Written by: Michał Sobieraj, Operations Manager, Digital Colliers
The Bloomberg story about airlines using AI to move seat prices in near real time is worth reading twice. Not because dynamic pricing is new in aviation. It isn't. What's new is the speed, and the fact that the arbitrage windows travellers used to exploit are collapsing. If you run pricing at a retailer, the interesting question isn't whether airlines are ahead. They obviously are. The interesting question is why, and what it would take for a mid-sized eCommerce operator to catch up.
The short answer: airlines don't have a better pricing algorithm than you. They have a better data model.
Airlines have one source of truth, you have seven
When a revenue management system at an airline decides what to charge for seat 14C on Tuesday's 07:20 to Madrid, it queries a single, authoritative view. Live inventory. Live demand signal. Live competitor fare from the GDS. Historical booking curves for that route, that day of week, that time of year. All of it sits in one place, updated continuously, with the pricing engine as a first-class consumer.
Now think about your stack. Cost sits in the ERP, sometimes stale by 24 hours. Inventory lives in the WMS, or across three WMS instances if you're multi-region. Competitor prices come from a scraping tool that runs overnight and dumps a CSV somewhere. Elasticity, if you have it at all, is a spreadsheet a merchandiser updates quarterly. The pricing engine, whatever tool you bought, has to reconcile all of this before it can make a decision. Most of the time it can't, so it falls back to rules a human wrote in 2022.
This is why weekly repricing is still the norm. It isn't laziness. It's that the data underneath can't support anything faster without breaking.
What repricing-ready actually means
Before you buy another pricing tool, audit whether your data can feed one. A repricing-ready model has four pieces, all live, all queryable from one place:
- Cost per unit, live from the ERP. Including landed cost, duty, and current FX. Not last month's average.
- Inventory per SKU per location, live from the WMS. With a sell-through velocity attached so the engine knows what's actually moving.
- Competitor price per matched SKU, refreshed hourly at minimum. Matched is the hard word here. Fuzzy matching across retailers is where most scraping projects quietly die.
- Elasticity per SKU, or at least per category. Derived from your own price test history, not a vendor's generic curve.
If any of these four is missing or stale, no pricing engine on the market will save you. You'll just be automating bad decisions faster.
Why this matters more in 2025 than it did in 2022
The margin environment has changed. UK eCommerce grew around 3% in 2024 versus 2023, and single-digit growth is now the baseline rather than a bad year. Customer acquisition cost across DTC brands is up roughly 40% since 2023, and Meta CPMs kept climbing through 2024 and 2025. Return rates on apparel sit at 19 to 20% of gross online sales, and higher in categories like fashion. Profitero's analysis suggests around 30% of SKUs at a typical multi-channel brand lose money per order once you back out returns and ad spend.
In that environment, the price you show at 10am on Tuesday matters. If your competitor drops 8% at noon and you notice on Friday, you've given away four days of margin or four days of volume, depending which way the wind was blowing. Multiply that across a catalogue.
The left-behind risk is quieter than you think
Roughly 95% of enterprise AI projects fail to reach production or ROI, and pricing projects fail for the same reason most of them do. Teams buy the model before they fix the plumbing. The vendor demo looked great on clean data. Your data isn't clean.
Operators who'll ship real dynamic pricing in 2026 are the ones doing the unglamorous work now. They're consolidating cost, inventory, and competitor feeds into one warehouse. They're building the SKU matching layer. They're running small elasticity tests per category so they have real curves, not vendor defaults. When they finally plug in a pricing engine, it works, because everything it needs is already there and already true.
The airlines figured this out twenty years ago. The retailers who figure it out this year will have a two-year head start on the ones who wait for a better tool to arrive.