RecommendationsFrom finding
From finding
to next Monday.
Every validated insight comes paired with a prioritized recommendation — target segment, expected impact, effort level, and numbered action steps. Something a team could pick up without a meeting.
Sorted, not scrambled
Prioritized the moment they're produced
HighDelist 972 properties from inactive hosts and auto-cancel their pending bookings
HighShow the full price breakdown before booking on overpriced listings
HighSuppress search ranking for 412 low-quality listings and require an improvement plan
MediumStricter cancellation policy for same-day bookings
MediumActivation campaign for 13,649 dormant hosts in India and China
MediumDedicated account program for the top 169 mega-hosts
LowOff-season campaign for 7,068 summer-only properties
Example recommendations from a vacation rental run.
Pulled out of the top lane
What a recommendation actually contains
Show the full price breakdown before booking
One of three high-priority recommendations.
High priority
Action steps
- Identify listings where the total booking amount exceeds 1.5× the expected base × nights.
- Surface the full breakdown — cleaning, service, taxes — directly in the listing card before checkout.
- Add a "Total shown includes all fees" badge to every affected listing.
- Run a two-week holdout A/B test on 20% of traffic in the affected segment.
- Roll out fully if the cancellation rate drops ≥3pp with p < 0.05.
Findings and actions stay connected
Every recommendation points back to the insight that justified it. If an insight gets invalidated by new data, its recommendation moves with it — no orphaned action items, no stale priorities.
app.decisionbox.io/recommendations
Keep reading
Try it in two minutes
Clone the repo, run docker compose up, and point it at your warehouse.