I deleted nearly a quarter of a nine-figure category and the team had its strongest year on record. The merchandising group hated me for a quarter. The P&L sent flowers in the next one.
This is the part of the job nobody puts on a slide deck: the case for less. Less choice. Fewer hero products. Tighter copy. A merchandising tree that doesn’t read like the index of an undergraduate textbook. I’ve sat in a lot of rooms where the answer to a weak quarter was more selection. The answer is almost never more selection.
§ 01 — The setup
The category was sprawling. Tens of thousands of active SKUs, multiple sub-categories, and a long tail that nobody had touched in years because nobody had time, and because the SKU intake was always faster than the SKU cleanup. The flywheel of “let’s just add it” had been spinning since 2014.
The hero products were buried. The customer journey looked like a librarian’s filing cabinet — technically organized, practically unusable. Our attach rate was flat. Our conversion was flat. Our return rate was creeping upward, which is the polite way of saying customers were buying the wrong thing and figuring it out later. Search relevance was getting actively worse as the catalog grew, because the surface area of “things that could match a query” was outgrowing the team’s ability to curate.
Every leader has a moment where they realize the answer isn’t do more. It’s do less, more loudly. This was mine.
§ 02 — The math behind less
People love to quote Sheena Iyengar’s jam study — 24 jam varieties on display converted at 3%, six varieties converted at 30% — and then proceed to do the literal opposite in their own product line. The instinct to hedge is stronger than the instinct to commit.
But the data on assortment overload isn’t fragile. It replicates across categories. McKinsey, Bain, and Forrester have all published recurring findings that customers shown a tightly curated set convert at meaningfully higher rates than customers shown a “complete” set — especially in categories where the customer is not a domain expert. The shape is consistent: above a certain threshold, every additional SKU lowers average conversion on the category page, because each new option adds cognitive load without adding meaningful incremental demand.
There’s a finance angle that gets ignored even harder than the conversion angle. Every SKU has a maintenance cost. Catalog ops, image refreshes, copy QA, returns processing, customer service escalations, inventory carrying cost if you stock it. In a nine-figure category where the worst tail products contribute fractions of a basis point to revenue, the unit economics on those SKUs are spectacularly negative before you count the conversion damage they’re doing to everything around them.
So the question isn’t “should we cut?” It’s “why didn’t we cut three years ago?”
§ 03 — How we decided what dies
I did not walk in with a hatchet. I walked in with a spreadsheet and a thesis: every SKU has to earn its slot, and most don’t.
The cut criteria, in order:
- Sales contribution. Pareto first. The bottom 30% of SKUs that contributed under 1% of revenue went on the watchlist. Not auto-killed — watchlisted, because some of them were strategic (price points, demographic coverage, search-intercept anchors).
- Attach rate and basket lift. Some low-revenue SKUs were genuinely useful — they pulled customers into the basket and those customers then bought the hero next to them. Those stayed.
- Return rate. Anything above the category median for returns got a hard look. High-return SKUs are not “popular.” They are expensive to ship twice.
- Search and browse traffic. SKUs that pulled zero direct search traffic and zero category-browse engagement over a full year were strong cut candidates regardless of revenue.
- Strategic hold. A handful of SKUs we kept for narrative reasons — entry price points, beginner-friendly options, the “trade-up” path. Defensible. Documented. The rest weren’t.
Twenty-two percent of the SKU set didn’t survive the filter. Some of them were vendor favorites. Some of them were originally launched by a leader three reorgs ago. None of them had a number that held up.
§ 04 — The internal fight
Every vendor PM whose SKU was on the chopping block had a version of the same argument: “But we still sell a few units.” Every vendor PM is correct, in the same way a person who keeps a 1998 receipt in their wallet is correct that technically it’s still legal tender.
The argument I had to win, repeatedly, in conference rooms with people who outranked me, is that the cost of a marginal SKU is not what it sells — it’s what it prevents the SKU next to it from selling. That second number is invisible on every standard dashboard. It only shows up in counterfactuals.
So I built the counterfactual. Matched-pair tests where we removed a marginal SKU from one storefront variant and held the other constant. The hero next to the deleted SKU lifted, every time, on average by mid-single digits. That’s not noise. That’s the tax we’d been paying to keep the museum open.
The second argument is psychological. Killing a SKU feels like killing a child. Adding a SKU feels like adding a feature. The brain accounts for them asymmetrically. As leaders, our job is to correct for that asymmetry on behalf of the customer and the P&L. Nobody outside the building cares about your SKU’s feelings. Customers care about whether they can find what they came for in under thirty seconds.
§ 05 — The receipts
After two quarters of execution:
- Attach rate up double digits in the cut subcategories. The hero products were finally visible, and customers who came for one thing left with two.
- Conversion up on browse pages, because there was less to wade through and the visual hierarchy actually meant something.
