Find the products that are worth promoting

Stop asking what sells. Start asking what’s worth selling right now.

Unit count on its own tells you almost nothing. Five hundred units could be three days of inventory or six months of it, and your ad platform can’t tell the difference.

Neither can you, until you divide by something.

So this tactic starts with the right export and ends with one number from 0 to 100 for every SKU.

High score, the product is ready to scale today. Low score, your ad budget is working against the business. Politely, efficiently, and at volume.

Fair warning: the score is only as good as the data underneath it. Garbage in, confidently wrong recommendations out.

Do it now:

• Pull everything into one spreadsheet: Inventory levels, sales from the last 7 and 30 days, price, product cost, shipping cost, return rate, and lead time on your next delivery. Most of it is already sitting in your e-commerce platform. For the rest, go ask whoever handles purchasing.
• Calculate days of inventory: Divide available stock by average daily sales. Five hundred units moving 100 a day is five days of cover, not “plenty of stock.”
• Flag the out-of-stock days: Zero sales during a stockout is not zero demand. Skip this step and you’ll punish your best products for being popular.
• Assign weights so the AI knows what you care about: For example: 25% margin after costs, 25% inventory coverage, 20% sales velocity, 15% returns, 10% seasonality, and 5% delivery confidence. A fashion store and a supplement store should not weigh seasonality the same way. Adjust accordingly.
• Ask for three outputs at once: The score, the recommendation, and the biggest risk (such as low inventory, weak unit economics, slow sales, or an unreliable delivery date).
• Refresh on a schedule: Daily or weekly, depending on how fast your inventory actually moves.

Use our base prompt to get started:

You are an e-commerce analyst specializing in planning advertising campaigns based on inventory and financial data. Below is product data in CSV format. Each row represents one SKU and contains the following columns [read more…]

Pro tip: Only count inventory you can genuinely sell. Subtract safety stock, pending orders, damaged units, and anything reserved for other sales channels.

The number that survives that subtraction is the real one.

Turn product data into five simple actions

Nobody on your marketing team should be reading margin reports and delivery schedules for four hundred SKUs before lunch.

They need one answer per product: scale, maintain, sell faster, or stop. That’s it.

So let AI do the sorting. Every product gets assigned to one of five modes, and the assignment updates itself whenever sales, inventory, or delivery timing shifts.

Step by step:

• Scale: Healthy margin, stable sales, enough inventory, and a restock date you’d bet on. Raise the budget gradually, widen the audiences, and get new creatives in the queue.
• Maintain: Profitable and consistent, but there’s no case for pushing harder. Hold the budget where it is and keep an eye on performance and availability.
• Clear: Inventory is running well ahead of sales velocity, or the season is closing in on you. Test bundles, multipacks, free gifts, and controlled promotions before you reach for a real discount.
• Protect: The product sells well and may not survive until the next shipment. Pull back on broad prospecting, keep branded campaigns and remarketing running, and start warming up an alternative.
• Pause: Out of stock, unprofitable after returns, sitting under a quality issue, or carrying broken feed data. Get it out of your campaigns before the algorithm starts buying clicks that can never become useful sales.

Then commit the decision somewhere the whole system can see it. Add the segment as an inventory_segment column in your spreadsheet and as a label in your product feed.

From there, let AI write the recommendation for every SKU with our prompt:

You are an experienced e-commerce and performance marketing analyst. Your task is to analyze inventory, sales, and financial data and assign each product to exactly one of five advertising segments:[read more…]

The data backs this up: According to Deloitte, six in ten retail buyers said AI-powered tools improved demand forecasting and inventory management.

Three-quarters of the executives surveyed expected AI to handle dynamic pricing based on demand and competitor activity.

Match the ad to the inventory situation

Once a product has a segment, nobody needs another inventory report. The analysis is done.

What’s missing is a short set of instructions for the person actually making the ad.

AI can turn the same product data into a messaging card: one objective, the angles that are fair game, the proof that supports them, and the claims to leave alone.

Hand that to a creator and they know immediately whether this ad should expand demand, accelerate sell-through, steer people toward an alternative, or filter the audience more carefully before anyone reaches checkout.

Playbook:

• Give AI the full picture: The product segment plus current price, availability, margin, return rate, product limitations, and the ads that are already performing.
• Ask for one messaging card: Advertising objective, recommended angle, main proof point, biggest risk, and prohibited claims. One card, five fields, no essay.
• For products ready to scale, go find new demand: New use cases, new buying situations, new audiences. Not a fourth variation of the ad that already works.
• For excess inventory, raise the offer value first: Bundles, multipacks, free gifts. A discount is the last lever you pull, not the first one you reach for.
• When inventory is low, stop creating demand you can’t fill: Point customers to an alternative product, a waitlist, or a back-in-stock notification instead.
• When returns are high, qualify harder: Spell out sizing, compatibility, how the product is meant to be used, and who it isn’t for. A sale you refund twice is worse than a sale you never made.

Use our prompt to create a ready-to-use messaging card and creative brief:

You are a senior performance marketing strategist and e-commerce copywriter. You specialize in transforming inventory, financial, and sales data into clear advertising communication guidelines [read more…]

Case study: In a report on excess inventory in the fashion industry, AI was used to plan demand, deliveries, and purchasing budgets against current sales.

Kering reported a 20% improvement in inventory forecasting accuracy.

Before you touch that budget

Don’t scale a product just because it sells.

Check the margin, the inventory you can actually sell, the velocity, the return rate, and the next restock date.

Turn that into five decisions: scale, maintain, clear, protect, or pause. Now your marketer knows where to push, where to hold, and where to walk away.

Then match the message to the situation. Excess inventory means raising offer value. Low stock means limiting demand. High returns mean qualifying buyers harder.

That’s the whole idea. Connect inventory, finance, and advertising data, then turn it into decisions your campaigns can act on.

And no, you don’t need a system for this. One spreadsheet, one score, and segments you keep updated will move the needle.