Digital Merchandising: What to Build for Category Managers (and What to Skip)

Build ranking overrides, boost/bury rules, and collection curation first — these are the must-have merchandising controls that let category managers act on what they already know. Skip full automation of every lever; save engineering budget for guardrails that let humans override the algorithm when it's demonstrably wrong.

Quick Answer: Category managers need a small set of high-leverage controls — manual ranking overrides, boost/bury rules, and curated collections — with clear guardrails for when a human override should beat the algorithm. Full merchandising automation is a delighter, not a baseline; build the controls first, automate later.

Most merchandising roadmaps start from the wrong question. Teams ask "what can we automate for merchants?" when the sharper question is "what decisions do merchants need leverage over, and where does automation actively get in their way?" Category managers aren't asking for a robot that runs the store. They're asking for a small number of precise, trustworthy levers — and a system that doesn't fight them when they pull one.

This matters because merchandising software is easy to over-build. Every stakeholder has a favorite automation idea — dynamic pricing feeds, auto-generated collections, AI-written product copy — and each one sounds valuable in isolation. But a merchandising team's real constraint isn't ideas; it's a fixed engineering budget and a category manager who needs to hit a revenue number this quarter, not admire a roadmap. The Kano model, developed by Noriaki Kano in the 1980s to separate satisfaction-driving features from baseline expectations, is the right lens for sorting the two.

What Category Managers Actually Do (and Why It Changes the Backlog)

Category managers spend most of their week on three activities: curating what shows up in a collection, correcting what ranks where, and reacting to inventory or margin shifts in real time. Any merchandising tool that doesn't directly serve one of these three activities is solving a problem the category manager doesn't have.

This is worth stating plainly because most merchandising platforms are built by engineers who've never run a category. The instinct is to build a recommendation engine or an auto-tagging pipeline first, because those are the technically interesting problems. But talk to an actual category manager and the requests are mundane: "let me pin this SKU to position two before the Friday drop," "let me exclude this out-of-stock item without deleting the mapping," "let me see why this product outranks a better seller."

Jobs-to-be-done thinking clarifies this fast. A category manager doesn't hire merchandising software to "personalize the catalog" — they hire it to hit a sell-through target without babysitting every page manually. Reframing features around a manager's actual job (rather than a generic "improve conversion" goal) is the same discipline covered in more depth in our jobs-to-be-done framework for ecommerce, and it applies directly here: every merchandising feature should map to a specific job a category manager is trying to get done this week, not a hypothetical future workflow.

The Three Core Activities, Mapped to Features

The pattern across all three rows is the same: managers want control with a safety net, not a black box that occasionally asks for permission.

Collection Curation: The Foundation Layer

Collection curation is the single highest-leverage merchandising surface because it's the first thing a shopper sees and the easiest thing for a manager to get visibly wrong. Build a curation tool that supports manual pinning, exclusion, and reordering with an instant, WYSIWYG-style preview — this is a must-have, not a nice-to-have.

Most catalog teams default to rules ("show all in-stock items tagged summer, sorted by margin descending") because rules scale. But rules break at the edges every single week: a bestseller goes out of stock and needs immediate removal, a supplier renegotiation makes one SKU suddenly more profitable than its rule-based rank suggests, or a marketing push needs three specific SKUs above the fold regardless of what the rule says. A pure rules engine with no manual override turns every edge case into a support ticket.

The fix isn't choosing rules or manual control — it's layering them correctly:

1. Rule-based default ordering handles the 80% of SKUs nobody has an opinion about.
2. Manual pins override the rule for a small, visible set of positions (typically the first 6-12 slots).
3. Exclusions remove an item from a collection without touching its underlying catalog data or tags.
4. An expiry or review date on every manual override, so a pin made for a holiday promotion doesn't silently persist into March.

That fourth item is the one most teams skip, and it's the one that causes the most long-term mess — stale pins from a campaign eighteen months ago quietly suppressing a collection's natural sort order, with nobody remembering why.

Ranking Overrides and Boost/Bury: The Controls Merchants Actually Reach For

Ranking overrides and boost/bury rules are the two controls category managers use most often, and they should be built as first-class, auditable actions — not buried three menus deep as an "advanced settings" escape hatch. This is where the automation-vs-control tension is sharpest, because the underlying ranking algorithm (relevance, popularity, margin-weighted, or some blend) is usually the most engineered part of the system, and every override feels like second-guessing it.

Boost and bury rules should be scoped, time-boxed, and visible. A boost that says "raise this SKU's rank by 15 positions for the next 14 days" is a fundamentally different — and safer — object than a boost that silently edits the underlying relevance score with no expiry. The former is inspectable and reversible; the latter accumulates as undocumented technical debt inside the ranking model.

A useful design constraint: every override should be visible to every other category manager, not just the one who created it. Merchandising teams are usually small, but a boost or bury rule created by one manager and forgotten by everyone else — including its creator, six months later — is a recurring source of "why is this ranking so weird" investigations. A simple audit log with creator, reason, and expiry date solves most of this for very little engineering cost.

Boost/Bury Design Decisions Worth Getting Right Early

Ranking logic itself — how relevance, popularity, and query intent get blended in the first place — is a deep topic on its own; if you're building or auditing the underlying algorithm rather than the override layer sitting on top of it, our guide on site search relevance and query understanding covers that layer in more detail. The override tools described here should sit cleanly on top of whatever ranking engine exists underneath, regardless of how sophisticated it is.

