept is that data readiness is not a destination. Andrew Dentinger, UpClear’s Product Director, emphasized this in his Summit session: “Data readiness is never a complete state. You’re always going to be continuously assessing what you have, and there’s always going to be a degree of approximation. You’re never going to get to a state where you are 100% confident in the data you have.”
This is especially true in CPG, where the volume and variety of data sources create mismatches that are difficult to fully resolve. Syndicated data providers don’t have 100% coverage. POS data from different retailers arrives in different structures. UPCs that exist in your internal system may not match the way the same products are coded by Nielsen or Circana.
These misalignments, while frustrating, are inevitable in our industry. While you can strive for improvements, you can’t delay your initiatives waiting for a perfect system to work in.
Most CPG brands can get the majority of their data in reasonable shape without too much effort. The remaining gap is what becomes a time sink. Avoiding this trap starts with building a clear framework for what “good enough” means. Without that, the conversation quickly becomes circular.
Don Baker, Partner at the practitioner consultancy The Partnering Group, observed this issue across his career. “A lot of companies are getting paralyzed in the conversation and aren’t acting,” he said. “Just make some progress somewhere. Pick a lane.” His partner at The Partnering Group often reminds him of the idiom: “Perfection is the enemy of good enough.”
According to Don, the way to break the paralysis is to stop treating data decisions as philosophical questions and start anchoring them to numbers. “If you’re not anchored to something you’re trying to change from a numbers perspective, you haven’t got a chance of getting a glove up on a lot of this stuff. Because then you’re just talking people’s preferences. And that’s a conversation that can go in circles forever.”
In practice, this means working backwards from what you’re trying to measure. If the goal is to improve forecast accuracy by three points of MAPE, that’s a number. You can figure out what data changes would actually move that number, and you can stop debating the things that wouldn’t. If improving SKU-level granularity in your planning data would improve MAPE meaningfully, it’s worth the effort. If it wouldn’t, it isn’t, regardless of how theoretically cleaner it might be.
One participant at the CPG Data Summit made a point about the risk of over-indexing on detail: “I’ve come to believe that at some point it’s kind of diminishing returns, because you’re not doing as good of work on a limited data set as you’re trying to get all this detail across it… It’s much better to have the organization focus on getting a smaller set really good than try to do everything.”
However, it’s important to consider the context when deciding how much to invest in your process. UpClear’s Andrew Dentinger made the point that the threshold for what counts as good enough depends on the use case. “If you’re working on more financial type scenarios, accuracy is increasingly important,” he said. “If you’re more on a use case around market performance and what have you, directional suffices.”
Not every use case requires the same level of precision. Knowing which ones do, and building your data processes around those first, is more productive than trying to achieve uniform quality across everything at once.
Once you’ve anchored your priorities to what you’re trying to measure, the question becomes: Where should you focus first? Don consistently points to three foundational areas shaping your trade system: pricing, product, and customer.
The most common failure mode Don sees is a lack of a clear price list and an ERP pricing routine that the trade system can actually read. EDLP allowances buried in the list price rather than reflected as a separate spend item, and customer prices that float order to order, create downstream problems that are hard to untangle. Without a true list price as a starting point, you can’t accurately measure what you’re spending at a customer or compare spending across customers.
The relationship between your internal product hierarchy and external data sources is a persistent source of misalignment. Don Baker posed some questions during his session. “Are the UPCs or SKU numbers coming from the syndicated providers the same way you have them set up in your system? And is there somebody watching that as SKUs get added or get deleted or get changed?”
In most organizations, the answer is no. Syndicated data doesn’t have 100% coverage and uses different structures than your internal data, which creates approximation challenges that don’t have clean solutions. In this case, the recommendation is to design around them rather than wait to solve them.
Customer hierarchy problems tend to emerge quietly. Customers buying through both direct and indirect channels, hierarchies that don’t roll up consistently, and markets defined differently by your internal system and syndicated providers all create situations where ROI work is either inaccurate or impossible to complete.
Don offered a rule of thumb for indirect retailer accounts: If you can’t get point-of-sale data to measure ROI, treat it as a financial exercise rather than a trade planning one, and don’t over-engineer the planning structure around it.
Accepting that your data will never be perfect is not the same as saying it doesn’t matter. The companies that make the most progress on data quality are not the ones that set the highest standards upfront. Rather, they are the ones that remove the process barriers that prevent them from improving at all.
In practice, that means starting somewhere specific rather than trying to solve everything at once. Pick the master data area that is causing the most downstream pain. That may be pricing misalignment, UPC mismatches with syndicated providers, or a customer hierarchy that does not roll up consistently. Then, start by building governance around that. Get the right people in the room, agree on simple rules, and start tracking where the gaps are.
Over time, the picture becomes clearer. You learn which data problems are actually affecting your KPIs and which ones are just noise. You learn where approximation is good enough and where precision genuinely matters. Eventually, you’ll build the internal muscle to keep improving incrementally rather than waiting for a wholesale transformation that never quite arrives.
As Don Baker put it, “Just make some progress somewhere. Pick a lane.”
Want to learn more about how to improve your data system without pulling a loose thread? Watch Don Baker’s full session from UpClear’s 2026 CPG Data Summit to go deeper on how to build a trade system that works for every function.
Master data is the core reference information that every other system and process in your business depends on. This includes your product hierarchy, your customer hierarchy, and your pricing structure. It is the foundation that trade planning, demand forecasting, deduction management, and post-event analysis all build on.
The main issue with master data quality is that nobody owns it. Master data tends to get managed to roles rather than to standards. The ERP team makes sure the syntax is correct. The commercial team adds new items when they need to. Finance books things the way the audit requires. No function is specifically accountable for the connections between them, leading to gaps that widen over time.
AI amplifies what it is given. A model trained on clean, well-structured data produces useful outputs. A model trained on misaligned or incomplete data produces confidently wrong outputs, at scale, faster than anyone can catch them. This means that applying AI to a bad dataset leaves you worse off than you started.
The main reason CPG/FMCG teams will never achieve “perfect” data quality is because the environment is dynamic. New products get added, customer hierarchies change, retailers reorganize, syndicated data providers update their databases, and ERP systems get upgraded. Each of those events creates the potential for new misalignments. Even if you reached a state of perfect master data quality today, it would start degrading tomorrow.
UpClear est une société de logiciels et l'éditeur d'Blue, une plateforme de gestion des revenus ( intelligence )platform utilisée par les marques de biens de consommation. Nous proposons une solution globale de gestion de la croissance des revenus (Revenue Growth Management), comprenant des fonctionnalités de TPM, TPO, IBP et RGM. Notre mission est de donner aux marques les moyens d'optimiser leurs performances en termes de revenus et le retour sur investissement de leurs actions commerciales grâce à un logiciel intelligent et collaboratif, offrant une source unique de données fiables, une automatisation rationalisée et des informations exploitables.
La plateforme RGM «Blue » ( Intelligence )Platform prend en charge l'ensemble des processus de gestion des revenus, du brut au net : planification opérationnelle annuelle, planification des comptes et exécution. Les solutions s'appuient sur l'analyse de données, l'intelligence artificielle et la gestion des données, qui permettent de relier les équipes et les systèmes d'entreprise.
Nous sommes là pour vous aider.
[email protected]
