Base-price models are the industry standard in STR dynamic pricing, popular because they’re simple and scalable. The math needs no probabilities or advanced statistics, and the input — widely available market data — is easy to come by. We use “base-price model” to describe any model that requires a base price set or managed by a user.
Where the data comes from
These models are fed primarily by data scraped off the OTAs — pricing and availability for the other short-term rentals in your area. A scraper is just a robot doing what you could do by hand: visit Atlanta listings on Airbnb, record price and availability for each, repeat thousands of times into a database. You can buy this data or build the scraper yourself.
From market average to a percentage curve
The scraped prices for your city are aggregated: for each future date, hundreds or thousands of observations condense into one market-average price. That curve has peaks and troughs — weekends, summer, holidays, events — created by your competitors’ variable pricing. The model then converts each day’s market average into a percentage above or below the average of the whole dataset (say, $213), producing a curve of percentages instead of dollars.
From percentage to your price
The base price is the starting point — the average rate you expect the property to command. Say I have a 3-bedroom in Atlanta I think will average $300; I set that as my base price, and the model multiplies it by each day’s market factor from the percentage curve. The result tracks the market trend exactly, just from a different starting point.
What makes it “dynamic”?
The model becomes dynamic through the frequency it’s run — but it only changes pricing when its inputs change. When the scraper runs again and picks up the price moves your neighbors have made, your pricing updates. Day-over-day changes are usually small: hundreds of scraped properties are hard to move, and managers typically make small adjustments — big swings usually signal an error.
The base-price model by itself is not dynamic; it becomes dynamic by the frequency of running it. But the model won’t change pricing unless the inputs change.
Benefits
- Simplicity — no historical data required, so a brand-new property with no booking history can run on day one.
- Accessible data — you can buy scraped market data or build a scraper for almost any market in the world.
- Easy to understand — more advanced models need dedicated revenue professionals with statistics and economics backgrounds.
The cost
The biggest criticism is that it doesn’t truly “optimize” revenue. Optimization is a mathematical process that uses probabilities to find the revenue-maximizing price; a base-price model just takes a static price and decides how much to nudge it up or down from market trends. Think of it as a “follow your neighbor” model — if your neighbors price well, it performs well; if you don’t trust their pricing, invest in an optimization model.
Why base-price models need so many comps
A base-price model avoids the forecast and optimization steps; it does not need to know the price, only how far to move a user-set base up or down from competitors. To do that it averages a broad market price, and a big dataset, hundreds of listings, smooths day-to-day noise into a forward curve, which is exactly what the model needs because your price is just a function of that average. Shrink the comp set and the model picks up local trends but becomes too responsive and erratic. To stay consistent it needs a lot of comps.
Why the whole industry ships the same model
Building a real optimization model is genuinely hard, so if customers accept a base-price model, a vendor saves enormous R&D. And base-price models scale almost anywhere because they need no historical data, no revenue processing and no currency conversion. The technology is easy to replicate, which is why many vendors and PMSs ship their own version of the same model, and why innovation stalls inside it.
What changes when there is no base price
An optimization model controls price differently, so it solves the comp-set problem too. It does not need a market average; it needs the small set of alternatives a shopper actually weighs. About 30 comps are attached to each listing, and for each night the 10 best matches that are still available go into the model. Shoppers screen down to a consideration set of about 10; they do not have the time or attention to weigh hundreds.
Go deeper: How optimization models work, Introducing search-based comp sets, Dynamic pricing isn’t optimized pricing.