Skip to content

Instantly share code, notes, and snippets.

@gsingal
Created April 11, 2026 15:58
Show Gist options
  • Select an option

  • Save gsingal/b8da3caee9c79aa690439f267367b00e to your computer and use it in GitHub Desktop.

Select an option

Save gsingal/b8da3caee9c79aa690439f267367b00e to your computer and use it in GitHub Desktop.
KPI Proxy Validation: What actually predicts business outcomes? (April 2026)

KPI Proxy Validation: What Actually Predicts Business Outcomes?

Before picking metrics, we ran the analyses to validate which proxies actually predict the outcomes we care about.


The Core Finding

Getting just 1 inquiry in 14 days nearly doubles lister retention (19.9% → 33.7%). Everything else we build — messaging strategy, ChurnKey, email lifecycle, pricing optimization — is secondary to this one insight. 60% of paid listers got zero inquiries this week.


7 Analyses, Summarized

1. Inquiry → Retention (the biggest signal)

Inquiries in first 14d Retention
0 19.9%
1 33.7% (+70%)
2-3 36.0%
4+ ~35%

Insight: The step function is at 1. Not 3, not 5. Get a lister one inquiry and retention nearly doubles. Beyond that, marginal returns.

2. Response Time → Thread Completion

Lister responds in... Thread completes
No reply 0%
< 1 hour 40.4%
1-4 hours 37.2%
4-24 hours 32.7%
24-48 hours 26.3%
> 48 hours 18.2%

Insight: Getting ANY reply is the biggest jump. After that, <4 hours is meaningfully better than >24 hours.

3. Reply → Searcher Re-engagement

Searcher... Messages another listing (30d)
Got ghosted 37.9%
Got a reply 59.3%

Insight: Searchers who get replies are 56% more likely to keep using the platform. Reply rate is a two-sided metric — it predicts both lister retention (via thread completion) and searcher retention.

4. Premium Boost Effect

Tier Avg inquiries/30d % with any inquiry
Free 3.17 55.1%
Standard 4.48 72.0%
Premium 5.63 70.6%

Insight: Premium gets 26% more inquiry volume but NOT more reach (70.6% vs 72.0%). The boost is per-listing, not per-probability. Important for Pro Plan — if we promise "more visibility," the data supports "more inquiries" but not "higher chance of at least 1."

5. Billing Cycle → Retention

Cycle Retention
Annual 50.7%
Quarterly 34.8%
Monthly 12.3%
Free 17.0%

Insight: Annual retention is 4.1x monthly. The annual tilt is already working (53.7% of paid subs are annual). Protect and expand this.

6. Fraud Impact

Last 30 days Value
Searchers who contacted a fraudster 37
Messages to fraudsters 47
Conversations with fraudsters 41

Insight: Low numbers — the fraud system is working. Measure in people harmed (37), not events (47).

7. Plan Distribution

Plan % of paid
Standard Annual 44.2%
Standard Monthly 15.6%
Standard Quarterly 13.9%
Premium Annual 9.5%
Free 8.4%
Premium Monthly 6.6%
Premium Quarterly 1.8%

The 8 Validated Top-Line Metrics

# Metric Current What it predicts Evidence
1 Listing liquidity (% paid w/ ≥1 inquiry in 14d) ~55% Lister retention: 19.9% → 33.7% at threshold Analysis 1
2 Reply rate (% messages getting a reply) 51.3% Thread completion: 0% → 40.4% Analysis 2
3 Median response time 34.1h Completion: 40% at <1h vs 18% at >48h Analysis 2
4 MRR (Stripe, with discounts) $36,085 Revenue health (direct)
5 Monthly churn rate (Stripe) ~25% Business sustainability (direct)
6 Annual plan % 53.7% Revenue durability: 50.7% vs 12.3% retention Analysis 5
7 Searchers who contacted fraudsters 37/mo Trust / searcher harm (direct) Analysis 6
8 Organic → signup → message funnel ~61.6% s→m SEO investment ROI Needs organic-specific data

What This Means for Decisions

Messaging Strategy: The #1 lever is getting zero-inquiry listers to ≥1 inquiry. That's 1,727 listers with 19.9% retention vs 33.7% if they get just one message. This is worth more than ChurnKey, email lifecycle, and pricing optimization combined.

Pro Plan: Premium gets 26% more inquiries per listing but the same reach percentage. If Pro promises "more visibility," we need to show it delivers more reach (% with ≥1 inquiry), not just more volume per listing.

Fraud: 37 searchers affected per month is low. The system is working. Don't over-invest here — the ROI of fraud reduction is diminishing.

SEO: We need the full funnel (impressions → visits → signups → messages) to know if content investment is paying off. Without it, 135K blog impressions could be vanity.

Churn: Decompose by inquiry status. Listers with 0 inquiries need demand (marketing/matching problem). Listers with inquiries who still churn need product improvements (pricing/UX problem). Different interventions for each.


Generated 2026-04-11. All analyses run against production read-only replica.

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