Category Management Software: Why Every Review Costs Three Weeks of Deck Building

Why category reviews eat weeks and produce decisions nobody tracks

A category review at a grocery retailer is a scheduled negotiation with real money on it: range, space, promotional funding and listing fees for the next twelve months. Preparation looks like this. An analyst pulls syndicated market data from Circana or NIQ and exports to Excel. Someone else pulls internal POS (Point of Sale) movement and margin from the warehouse. Space data comes from whoever holds the planogram tool. The supplier arrives with their own deck built on their own view of the same market, and the first forty minutes of the meeting are spent arguing about why their share number does not match yours. The answer is always the same: their category definition includes a segment yours does not.

Three weeks of work produces a 90 slide deck. The review runs, decisions get made, and someone writes them in the minutes. Nine months later, in the next review, nobody can say whether the agreed range changes were actually implemented, whether the space shift happened in every store, or whether the funding committed against those changes was ever earned. The cycle starts again from a blank deck.

The money at stake is not the analyst time, although at a mid size retailer that is several full time equivalents doing slide assembly. The money is in decisions made against a market picture that is stitched together by hand every cycle, and commitments that nobody is holding anyone to.

Problem 1: two hierarchies that will never agree on their own

Your merchandising hierarchy exists to run a business: departments, classes, subclasses, built around how you buy, how you space and how you report. The syndicated hierarchy exists to describe a market and it is built around how a consumer shops and how the panel is measured. They do not line up, and they cannot be made to line up permanently because both change.

Circana and NIQ are the market view and you are not going to replace them, nor should you try. What they cannot do is see your loyalty data, your true landed margin, your store level space, or your private label economics. Their analytics layers answer questions about the measured market, and their delivery model is a subscription plus analyst hours. Blue Yonder and DotActiv sit on the other side, strong on space and assortment mechanics, and they will not tell you what happened in the market outside your four walls.

What a custom build does: hold the mapping as a maintained, versioned object rather than a spreadsheet that gets rebuilt. Every syndicated segment maps to your subclasses with an owner, a date and a reason. When a new segment appears, it lands in an unmapped queue that somebody clears, and every report built on the mapping either updates or flags itself as stale. This is the least exciting part of the build and it is the part that pays, because it is what removes the first forty minutes of the meeting and the three days of reconciliation before it.

Problem 2: the deck is the deliverable, so the analysis is never reusable

Because the artefact is a PowerPoint, every insight in it is dead on delivery. Next cycle, the same charts get rebuilt from newer data, by a different analyst, with slightly different filters, and nobody can reproduce last cycle's number to compare against it.

The category tools that exist sit at either end of the problem. Circana and NIQ produce reports you export. DotActiv produces space output. Neither owns the review as a repeatable process with a fixed definition of each measure, which is why every retailer ends up with its own Excel template that one person maintains and everyone quietly modifies.

What a custom build does: define the review once as a structured document with fixed measure definitions, then generate it from live data for any category on demand. Share, distribution, rate of sale per point of distribution, penetration, repeat rate, margin after funding, space to sales index and days of supply are all calculated the same way every time, by everyone. Deck generation is real here and it is where a language model does honest work: charts and tables come from the data model, and the model drafts the commentary underneath each exhibit, which the category manager then edits. It does not decide anything. It writes the paragraph that says volume fell 6 percent driven by the loss of distribution in two subsegments, which is a sentence an analyst currently spends an hour writing forty times.

Problem 3: supplier proposals arrive as slides and get read as opinion

Suppliers send range proposals as PowerPoint or Excel. A category captain arrives with a recommended shelf and a rationale. Your team either accepts a picture or rebuilds the whole thing to check it. There is no middle path, and the asymmetry means the captain's view carries more weight than it should simply because it arrives pre packaged.

What a custom build does: give suppliers a structured submission route, and parse the ones who will not use it. A proposal becomes a set of range change requests, each with the item, the action, the claimed rationale and the source data. Then your own numbers run against every proposed change automatically, so the meeting starts with your assessment of their proposal rather than their assessment of it. For suppliers who send a deck regardless, document extraction pulls the proposed range table out of the file and into the same structure. That is the difference between negotiating and being presented to.