The Automation-vs-Control Tension, and a Guardrail Model for Resolving It

The core tension in merchandising software is that automation improves average outcomes while overrides protect specific outcomes a manager can't afford to get wrong — and a good system needs both, with a clear rule for which one wins when they disagree. The mistake most teams make is treating this as an either/or roadmap decision instead of building an explicit guardrail model that specifies when a human override should beat the algorithm.

A workable guardrail model has three tiers:

1. Automated by default — day-to-day sort order, personalized ranking for logged-in shoppers, out-of-stock suppression. The algorithm should own these unless something is actively wrong; humans shouldn't need to touch them for the system to work.
2. Override with visibility — boost/bury, manual collection pins, margin-weighted re-ranks during a promotion. The algorithm proposes, the manager can override, and the override is logged and time-boxed as described above.
3. Human-only, no automation — brand-safety exclusions (e.g., never surface a recalled or legally restricted product), contractual placement commitments to a supplier, and anything touching pricing displayed to a regulator-sensitive category. These should never be reachable by an automated rule at all, regardless of how confident the model is.

The guardrail question to ask for every new feature request is: "what does the algorithm not know that the human does?" A category manager often knows things no ranking signal captures — a supplier relationship is ending, a product photo is temporarily wrong, a regional compliance issue makes an item unsellable in one market. Automation should defer whenever the human's information advantage is structural, not just occasional.

This tiered model also gives you an honest way to talk to stakeholders who want "full automation" — you can point at tier three and say plainly: some decisions should never be automated, and building an override path for them isn't a fallback, it's the correct design.

A Kano-Based Prioritization of Merchandising Features

Sorting merchandising features through a Kano lens — basic expectations, performance features, and delighters — turns a long backlog into a short, defensible build order, and it's the single most useful exercise for a team with a fixed budget deciding what category managers actually need first.

Kano's framework, built from structured customer surveys pairing a "feature present" and "feature absent" question, sorts requirements into three buckets that behave differently as you invest more in them:

• Must-haves (basic expectations): absence causes real dissatisfaction, but presence doesn't create delight — it's simply table stakes. Manual collection curation and ranking overrides sit here. A category manager doesn't celebrate having them; they leave if they don't.
• Performance features (one-dimensional): satisfaction scales roughly linearly with how well they're built. Boost/bury rule sophistication, override audit logging, and preview speed all fall here — better execution directly improves the manager's day.
• Delighters (attractive/exciters): absence isn't noticed, but presence creates disproportionate enthusiasm. Auto-generated collection suggestions, predictive re-ranking ahead of a demand spike, or AI-drafted merchandising copy are delighters — genuinely nice, but not what retains a category manager if the basics are broken.

Applying Kano to a Real Merchandising Backlog

The practical lesson from Kano research (and echoed in later product-prioritization work, including the RICE framework popularized by Intercom) is that delighters have a way of becoming table stakes over time — today's exciting AI ranking suggestion is next year's expected baseline. That's not a reason to skip must-haves in favor of delighters; it's a reason to nail the must-haves fast, because the delighter bar keeps rising regardless of what you build.

This is also where a structured Kano exercise pays for itself against a stakeholder who insists every idea is a "must-have." Running the survey-based methodology, or a lightweight internal proxy of it, forces the "everything is critical" conversation into an actual sorted list — which is exactly the exercise Prodinja's Kano analysis is built to walk a PM through: separating the must-have merchandising controls from the delighters when you're deciding what to build for category managers on a fixed budget.

Where Merchandising Sits in the Broader Commerce Experience

Merchandising tools don't operate in isolation — a ranking override upstream only matters if the downstream experience (search, recommendations, checkout) doesn't undo the manager's intent. It's worth briefly placing merchandising in that larger system before finalizing a build list.

A boosted product that a shopper never sees because on-site search re-ranks it away, or a curated collection that gets bypassed entirely by a "people also bought" recommendation module, defeats the point of the override. If you're also scoping recommendation logic, our piece on AI recommendations beyond "people also bought" covers how those modules should respect (or knowingly override) merchandising intent rather than silently contradicting it. Similarly, a beautifully curated collection that leads into a broken cart flow loses the sale anyway — see our guide on checkout flow optimization for the highest-stakes part of that journey. And because a shopper's path through collections, search, and product pages is really one continuous experience, it helps to map merchandising decisions against the broader customer journey rather than treating each surface as an independent backlog. For teams scoping merchandising as part of a wider ecommerce platform rebuild, our ecommerce and retail complete guide is the right starting point for sequencing that larger roadmap.

Key Takeaways

• Build manual controls before automation — collection curation and ranking overrides are Kano must-haves; category managers notice their absence far more than they celebrate their presence.
• Time-box every override — a boost, bury, or pin without an expiry date accumulates into undocumented technical debt inside the ranking model.
• Make overrides visible to the whole team — an audit log with creator, reason, and expiry prevents duplicate or conflicting rules across managers.
• Use a three-tier guardrail model — automated by default, override with visibility, and human-only-no-automation for brand-safety and compliance decisions.
• Sort the backlog with Kano, not opinion — must-haves get built first regardless of how exciting a delighter sounds in a roadmap review.
• Remember delighters drift toward baseline over time — today's AI-powered ranking suggestion becomes next year's expected feature, so nail the fundamentals fast.
• Merchandising doesn't work in isolation — check that search, recommendations, and checkout don't quietly undo a category manager's override.