Thinklytics

Southwest Hotel Group · Gaming & Hospitality · Las Vegas, NV · 14 weeks

Dynamic pricing AI deployed across 12 properties

Southwest Hotel Group relied on manual rate setting, using weekly competitor reports and past occupancy data. We developed and implemented a dynamic pricing AI for 12 properties that adjusted rates in real time. This system increased revenue by $3.8 million in the first year.

Challenge

The revenue management team relied on weekly competitor rate reports and 90-day occupancy data to set prices manually. This approach missed same-day demand shifts from local events, weather changes, and competitor moves. As a result, they were losing an estimated $3M to $5M in annual revenue from poor pricing decisions.

Approach

We created a pricing model that pulled real-time data from competitor rates, local events, weather, and booking speed. It updated rate recommendations every four hours for each property and room type. We also developed a dashboard displaying these recommendations with clear explanations, enabling the team to accept or adjust prices instantly.

Outcome

We rolled out dynamic pricing across all 12 properties by week 14. Over the first year, this approach added $3.8 million in revenue compared to the previous manual pricing method. Meanwhile, the revenue management team’s override rate fell from 40 of every 100 in the first month to 12 of every 100 by month six, reflecting growing confidence in the model’s recommendations.

We spotted real-time signals that manual tracking completely missed.

Alright, here’s what was going down. We used to rely on these slow weekly competitor reports and old data to set prices. That meant we were always a step behind. Like when a Taylor Swift concert got announced just three days out, or when a competitor dropped their prices by $18 out of nowhere in the afternoon, we had zero clue until it was too late. So, we built an AI that watches all this stuff in real time and adjusts prices instantly. No waiting around. Just quick moves that keep us ahead.

How We Got More People Using the Tool by Making Insights Super Clear

The revenue team wasn’t sold on this black-box pricing model at first. So, we built a simple dashboard that showed the recommended price plus the top three reasons why. Once they could actually see what was behind the numbers, trust started to build. The payoff? Price overrides fell from 40 of every 100 down to 12 of every 100 in just six months.

How We Spotted Model Slip-Ups and Boosted Accuracy by Adding Overrides

Every time we overrode the model, we retrained it right away. If that override improved things, the model gave more weight to that choice. If it made things worse, the model pulled back fast. Over a year, this back-and-forth pushed accuracy up by six more correct picks per 100 cases.

Results

  • $3.8M Incremental revenue in first year
  • 12 Properties on dynamic pricing
  • 4 hrs Rate optimization cycle
  • 12 of every 100 pricing decisions Override rate at month 6

We knew manual pricing was holding us back, but we didn’t expect it to cost $3.8 million in the first year. Now, the model pays for itself every month.

VP of Revenue Management, Southwest Hotel Group

In the client's terms

The question

Which of this work should a person still be doing?

The situation

Revenue management set prices by hand from weekly competitor reports and 90-day occupancy history across twelve properties. Same-day demand from local events, weather and competitor movement was invisible, at an estimated $3M to $5M a year in foregone revenue.

What we found

The team was not slow, the inputs were. A weekly report cannot price a Tuesday that changed on Tuesday, and the decision cadence was set by the reporting cadence rather than by the market.

What it produced

Live across all twelve properties by week 14 and $3.8M in incremental revenue against the manual baseline over the first year. Override rate fell from 40 in 100 in month one to 12 in 100 by month six, which is the number that shows whether people trust it.

The roadmap

A pricing model fed by live competitor rates, local events, weather and booking pace, producing recommendations every four hours per property and room type. Recommendations were surfaced with the reason attached rather than as a number, so the team could accept or override with something to argue against.

The work behind this result

Related case studies

Frequently asked questions

Does dynamic pricing work for hotel groups?

At Southwest Hotel Group it produced $3.8M in incremental revenue across twelve properties in the first year, measured against the manual pricing baseline. The gain comes from acting on same-day signals such as local events, weather and competitor movement that a weekly report cannot carry.

How do you get revenue managers to trust a pricing model?

Show the reason, not just the price. Each recommendation here carries the top three factors behind it, so the team accepts or rejects an argument. Override rate fell from 40 in 100 in month one to 12 in 100 by month six.

How often should hotel rates be re-optimised?

Every four hours per property and room type at this group, because demand moves inside the day. The previous cadence was weekly competitor reports and 90-day occupancy history, which missed an estimated $3M to $5M a year.