Skip to content

Instantly share code, notes, and snippets.

@gsingal
Created April 12, 2026 20:59
Show Gist options
  • Select an option

  • Save gsingal/6dcad3eadc70c77fef16c9a73e190fb8 to your computer and use it in GitHub Desktop.

Select an option

Save gsingal/6dcad3eadc70c77fef16c9a73e190fb8 to your computer and use it in GitHub Desktop.
Inquiry → Retention Deep Dive: window comparison, timing, segmentation (April 2026)

Inquiry → Retention Deep Dive

First-time paid subscribers only (n=2,099). Endpoint: subscription still active. Since July 2025.


1. Window Comparison: First 7d vs 14d vs 21d vs 28d

Question: Which window best separates retainers from churners?

Window 0 inquiries → retention ≥1 inquiry → retention Lift (0 → ≥1) Lift (0 → 2-3)
First 7d 27.2% (n=1,629) 38.2% (n=470) +11.0 pts +10.2 pts
First 14d 25.8% (n=1,421) 38.0% (n=678) +12.2 pts +13.5 pts
First 21d 24.3% (n=1,282) 37.9% (n=817) +13.6 pts +17.0 pts
First 28d 22.1% (n=1,158) 39.0% (n=941) +16.9 pts +16.4 pts

Finding: The lift INCREASES with longer windows. 28d has the strongest separation (22.1% vs 39.0% = +16.9 pts). But 7d already captures most of the signal (+11.0 pts). The diminishing returns inflection is between 14d and 21d.

Recommendation: Use 14-day window as the standard metric. It captures 72% of the 28-day lift while being actionable within a billing cycle for monthly subs.

2. Granular Inquiry Count (Entire Duration)

Inquiries (total) n Retention Interpretation
0 707 8.5% The dead zone — 91.5% churn
1 240 27.9% The step function — 3.3x improvement from ONE inquiry
2 145 27.6% Flat — second inquiry doesn't help
3 131 37.4% Starts climbing again
4 116 35.3%
5 85 40.0%
6-10 263 44.9% Steady improvement
11-20 241 52.7% Inflection — majority retention
21+ 171 51.5% Plateau — more inquiries don't help past ~11

Finding: Two step functions:

  1. 0 → 1 inquiry: 8.5% → 27.9% (3.3x lift). Getting any inquiry is transformative.
  2. 1-5 → 6-10 inquiries: ~30% → 45% (1.5x lift). Sustained inquiry flow matters.
  3. Past 11: plateau at ~52%. More inquiries don't help — other factors dominate.

The critical threshold is 1. But for sustained retention, the target should be 6+ inquiries over the listing's life.

3. When in the Billing Cycle Do Inquiries Matter? (Monthly Subs Only)

Period No inquiry → retention Any inquiry → retention Lift
First 7 days 10.3% (n=918) 19.3% (n=296) +9.0 pts
Days 7-14 9.9% (n=937) 21.3% (n=277) +11.4 pts
Days 14-21 10.2% (n=977) 21.9% (n=237) +11.7 pts
Last 7 days 9.3% (n=962) 25.0% (n=252) +15.7 pts

Finding: Recency wins. An inquiry in the last 7 days before renewal has the strongest retention signal (+15.7 pts vs +9.0 pts for first 7 days). This makes intuitive sense: a lister who got an inquiry yesterday feels value; one who got an inquiry 25 days ago has forgotten.

Implication for the metric: "Listing liquidity in the last 14 days" is better than "listing liquidity in the first 14 days" for predicting renewal. The rolling-window metric we proposed (% with inquiry in last 14d) is correct.

Implication for the product: If we could nudge demand toward listers approaching their renewal date, the retention impact would be maximal.

4. By Tier (Standard vs Premium) — 14-Day Window

Tier + Cycle No inquiry Any inquiry Lift n
Standard Monthly 5.8% 17.9% +12.1 pts 976
Standard Quarterly 31.0% 41.4% +10.4 pts 241
Standard Annual 50.8% 72.5% +21.7 pts 507
Premium Monthly 24.3% 29.1% +4.8 pts 197
Premium Quarterly 57.9% 75.0% +17.1 pts 31
Premium Annual 53.5% 70.5% +17.0 pts 147

Finding:

  • Standard Monthly has the lowest base retention (5.8%) and a strong lift (+12.1 pts). This is the segment where inquiries matter most — and where most churn happens.
  • Standard Annual has the highest lift (+21.7 pts). Annual listers who get inquiries retain at 72.5% — they're committed AND getting value.
  • Premium Monthly has the weakest lift (+4.8 pts). Premium listers may be less sensitive to inquiries — they're paying for features (boost, verification), not just demand. Or: the sample is small (n=197).
  • The inquiry effect is universal — every segment retains better with inquiries. But the magnitude varies 5x (from +4.8 pts to +21.7 pts).

5. Summary of Findings

Question Answer
Best window? 14 days (captures 72% of 28d lift, actionable within a billing cycle)
First N days or last N days? Last N days (rolling window). Recency matters more: +15.7 pt lift in last 7d vs +9.0 in first 7d
Threshold? ≥1 inquiry is the critical step function (8.5% → 27.9%). Secondary threshold at ≥6 total (30% → 45%)
Endpoint? Still active subscription (survived past first billing cycle)
First-time only? Yes — apples to apples. Repeat subscribers have different retention patterns
Segment differences? Standard Monthly: biggest population, strongest need (+12.1 pts). Standard Annual: highest absolute lift (+21.7 pts). Premium Monthly: weakest effect (+4.8 pts)

6. What This Changes in the Dashboard

The original metric was "% of paid listings with ≥1 inquiry in 14 days" measured at 55.7%.

Validated configuration:

  • Window: 14 days (rolling, not from listing creation)
  • Threshold: ≥1 inquiry
  • Endpoint: Active subscription retention
  • Population: First-time paid subscribers (for validation), all paid subscribers (for the dashboard)
  • Current value: 55.7% (already measured)
  • Implication: 44.3% of paid listers got zero inquiries in the last 14 days. Those listers retain at 25.8% vs 38.0% for those with inquiries. Closing this gap is the highest-ROI intervention.

Potential new metric: "% of monthly standard listers with ≥1 inquiry in last 7 days" — this is the highest-risk segment (5.8% base retention) with the strongest inquiry signal (+12.1 pts). Could be a targeted alert: "142 monthly standard listers approaching renewal with 0 inquiries this week."


All analyses: first-time paid subscribers, July 2025 – March 2026. Endpoint: stripe_status = 'active'. n=2,099 eligible.

@majones919
Copy link
Copy Markdown

Overall- I agree with this.

Some other things to consider

  1. Host as min rental requirement (ie- 3 month min). If they get a message for 4 weeks- they see zero value and try to cash in on 14 day refund. We currently have no min stay field. Obviously if it’s in description a direct message should catch this. But MLB does not.
  2. same as above but on gender preference. If a host is only renting to women and 3 men message her in first 14 days, the message also has no value and we hear about it in support.
  3. the last issue that’s common- the ghost traveler. Messages hosts. And then fails to respond. Again- zero value for host which usually turns to churn.

All this to say messages alone aren’t the only metric- but the value of the message.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment