PriceRadar: why a 0% median price change can hide 20% of offers getting more expensive

I’m working on PriceRadar, a free price-comparison service for Ukrainian stores. One product decision has been to keep a seller’s crossed-out price, comparisons across shops, and our recorded price history separate.

Our closed September report gives a concrete reason. Among 12,455 identical retailer-offer pairs measured on both 8 and 30 September 2026, the median price change was 0%. But 20.1% rose by more than 1%, 8.3% fell by more than 1%, and 71.6% stayed within ±1%.

A flat headline does not mean every shopper saw flat prices. Nor does an average of today’s whole catalogue answer the same question: the mix of products can change as offers enter or leave.

The report shows the sample, category breakdowns, exclusions and an aggregate CSV:

https://priceradar.com.ua/research/monthly/2026-09

Limits matter: this is our tracked assortment, not all Ukrainian retail, not a sales-weighted inflation measure, and not the full calendar month. Pairs missing an endpoint are excluded; no last-price carry-forward. The public CSV is aggregated, not a raw offer-history dump.

For people building shopping or analytics interfaces: would you lead with the median, the share of offers that changed, or show both side by side? That is the presentation question this example raises for us.

Disclosure: I’m affiliated with PriceRadar. Browsing is free; some retailer links on the service are affiliate links. AI helped prepare this post from the published figures.