Growth & conversion analytics · Python + Excel
People are browsing more, but booking less.
I analysed 254,177 real user sessions on a hotel-search platform to find where the funnel leaks. The headline: across the two months, engagement rose while the booking rate fell — visitors clicked into more offers but completed fewer bookings. This dashboard shows where conversion holds up, where it breaks, and what to fix first.
1 May – 30 Jun 2022
Python · pandas
Excel
Funnel · clickout ratio · booking rate
The numbers at a glance
254,177
user sessions analysed
1.58
avg. clickouts per session
1.57%
booking rate (sessions that book)
4,105
total bookings in the window
Where the funnel leaks
Sessions → clickouts → bookings
Sessions generate plenty of offer clicks (402,003 clickouts in total), but only 1.57% end in a booking. The leak isn't at the top of the funnel — it's at the final step.
402,003 clickouts — ~1.58 per session, strong offer engagement
Bookings4,105 · 1.57% of sessions
The two-month trend — engagement up, bookings down
Segment explorer — who converts, who doesn't
Clickout ratio by traffic source
Pick a dimension and a metric. Bars are sorted best-to-worst; the dashed line marks the platform average. Hover any bar for session volume and both metrics.
Break down by
Measure
What the data says to do
Traffic — value ≠ volume
Source Type 2 (94,331 sessions) is the money source: best engagement (1.71) and best booking rate (2.63%). Types 8 & 10 pull real volume but almost never book (0.23% / 0.05%).
Shift spend toward Type 2; stop paying for the volume of Types 8 & 10.
Device — the biggest one leaks
Device 20 handles the most sessions (96,032) yet books at just 0.65% — the single largest conversion opportunity. Device 14 books best at 4.61%.
Audit Device 20's checkout — a small lift there moves the whole number.
Engagement isn't booking
Turkey tops engagement (COR 1.88) but books at only 0.89% — below the US, which engages least (1.49) yet books best at 2.74%. Clicking ≠ converting.
Treat engagement and conversion as two problems; localise pricing where clicks don't convert.
The trend to reverse
Weekly booking rate peaked at 1.65% in late May, then fell to 1.45% by late June — while engagement kept rising. Interest is growing; conversion is slipping.
Instrument the booking step and A/B test friction, trust signals and price display.
Why the booking step is the priority
+0.5 pt booking rate ≈ 1,270 more bookings
Only 4,105 of 254,177 sessions booked. Lifting the booking rate by half a percentage point would add roughly 1,270 bookings over this window — about a third more, from traffic the platform already has. Engagement is healthy, so the cheapest growth isn't more visitors; it's closing the gap at the booking step. That's where I'd focus and measure.