Problem 4: agreed actions vanish between reviews

This is the failure with the largest financial consequence and the least attention. A joint business plan commits range changes, space changes, promotional slots and supplier funding against agreed volume targets. The plan is signed. Then delisting depends on the buying team, the space change depends on the next reset cycle, the promo slots depend on the promotional calendar, and the funding depends on finance raising the accrual and someone claiming it. Four different systems, four different owners, no shared status.

What a custom build does: every agreed action becomes a tracked object with an owner, a due date, a system of record where completion can be verified, and a funding line if money attaches to it. Range changes reconcile against item status in the merchandising system. Space changes reconcile against published planograms. Promotional slots reconcile against the promo calendar. Funding reconciles against what was actually accrued and claimed. Then the opening slide of the next review is not a market overview, it is a scorecard of what both sides committed to last time and what actually happened. That single change alters supplier behaviour more than any analysis in the deck.

Problem 5: the scorecard measures the category and ignores the shopper

Most category reviews are built on units, value, share and margin. The questions that decide range are shopper questions: does this item bring people into the category who buy nothing else, does delisting it lose the basket or just the line, is the growth in this segment new buyers or heavier buying by the same buyers. Those answers live in your loyalty data, which is usually not in the deck because the analyst could not join it in time.

What a custom build does: bring transaction and loyalty data into the same model as the market data, so penetration, repeat, source of growth and basket association are standing measures rather than a special request. Basket association in particular changes delist decisions, because the item with poor rate of sale that appears in high value baskets with nothing else in the category is not the item to cut, and no syndicated report will ever tell you that.

What this costs and how long it takes

A focused first release, meaning the hierarchy mapping engine, a fixed measure library, one generated review template covering three to five categories end to end, and the action tracker, runs $80,000 to $160,000 and ships in 12 to 18 weeks. A full platform adding supplier submission and proposal parsing, loyalty and basket analysis, space integration, funding reconciliation, and self service for the whole category team runs $200,000 to $500,000 phased over 8 to 14 months.

What drives the number up here specifically: how many syndicated data feeds you subscribe to and in what formats, since a NIQ delivery and a Circana delivery are separate ingestion projects; whether you are a retailer with your own POS data or a supplier working only from syndicated plus retailer portals, because supplier side means ingesting a different portal format per retailer; and the state of your merchandising hierarchy, since a hierarchy mid restructure means mapping a moving target. What keeps it down: start with the three categories where the money is and the category manager who most wants it, and generate exactly one review format rather than trying to satisfy every stakeholder's preferred layout in release one.

Build versus buy, and when buying is the right call

Buy if you run a small number of categories, if reviews are annual, or if your team is under five analysts. Your Circana or NIQ subscription plus a strong Excel template genuinely covers that, and a build will not return the money. Buy if your real problem is data access rather than assembly, meaning your analysts wait weeks for a warehouse query, because that is a data platform problem to fix first.

Build when two or more of these hold. Review preparation consistently takes more than two weeks per category and most of it is assembly rather than thinking. Nobody can produce a reliable status on the actions agreed in the last cycle. Your suppliers routinely present numbers you cannot reconcile with your own inside the meeting. Loyalty and basket data exists in the business but never makes it into a review. Or you are a supplier acting as category captain for multiple retailers and rebuilding the same analysis against four different retailer hierarchies every quarter, which is the single clearest build case in this category.

The honest tipping point is when the review process itself becomes the bottleneck on range decisions. If you are reviewing categories less often than you should because reviews are expensive to prepare, you are paying for the process in decisions you are not making.

How to choose a developer for category management software

Ask them how they will handle hierarchy mapping before anything else, and listen for whether they treat it as a versioned object with ownership or as a lookup table. If it is a lookup table, every historical comparison in your system will silently break the first time a segment definition changes.

Ask what they have actually ingested. A Circana feed, a NIQ feed, a retailer supplier portal export and a GS1 item feed are four different problems and experience with one does not transfer cleanly. Ask for the specific source and format, not a claim about handling data.

Ask how measures are defined and where those definitions live. Rate of sale per point of distribution calculated three different ways across three screens is the fastest way to lose the category team's trust, and once lost they go back to Excel and the project is dead.

Ask who owns the code and the data model, and settle it before kickoff. At Digital Heroes the client owns the repository and the infrastructure accounts from the first commit. Your hierarchy mapping and measure library are institutional knowledge about how your business defines its categories, and no agency should be holding them.