Unit 5 · Maximizing Performance Through Iteration · Lesson 5.13
When your comp set outprices you
A listing priced below the 25th percentile of its comp set, in a market running healthy occupancy, with a listing that still converts browsers into bookings, should not be sitting empty. When it is, the comp set is usually the problem. This case study walks a four bedroom Gatlinburg property from an overly ambitious comp set to a realistic one, and reprices it to restart bookings.
The full lesson text below is an edited transcript of the video, published 2026-08-24. The complete course is free at the playbook.
The ambitious comp set trap
The mentality behind this failure is easy to sympathise with: my property is great, so I should only be comparing myself to great properties. It is the complete opposite of what a comp set is for. A comp set assembled from the highest nightly rates in the market, or from the listings with the highest estimated annual revenue, does not tell you what your listing is worth. It tells you what those listings are worth, and it quietly makes yours look cheap by comparison.
That is not a cosmetic error. Every pricing decision that follows is read off that comparison, so feeding it the wrong reference returns the wrong answer with total confidence. Like anything else, if you are looking at the wrong information you are going to get the wrong outcome, and the frustrating part is that the numbers keep looking reasonable the whole way down.
The listing and the symptom
The listing is a four bedroom property in Gatlinburg, Tennessee. The steps that split its booking window run 0 to 7 days out for the hedge, 8 to 51 for the control, and 52 to 365 for the test. Cleaning runs at $350, tax is 13.5%, and the booking platform takes 15.5%.
The conversion metrics were checked first, and they were fine. When people saw the property they clicked it, and when they clicked it they went on to book. The listing had not been removed from the platform, and there were no bad reviews. Nothing about the listing itself had fundamentally changed.
The symptom is stark: no bookings at all, across every step, while the listing sat priced low against its comp set. The instinct in that situation is to drop rates further. That instinct is exactly what a bad comp set is about to punish.
The first read of the data
The rates in play were $375 midweek and $550 on weekends in the hedge, $450 and $650 in the control, and $650 and $850 in the test. Against those, the bookings column was a row of zeros. Whatever those price points were doing, the people in that market were not responding to them.
In the hedge, market occupancy was running ahead of expectations, with weekends at 71.67%. That is high, and at that level you would very much expect bookings to appear if the price point were within reason. Midweek at $375 sat between the 25th and 50th percentile, which looks perfectly sensible. Weekends at $550 sat barely above the 25th percentile, which alongside 71.67% market occupancy makes no sense at all.
The control told the same story more sharply. Market occupancy was 37% midweek and 52% on weekends against expectations of 50% and 80%, the listing had no occupancy whatsoever, and the price point was below the 25th percentile. In the test, the market had not built occupancy that far out yet, but the expectation drawn from a year earlier was high both midweek and on weekends, and again the listing sat below the 25th percentile.
Below the 25th percentile, in a market that is filling, with a listing that converts, and still nothing. Unless the rates had only just been implemented and the market had not been given time to respond, the data being compared against had to be wrong.
What the contradiction means
Three readings arriving together are the signature: your occupancy is low, market occupancy is high, and your price position is low. When all three are true and the listing itself is healthy, the comp set is misleading and needs revision. If you are genuinely one of the cheapest listings in a market that is booking, you get booked.
That is the moment to stop cutting. Another rate drop off a broken reference does not fix anything. It just sells nights cheaper than they need to be sold, on a calendar that is empty for a different reason entirely.
Rebuilding the comp set
The rebuild used the same market and the same bedroom category, still four bedroom properties in Gatlinburg. What changed was the selection rule. Instead of reaching for the highest possible nightly rates or the highest possible annualised estimated revenue, the properties were chosen listing by listing on what is relevant to the guest, a group with a similar type of appeal, which is what produces an honest range of nightly rates.
The rebuilt comparison looked nothing like the first one. Its 25th percentile came in at $240 a night, with the rungs above it at $308, $415, $417 and $508. That is the circularity worth understanding: if you start with a range of nightly rates in mind, you will always end up with that range as your outcome, and the outcome will be misleading.
The same rates, repositioned
Market occupancy barely moved between the two comparisons: 29% midweek in the hedge, which is low, and 73% on weekends, which is strong. The price positioning moved a very long way.
Against the realistic comp set, the hedge midweek rate of $375 sat between the 50th and 75th percentile, closer to the 75th than the 50th. The weekend rate of $550 sat between the 75th and 90th. The same unchanged rates that looked cheap against the ambitious comp set were in fact priced near the top of the listing’s real peer group.
Nothing about the listing was broken. It was expensive, and the first comparison hid it.
Setting the new price points
Hedge midweek, with market occupancy at 29% and expected to land near 35%, moves slightly above the 25th percentile, to $250. That is a heavy drop from $375, and it is deliberate: with no bookings on the calendar whatsoever, the new plan should be very conservative, and it can be reassessed once it starts getting traction.
Hedge weekends, with occupancy around 73%, move up toward the 50th percentile, to $400. Not at the 50th and not above it, because the hedge is the window where fewer people are still looking, and the job in that window is filling last minute vacancies.
Control midweek, with expected market occupancy near 50%, lands between the 25th and 50th percentile, provisionally $325. Control weekends, expected at 81% against 60% currently, took a first pass at $550, then got held back, because December 5 to January 17 is a peak period sitting inside the control window and the blended percentile read will be pulled upward by it.
The test step carries occupancy of 52% midweek and 73% on weekends from a year earlier, with the current trajectory 1.7% down and 2.5% up, close enough to flat to ignore. Test rates are set 20% above the control, so they have to wait until the control numbers are settled.
Reading the month by month detail
Two numbers needed a closer look: the control weekend rate, and confirmation of the control midweek rate. Both come from reading December weekend by weekend and stopping before the late December peak ramp, so that the peak does not contaminate the average.
On weekends, the 50th percentile readings ran in the low $400s, 417, 400 and 429 among them, averaging $423.91 across the weekends before the ramp. The 25th percentile averaged almost exactly $325. With weekend occupancy expected to be high, the 50th percentile is the right reference, and the rate is set slightly above that average, at $450.
Midweek, over the same pre-peak stretch, the 25th percentile averaged $250 and the 50th averaged $320. The band is therefore $250 to $320, and the provisional $325 sits just outside the top of it. Control midweek is revised down to $295.
With the control settled, the test follows: $295 times 1.2 is $354, rounded up to $355, and $450 times 1.2 is $540.
What the next iteration should show
The new iteration opens on 26 October with the step boundaries unchanged at 7 and 51 days. The plan is $250 and $400 in the hedge, $295 and $450 in the control, $355 and $540 in the test, against the old $375 and $550, $450 and $650, $650 and $850. That is between roughly a quarter and nearly half below where the listing was.
The purpose of the drop is not the drop. It is to restart bookings so that the next decision can be made from real data instead of an empty calendar, and so that dates stop rolling past unsold when they could have been earning. It is not the long term strategy.
The expectation for the next read is a decent number of bookings in the hedge and the control, and a couple in the test. Then the question becomes whether the pacing is working, whether there is a risk of overbooking if the strategy continues, or whether the calendar is still underbooked. Still no bookings at these numbers would be very concerning.
The discipline in the meantime is to not shut it off prematurely. Let a few bookings come through, reassess in the next iteration, and most likely start bringing prices back up, the test step included.