- Returns down in the affected subcategories, because customers were buying what they actually wanted instead of what they thought they wanted after thirty minutes of scrolling.
- Search relevance improved without any algorithm changes — because we’d thinned the candidate pool of irrelevant matches.
- Operations cost down in catalog ops, customer service, and reverse logistics. Small per-SKU. Meaningful in aggregate.
The category had its strongest year. The merchandising group, who’d hated me for the first quarter, sent me an invitation to the planning offsite for the next one. Nothing softens a former opponent like a number that moves.
§ 06 — The objections, answered
For anyone about to run this exercise inside their own organization, here is the short list of objections you will face and the responses that actually work, ranked by how often they came up.
“But it’s a long-tail strategy. The internet rewards infinite inventory.” Chris Anderson wrote a beautiful book in 2006. He was largely right about a specific economic moment. The argument has been over-applied for two decades, especially in categories where inventory carry, returns, and discovery friction are non-trivial. Long-tail economics work for digital goods with near-zero marginal cost. They work less well for physical inventory in a category where every additional SKU degrades the conversion rate on the SKUs you actually care about. Run the math on your category before you cite the meme.
“What if a customer wants the one we cut?” They will, occasionally. The customers who want a deleted SKU are, almost without exception, the same customers who would have bought something else from you anyway. The substitution rate on cut SKUs, when you measure it, is consistently high — usually north of 70% in apparel and home, often higher in consumables. The “we’ll lose the customer entirely” scenario is statistically rare. The “we lose two adjacent customers because of choice paralysis” scenario is the one bleeding you in the dark.
“The vendor will pull their entire line.” Sometimes. Less often than the vendor’s account manager threatens. The vendors who can credibly walk are the ones whose hero SKUs you would never have cut in the first place. The vendors threatening over a deletion in the bottom 5% of their own catalog are, in nearly every case I have seen, bluffing. Call the bluff on a small one first. The pattern usually corrects itself.
“We can’t afford to lose the revenue.” The revenue you lose from cutting a marginal SKU is, by construction, marginal. The revenue you gain from un-burying the hero next to it is not. The category math is almost always net-positive within one to two cycles. If you genuinely cannot survive a quarter of small disruption, you have a deeper problem than your SKU tree.
Every one of these objections is loud in the room and quiet on the spreadsheet. The job of leadership in a moment like this is to make sure the spreadsheet wins.
§ 07 — A worked example: where to cut first
The bottom 30% by revenue is the start, not the answer. The bottom 30% includes both the cuts you should make and the strategic holds you should defend. Triaging that list — fast, without losing nuance — is the real work.
The shortest version of the triage that worked for us, on a single SKU at a time:
- Does this SKU exist to anchor a price point? A cheapest-in-category option, an entry SKU for a beginner segment, a premium tier that frames everything below it as a good deal — these are strategic holds even when the units are slow. Defend them by name.
- Does this SKU intercept search intent we can’t get any other way? If customers are searching a specific keyword that lands them on this SKU and the SKU is the bridge to the rest of the catalog, it is doing brand work the dashboard cannot show. Defend it.
- Is this SKU pulling new customers in? Check first-time-buyer rate as a percentage of total buyers. A SKU with disproportionately high first-time-buyer share is an acquisition asset, not a long-tail straggler. Defend it.
- If none of the above — does the SKU have a story? Sometimes a vendor PM can articulate a real reason a SKU exists. A founder story, a category-defining piece, a halo product that elevates everything around it. The honest test is whether the story works without the speaker’s hand-waving. If it does, defend it. If it requires the speaker to be in the room to make sense, it doesn’t.
- If none of the above — cut. Move on. Do not relitigate. Do not workshop. Cut and observe what happens to the hero next to it.
This triage takes about ninety seconds per SKU once you have the data infrastructure for it. A thousand SKUs is a long week. A nine-figure category, run by a category manager who has not done this exercise in three years, is a six-week project — and one of the most leveraged six weeks of work that team will do all year.
§ 08 — The take
Every category in every retail organization on Earth is hoarding SKUs they’re afraid to kill. Some of it is institutional inertia. Some of it is fear of vendor blowback. Some of it is the perfectly rational human reluctance to be the person who killed someone’s pet project. All of it is a tax — the customer pays in cognitive load and the company pays in margin.
If you run a category, do this exercise this quarter:
- Pull the bottom 30% of SKUs by revenue.
- For each one, write down — in one sentence — the reason it exists. Not its sales number. Its reason.
- If you can’t write that sentence, it shouldn’t exist.
You’ll find that ten or fifteen percent of your catalog has no defensible reason to exist. Kill it. Watch what happens to the rest.
The hardest part of merchandising isn’t what you add. It’s what you have the discipline to remove. Less is louder. Less is faster to ship. Less is what your hero product needs to actually be the hero.
Math, not vibes.