Should I keep, discount, or discontinue this SKU? (small business)

Top picks

Protocol

1. 1

   Step 1 of 6

   Paste facts (5–8 min).

   Assemble only what you actually know. Mark gaps [MISSING] / [NEED FACT]. Do not invent rates, BATNA, legal outcomes, or benchmarks. This small-business line-level existence gate is not a job quote and is narrower than a whole-catalog price raise. No invented market share, elasticity, or clearance lift percentages.

   Paths at a glance

   Path	Lean this way when (from your paste)
   A — Keep	Margin and demand meet your floors; support cost acceptable; strategic reason to retain.
   B — Raise price or shrink scope	Demand exists but margin/support miss floors—fixable without killing the line.
   C — Clearance / sunset	Exit needed but inventory/customers need an orderly wind-down you can staff.
   D — Discontinue now	Zombie line: thin demand, poor margin, high support—and delay costs more per your notes.

2. 2

   Step 2 of 6

   Lock decision-tree rules (step 2).

   Paste the rules first so the model cannot mint market share, competitor SKUs, or clearance lift %.

   Copy-paste prompt

   Copy rules

   Help me run a KEEP / DISCOUNT-OR-FIX / DISCONTINUE SKU decision tree.
   Hard rules:
   - Use ONLY margin, demand, and support facts I paste. Do not invent market share, competitor SKUs, or clearance lift %.
   - Paths: Path A Keep / Path B Raise price or shrink scope / Path C Clearance then sunset / Path D Discontinue now.
   - Optional customer sunset note ONLY after I choose C or D, using only dates/offers I authorize.
   - Not legal advice on cancellations. I decide.
   Confirm in one sentence, then wait for my pack.

3. 3

   Step 3 of 6

   Fill pack (step 3a) and generate (step 3b).

   Paste the pack, then send the generate command. The model must not pick a path for you.

   Paste pack

   Copy fact pack

   SKU / SERVICE NAME: [ ]
   REVENUE / UNITS (period): [or MISSING]
   GROSS MARGIN: [or MISSING]
   MY MARGIN FLOOR: [or NEED FACT]
   SUPPORT COST / TIME DRAIN (facts): [or MISSING]
   DEMAND TREND NOTES: [ ]
   STRATEGIC REASON TO KEEP (if any): [or NONE]
   INVENTORY / CONTRACT OBLIGATIONS: [or NONE]
   CUSTOMER COMMUNICATION CONSTRAINTS: [or MISSING]
   PRICE RAISE OR SCOPE SHRINK I COULD TRY: [or NEED FACT]

   After you send the filled pack, copy the Generate command in the next box and paste it into the SAME chat so the AI runs the decision tree / produces the output.

   Generate command — copy and send this next (same chat)

   Copy generate command

   Build the decision tree now:
   1) Facts vs [MISSING]
   2) Scorecard table: Criterion | My fact | Lean (+/0/− or NEED FACT) | Note
   3) Path A Keep / Path B Raise price or shrink scope / Path C Clearance-sunset / Path D Discontinue now — for each: when it fits, what must be true from my pack, 3 owner-verify bullets
   4) Do NOT pick for me. Do not invent facts.

4. 4

   Step 4 of 6

   Choose a path (step 4); optional wording.

   Only after you pick a labeled path. Owner-verify: facts still match your pack; [MISSING] gaps filled or accepted; any sendable wording uses only terms you authorize.

   Path wording

   Copy path wording

   I chose Path [A/B/C/D]. If C or D, draft a short customer/sunset note using only dates, substitutes, and offers I authorize: [ ].
   No invented discounts. Mark [NEED FACT].

5. 5

   Step 5 of 6

   Second-model check (step 5).

   In the other tool, paste the tree plus your pack:

   Second-model check

   Copy second-model check

   Here is a keep/discontinue SKU tree from another assistant plus my pack.
   Find: (1) invented market share or clearance %, (2) discontinue forced despite strategic keep reasons I listed, (3) offers not in my authorize list, (4) confusion with whole-catalog price raise.
   Return only corrections.

6. 6

   Step 6 of 6

   Human decide and act (step 6).

   Pick one path. Fill [NEED FACT] gaps from real sources. Send or implement yourself—AI drafts are not approvals, contracts, classifications, or tax elections. If the line stays and you still quote it, use the service-estimate how-to. Whole-catalog pricing is a different tree.

When not to

• Do not invent market share, elasticity, or competitor catalog moves.
• Contracted deliverables or regulated product withdrawals—follow those rules/counsel.
• Skip if margin and support time are unknown—measure one period first.
• Whole-catalog pricing belongs on the raise-prices recipe instead.