What we learn from watching a few million prices change. Written for the people who have to decide what to charge.
Everyone buys it expecting to discover where they are too expensive. That is almost never where the money is.
The unglamorous question that decides whether a price comparison is worth anything, and why the answer changes completely per industry.
Regulation fixes or caps one slice of the range, usually reimbursed prescription medicines, leaving over the counter, self-care, cosmetics and delivery pricing entirely to the pharmacy.
Less about the token bill than most people assume, and far more about what happens when you leave.
Running a price tracker in-house costs mostly people, not servers: expect parser upkeep and match review to dwarf the fetching bill, roughly three quarters of the total.
Report gross profit in currency, price realisation, win rate at quoted price and margin by group; treat price index, change counts and average discount with suspicion.
Convert every offer to one base unit by dividing total price by pack count times content per pack, then flag the pairs that cannot convert.
Minimum advertised price controls the displayed price, not the transaction price. Enforce it with scheduled monitoring, a fixed evidence format and identical consequences, and check the rules per country.
Generic tools match on brand and title text, while wine prices are set by vintage and bottle volume, two fields that are inconsistently placed and often missing.
Ask how the vendor matches your products to competitor listings, how often that match is wrong, and how you find out; everything else is downstream.
A price index is your price divided by a reference price times 100. The hard parts are choosing the reference, weighting by revenue and matching the products correctly.
Rule-based repricing applies logic you can read back and audit; dynamic pricing lets a model set prices from demand. Most companies need rules, not models.
A confidence score estimates how likely two listings are the same product, which makes it good for routing pairs to review and poor as proof.
Decide margin or market share per SKU using four tests: comparability, repeat purchase, real volume economics and basket pull. Company-wide pricing policies apply one answer to opposite cases.
A published trade price is not the price your customer pays. Compare on the delivered, discounted, waste-adjusted figure, because that is what the buyer is choosing between.
A competitor's shelf price describes a product, not a person, so it is not personal data; the risk comes from names and reviews on the same page.
Most competitor price errors are silent successes: the page loaded, the parser ran, and the number it produced is not what a customer would actually pay.
Track competitor prices with a small matched set, polite collection from public pages, timestamped observations, and a human review step before anyone acts on a number.
Exit a price war by raising a narrow, low-visibility set of lines first, holding the new level long enough to be seen, and explaining it without mentioning competitors.
A comp set records what a few named hotels were asking on one channel at one moment, not what they sold, at what occupancy or after commission.
No EU law bans collecting a competitor's published prices; the risk sits in database rights, accepted terms of use, unauthorised access rules and how hard you fetch.
Match on GTIN, EAN or brand plus MPN where identifiers exist, and rebuild identity from brand, model, quantity and pack count where they do not.
Competitive price monitoring is the repeatable collection of competitor prices for products you have deliberately matched to your own, kept with history and used as pricing evidence.