The collection is the easy 30%
Fetching competitor prices is straightforward engineering. What sinks these projects is everything after: knowing which of their products corresponds to which of yours, handling variants and bundles, and turning a wall of numbers into something a human acts on.
Budget accordingly. If a proposal is mostly about collection, it is describing a third of the work.
Product matching is the actual project
The same item appears as different names, codes, pack sizes and bundles across retailers. Matching needs a combination of approaches and a human in the loop.
- Exact identifiers first — EAN, GTIN, MPN where available. Reliable and often missing.
- Normalised text matching on brand, model and specification.
- Semantic similarity for descriptions that differ in wording but not meaning.
- Human review for the middle band, with the decision remembered so it is never asked twice.
Expect 60–80% automatic matching on a typical catalogue, with the remainder needing review. Anyone promising fully automatic matching across messy retail data has not tried it on your catalogue.
Handle variants and bundles deliberately
A 500ml bottle is not a 1-litre bottle, and a three-pack is not three singles. Normalise to a comparable unit — price per litre, per unit, per kilogram — or your comparisons will be confidently wrong.
Bundles are harder still and worth an explicit decision: exclude them, flag them separately, or estimate a component price with a clear caveat. Pick one and document it.
Alert on decisions, not on changes
A daily email listing every price movement gets filtered within a fortnight. Alert only when something crosses a threshold that would change what you do.
- A competitor drops below your price on a product where you compete on price
- You are now more than X% above the market on a high-volume line
- A competitor is out of stock on something you hold — often the most actionable signal available
- A sustained trend rather than a one-day promotion, which is usually noise
Watch stock as well as price
Competitor stock-outs are frequently more valuable than price movements and are consistently overlooked. A rival out of stock on a line you hold is a short window where advertising spend and search position are worth more than usual.
Collecting availability alongside price adds little to the build and often produces the highest-return alerts in the system.
Put the output where decisions happen
A separate dashboard is visited enthusiastically for three weeks and then forgotten. Feed the data into the tool where pricing decisions are actually made — the pricing sheet, the ERP, or a scheduled summary in the channel the team already uses.
The measure of success is whether prices changed as a result, not whether the dashboard was accurate.