Selling on Wildberries and Ozon Without the Night Shift: 10 Prompts That Handle Assortment, Pricing, and Supply
A marketplace seller's day rarely ends when the office closes. You finish a supply plan, then realize the commission changed overnight. You open the unit economics spreadsheet, and the numbers no longer add up because logistics tariffs shifted again. Multiply this by 40 SKUs across two platforms, and you are not running a business — you are running a treadmill.
The real bottleneck is not the work itself. It is the context switching. Every task — describing a new SKU, recalculating margin, drafting a claim for a lost shipment — forces you to reload a different mental model. That is exactly where large language models earn their keep. Not as a magic button, but as a tireless analyst that never forgets the formula you gave it three weeks ago.
Below are ten prompts I use for Wildberries and Ozon operations. Each one is copy-paste ready, but the value is in the reasoning behind it. Treat them as templates: paste your own category, tariff, and SKU data, and the model does the heavy lifting. A word of caution upfront — never paste real customer personal data or your API keys into a public chat. Use placeholders, and verify every number the model returns against the official seller portal.
1. Unit Economics Recalculation After a Tariff Change
Marketplace commissions, logistics, and storage fees change constantly. The prompt forces the model to show its work instead of guessing.
You are a marketplace unit-economics analyst. Recalculate margin for the SKU below.
Inputs: selling price, cost of goods, WB commission %, logistics per item,
storage per item per day, average days in warehouse, acquiring %, tax regime.
Rules: show every line as a formula, then the final margin in absolute and %.
Flag any input that looks unrealistic. Do not invent missing numbers — ask.
Example: feed it a 1,490 RUB item with 25% commission and 180 RUB logistics. The model returns a line-by-line breakdown and flags that storage cost was omitted.
2. Assortment Audit: Which SKUs to Kill
Dead stock quietly eats your storage budget. This prompt turns a raw sales export into a decision list.
Here is a CSV export of my SKUs: article, category, sales last 30 days,
returns, current stock, storage cost. Classify each SKU as: grow, hold, fix,
or liquidate. Justify each verdict with the actual numbers from the row.
Output a markdown table sorted by priority.
3. Card Description That Survives Moderation
Marketplace moderation rejects cards for vague or keyword-stuffed text. This prompt writes copy that reads like a human wrote it.
Write a product card for [product]. Audience: [buyer]. Include: title under
60 chars, 5 bullet benefits, 1 paragraph description, and 5 SEO keywords
naturally embedded. No superlatives like "best" or "unique". No medical claims.
4. Competitor Price Monitoring Script
Manually checking rivals is the fastest way to lose an afternoon. This prompt generates a parser skeleton you can run yourself.
Write a Python script using requests and BeautifulSoup that reads a list of
product URLs, extracts price and rating, and writes results to CSV.
Add a 2-second delay between requests and a realistic User-Agent header.
Explain which parts may break if the site loads data via JavaScript.
The honest caveat matters: dynamic storefronts often render prices client-side, so a plain HTTP request may return nothing. The prompt asks the model to say so rather than pretend the script always works.
5. Supply Plan From Sales Velocity
Overstocking locks cash; understocking kills ranking. This prompt builds a replenishment plan from velocity data.
Given: average daily sales per SKU, lead time from supplier in days,
safety stock days, current warehouse balance. Build a supply plan for the next
4 weeks. Show formula: needed = (daily sales * (lead time + safety)) - balance.
Mark SKUs where balance already exceeds 6 weeks of demand.
6. Return-Rate Root Cause Analysis
High returns usually trace back to one of three causes: sizing, photos, or description mismatch. This prompt separates them.
Here are return reasons as free text for one SKU. Cluster them into themes,
count each theme, and rank by frequency. For the top theme, propose 3 concrete
fixes (photo, size chart, description wording). Do not fabricate counts —
base them only on the text provided.
7. Promotion Participation Calculator
Flash sales promise traffic but can destroy margin. This prompt decides whether to join.
Promotion offers X% discount. My current margin is Y%. Logistics and commission
stay fixed. Calculate break-even discount and the sales volume increase needed
to keep the same absolute profit. Show the math step by step.
8. Claim Letter for Lost or Damaged Goods
A well-structured claim gets resolved faster. This prompt drafts it in the right tone.
Draft a formal claim to the marketplace logistics department. Facts: shipment ID,
date, missing quantity, documented evidence. Tone: factual, no emotion.
Structure: subject, facts, legal reference to the offer agreement, requested action.
Leave placeholders for data I must verify.
9. Review Response Templates by Sentiment
Reviews shape ranking, but copy-paste replies look robotic. This prompt builds a tiered response system.
Create response templates for 5 review types: 5-star, 4-star, 3-star neutral,
negative about quality, negative about delivery. Each under 300 chars, warm but
not over-apologetic. No promises you cannot keep.
10. Weekly Ops Digest for the Team
If you have a small team, a weekly digest keeps everyone aligned.
From this week's metrics (sales, returns, stock alerts, price changes), write a
short digest: 5 bullet wins, 3 risks, 2 decisions needed. Plain language, no jargon.
What Actually Changes
The pattern across all ten prompts is the same: you stop being the calculator and start being the decision-maker. The model handles the arithmetic, the drafting, and the first pass of analysis. You handle judgment, verification, and the calls that carry real money.
Two habits make this work long term. First, keep a personal prompt library with your tariff assumptions baked in — the model is only as good as the context you feed it. Second, always verify numbers against the official seller portal before acting; an LLM is an assistant, not an accountant of record. Start with prompt #1 today, and by next month the night shift becomes optional.
Comments