Variant Triage for Merchandisers: Cutting 200 Concepts to a 12-SKU Drop in One Afternoon

Variant triage is the merchandiser's discipline of turning a large set of AI-generated design directions into a buy-ready assortment in a single working session. The premise is simple. The design team produces volume. The merchandiser produces conviction. This piece is the afternoon framework: a four-step rubric, a 2x2 triage matrix, and the four cuts that take 200 concepts down to 12 SKUs without losing the strongest ideas. It is written for the in-house merchandiser who already owns sell-through, margin, and assortment plans, and wants a repeatable way to absorb AI-scale concept volume without drowning in it.

The reason this matters now is that concept generation has stopped being the constraint. A design team running an AI sprint can produce 80 to 200 directions on a brief in days. Without a triage system, the merchandiser becomes the bottleneck and the drop slips. With one, the merchandiser becomes the use point and the drop ships faster than it ever did under the old cycle.

Why 200 concepts is the new starting point

Under the old cycle, the merchandiser saw 10 to 15 concepts after the design team had already done the cull. The implicit pre-filter was time. Designers could only draw so many. Now the design step is cheap, so the pre-filter has to move. The merchandiser receives a larger set, but the set arrives with consistent metadata: a base spec, a fabric platform, a price band estimate, and any color and trim variants the design team explored. The triage is faster because the inputs are uniform.

If your team is still receiving concepts as mood-board collages with no spec attached, fix that first. Triage on uniform inputs is what makes the afternoon framework work. Triage on inconsistent inputs is just opinion in a room.

The afternoon framework, end to end

Three hours, end to end. The framework holds as long as the prework is done. The prework is the spec uniformity covered above, plus a current assortment plan with open slots named by category, price band, and target margin. If the slots are not named going into the session, the session turns into a workshop, not a triage.

Step 1: Hard cut on brief fit

The first 30 minutes are solo work for the merchandiser. Cut anything that fails brief fit on three dimensions: brand voice, fabric platform, and price band. Brand voice cuts are the fastest, the merchandiser knows them on sight. Fabric platform cuts come from the spec: if the concept calls for a fabric the brand is not running this season, cut. Price band cuts come from the spec estimate: if the SKU lands more than 15 percent outside the target band, cut. By the end of 30 minutes, the set is usually around 90.

Step 2: Margin and price-band scoring

Bring in the planner. Score each of the remaining 90 from 1 to 5 on margin fit. Margin fit is target margin minus estimated margin for the SKU, normalized so 5 is on-target and 1 is more than 8 points below. The planner reads the spec, the merchandiser reads the assortment slot, and the score is mutual. Set a floor: anything scoring 1 or 2 drops out unless it has a clear demand reason to survive Step 3. The set lands around 50.

Step 3: Demand signal overlay

Bring in the data lead, or do this solo if your team is small. Pull the demand signal for each of the 50: sell-through index on the closest historical SKU, search trend for the named category over the last 90 days, and any first-party signal from the brand's own waitlist or wishlist. Score 1 to 5 on demand. The output is 30 concepts with two scores: margin fit and demand signal.

Step 4: The 2x2 triage and the assortment build

Plot all 30 on the 2x2. The four quadrants drive four different calls.

• Greenlight, top right. High margin fit, high demand. These are the spine of the drop. Most drops fill 8 to 10 of the 12 slots from this quadrant.
• Repricing test, top left. Strong demand, weak margin. Ask the planner if a fabric swap or trim simplification gets the margin to a 4. If yes, promote to Greenlight. If no, hold.
• Hold for next drop, bottom right. Strong margin, weak current demand. These often become next season's hero pieces when the demand signal turns. Park them in a named queue.
• Cut, bottom left. Both scores weak. No drop slot, no future slot, no test. Cut without a meeting.

The assortment build is mechanical once the quadrants are drawn. Walk the named slots in the drop plan one at a time. For each slot, pick the highest combined score from Greenlight that fits the slot. If a slot has no Greenlight candidate, pull the strongest Repricing test, run the swap, and re-score. Name an alternate per slot in case the tech pack hits a manufacturability issue in Days 12 to 14.

What 12 SKUs from 200 concepts actually looks like

The picture above is what the merchandiser walks out of the room with: 12 SKUs across four categories, one per slot, each one already scored on margin and demand, each one with an alternate behind it. The set goes to the design and production team for tech pack handoff with a single decision log attached. The log lists which concepts were cut at each step and why. If anyone asks why a specific concept did not make it, the answer is on the page.

The four cuts that go wrong

1. Skipping the hard cut on brief fit. If the merchandiser tries to score 200 on margin, the session is over before it starts. Cut on brief fit first, always.
2. Letting demand signal override margin without a repricing test. A high-demand SKU at a 1 on margin is a sell-through win and a P&L loss. The Repricing test exists for this exact case.
3. Greenlighting more than 12. The point of the triage is a focused drop. If the Greenlight quadrant has 18 candidates, the brief was too broad. Note it for the next sprint and still ship 12.
4. Failing to name alternates. Tech pack issues happen. If a slot has no alternate, the slot ships empty or the drop slips.

How this connects to the merchandising operating model

The afternoon triage is one routine inside a larger merchandising and launch model. The model covers assortment planning, drop calendar, pricing logic, and post-launch read. Our AI merchandising and launch pillar walks through how the routines stack and where the merchandiser sits at each step. If your team is running the triage without the surrounding model, you will get a good drop and lose the institutional memory between drops. The pillar fixes that.

Frequently asked questions

Can a small team run this without a planner or a data lead?

Yes. A merchandiser working solo can do all four steps in a single afternoon by carrying the planner and data lead views as checklists. The output is slightly less defensible to a CFO, but the drop quality holds.

What if our drops are larger than 12 SKUs?

The framework scales by category, not by total SKU count. Run the triage per category with its own slot count. A 36-SKU drop with three categories runs as three triages, each producing 12 SKUs against named slots.

How often should we recalibrate the scoring?

Every two drops. After two drops, compare actual sell-through and actual margin on Greenlight SKUs to the scores they received. If the correlation is weak, the rubric needs adjustment. If the correlation is strong, leave it alone.

Does the merchandiser need to be in the room with the creative director for Step 4?

Yes. The 2x2 is where commercial logic and taste meet. The creative director catches the cases where a high-scoring SKU dilutes the brand voice of the drop, and the merchandiser catches the cases where a beautiful concept will not sell. Both calls have to happen at the same table.

Further reading

• AI merchandising and launch, the operating model around the triage
• Pre-production workflow, what happens to the 12 after sign-off
• AI fashion workflow software, the system the design sprint and the triage sit inside

Run the next triage with us

If you want to run the afternoon framework on your next drop with our team in the room, book a 30-minute working session. Bring the concept set and the assortment plan, and we will walk the rubric with your merchandiser end to end. Book a working session.

Related: Merchandising & Launch