Zero October bookings on a listing priced under market. The fix wasn't the rate.
A standing AI employee ran three straight days of automated checks that came back "nothing left to cut" on price — which is exactly what redirected the search to the guest's total bill instead. Four days after the fix, the listing went from empty to beating market.
The trigger
A standing job description caught this one — Occupancy Growth Manager, running five checks each day and each week: a 7am gap scan, a Monday scoreboard, a nightly drift check, a Thursday demand watch, a Wednesday listing review.
The diagnosis
The listing sat under the market's 25th percentile — the number that usually means "drop the price." Every normal signal pointed the other way: top-5 search rank, a rate already cheap. The deep-dive found the number that mattered instead of the obvious one: fees ran 24–60% of the total guest bill, on top of zero reviews since a recent relaunch. The rate looked cheap; the checkout screen was expensive.
Before / after
- Base rate: $141, floor: $87
- Linen fee $45, resort fee 3%, extra-guest fee $45
- Zero reviews since relaunch
- 0 of 31 October nights booked
| Day | Next 30 occupancy | vs. market | Performance index |
|---|---|---|---|
| Sep 15 (before) | 0 of 31 Oct nights | under 25th percentile | ~0 |
| Sep 17 | 26.7% | 34.3% | 0.78 |
| Sep 19 | 36.7% | 35.2% | 1.04 — beating market |
What a person would have cost to catch this
(at $85–150/hr, 17–30 hrs)
(can pull the numbers, rebuilding the pricing logic is not the skill)
(likely trades the fix for less profit further out)
Estimated cost of missing it: roughly $2,300–2,700 of October revenue on one small unit. Watching a listing this closely, every week, for itself on one operator across a portfolio, is a checking discipline that doesn't happen at all without an automated employee — nobody has the hours.
The real ROI: what the AI employee actually cost
Every case-study claims a great outcome. This is the number most skip: what Viktor itself billed to produce this one. Pulled directly from the account, not modeled — $22 of it is an exact, isolated thread cost; the rest is a documented range because two pieces of the work shared a thread with unrelated work that week.
The real revenue: what guests actually paid
The $2,300–2,700 figure above was an estimate of revenue that would have gone unbooked. Here's the number underneath it, queried directly from the booking platform — the actual total guests paid for the reservations that landed after the fix.
| Booked | Stay dates | Nights | Guest paid |
|---|---|---|---|
| Sep 15 | Oct 9–11 | 2 | $504.72 |
| Sep 17 | Oct 2–4 | 2 | $475.68 |
| Sep 18 | Oct 22–25 | 3 | $660.79 |
| Sep 19 | Oct 17–19 | 2 | $539.80 |
| Total — 4 reservations | 9 | $2,180.99 | |
One property, one week, one automated employee. See what your own listings are quietly leaking.
Start your diagnostic →Results are not typical. Past performance does not guarantee future results. Property identity anonymized except where the account owner has authorized disclosure of their own portfolio.