Package holidays customers are value-conscious; they compare prices across sites and expect to get the best deal – or at the very least a good one. A holiday that is too expensive will not be booked, and the customer might not come back to the website. A holiday that is priced too cheaply is a missed opportunity to generate revenue. Finding the right balance is incredibly important and Price Intelligence is critical to the business operations of On The Beach. Every day we process millions of prices across multiple OTAs. When a routine manual check revealed that the prices we received for a competitor didn't match what a human saw on the same site, for the same hotel, on the same day, we had to ask ourselves whether the competitive intelligence underpinning our pricing decisions could be trusted at all.

The differences were systematic, directional, and suspiciously round. This talk covers what we did next: how we designed a controlled experiment to distinguish between legitimate price volatility and deliberate adversarial manipulation. We'll walk through the stratified sampling approach, the framework for classifying products as clean or contaminated, and the moment when we discussed the results. We'll cover how industry margin constraints can be used to bound what's plausible, how route-level volatility data helps separate signal from noise, and why your effective sample size is probably smaller than you think when observations aren't independent.

Whether you're scraping competitor data, consuming third-party intelligence feeds, or just trying to make decisions from data you don't fully control - the core question is the same: how do you know your data source isn't fighting back?