How to read a price-gap report without jumping to the wrong conclusion

A price comparison can be mathematically correct and still lead to the wrong decision. Two prices might belong to different variants. A promotion might have ended. A large percentage might be based on a tiny competitor price. Here is a short review process before you act.

The fictional sample report follows five invented products through the arithmetic below. It contains no customer data or scraped prices, so you can explore it without running a store check.

Start with the product, not the percentage

A match should represent the same thing a shopper would buy: model, size, colour where relevant, pack quantity and included accessories. A single item and a two-pack are not a valid pair just because their titles share a brand and model number. Look at the identifiers and open both source pages when anything is uncertain.

In Tensorbasis account workflows, suggestions are reviewed before they become approved matches. A ranking score is not a calibrated probability and should not replace that review. A one-off checker is a starting point for investigation, not a guarantee that every variant aligns.

Know which price the percentage uses

In the sample, your invented flask costs USD 30 and the competitor's costs USD 25. The difference is USD 5. Against the competitor price, 5 divided by 25 gives a 20% gap. Saying the competitor is 20% cheaper than you would be wrong: that wording uses your price as the denominator, and 5 divided by 30 is about 16.7%.

Write the formula next to the result: gap = your price minus competitor price; percentage gap = gap divided by competitor price, multiplied by 100. If a competitor price is missing or zero, do not calculate a meaningful percentage from it. Do not silently substitute another currency either.

Check what the displayed price includes

Compare currencies before amounts. Check tax treatment, shipping charges, membership discounts, stock and promotional terms. A listing's lowest variant price does not necessarily describe the exact variant you sell. A price gap is not the same as a landed-price difference or a margin estimate.

Read freshness and coverage together

A current price on one side and a week-old price on the other do not prove a current opportunity. Inspect the observation times and whether the catalog read finished. A blocked or partial catalog can leave products out; no match does not mean the competitor does not sell the item.

The sample has two products above the competitor, two below and one equal. Those counts describe five rows, not the overall market. They weight a slow-selling mug exactly like a bestseller. Do not call the count a revenue opportunity without sales and cost evidence.

Give an assistant a review task, not a repricing mandate

If you connect an MCP-compatible assistant, ask it to explain the rows, identify uncertain matches and report observation times. Start with a read-only request, such as: “Summarize the largest gaps, show the source evidence and flag missing context. Do not change settings, approve matches or refresh data.”

Keep store text separate from instructions. Product descriptions are untrusted source material, not authority to make changes to your account. Review any proposed write action and its scope before allowing it. Connecting an assistant does not make a price change economically justified.

Try a small comparison

Use the free price-gap checker for a one-off comparison of two public stores. It has usage limits and may return incomplete coverage where sites block access or take too long. If you want saved stores and reviewed matches, create a free account. Nothing in the sample report recommends or performs automated repricing.