B Babelio Playbook 14 / 24
Refresh: weekly EN · RU
Operating System · Artifact 14

KPI Dashboard

What we measure, the exact formula, the source of truth, how often it refreshes, and the red line that pages someone. 20 metrics across seven funnels. Anyone in the company can open this and answer one question: are we on track?

Metrics tracked
20
Weekly top set
5
Funnels
7
P0 tripwires
4

North Star

One number above all others. MAU is vanity; the truth is whether a learner trusted Babelio enough to leave it running over a real lecture or stream, repeatedly, every week (product.md §1).

North Star Metric
≥ 4 native-desktop sessions translated end-to-end, per active user, per week
Session = a call / lecture / stream > 5 min the user kept Babelio running on. It compounds (habit = retention) and maps to revenue (per-active-hour COGS). Red line: median < 2 sessions/wk for two weeks = the habit isn't forming.

Top 5 — read every Monday

These five are reviewed at the Monday-morning leadership sync (13-operating-cadence.html). They are the make-or-break numbers for a prototype-stage, community-led consumer product: does the habit form, does the loop exist, do they pay, do they stay. Everything else is diagnostic.

01
North Star · sessions/wk
≥4
median active user. Red <2.
02
Activation rate (D7)
45%
first dual-track + first export. Red <40%.
03
Loop factor K
≥0.3
prove the loop. Red <0.15 = single-player.
04
Net-new paid
+wk
paid adds − churned. Red: 2 wks flat/neg.
05
Weekly logo churn
≤1.7%
≈7%/mo. Red >2%/wk (~8%/mo).
Why these five At prototype stage the entire thesis rests on two unproven things: does the artifact-share loop exist (K) and does the habit form (North Star + activation). If K stays <0.15 and community CAC exceeds $37, LTV:CAC falls below 3:1 and the unit economics don't clear (growth.md §6). That is why K is a weekly number, not a quarterly one — it is the single most important reading in the first 90 days.

The full spec · 20 metrics, seven funnels

Every metric has a formula, a single source of truth, a refresh cadence, a red line, and a named owner. WK = part of the weekly top-5. Source of truth: PostHog (product analytics + cohorts), Stripe/Polar (billing), babelio-evals (quality), the in-app usage ledger (active-minutes & COGS), and the Sean-Ellis PMF survey.

Acquisition 3
Metric Formula Source of truth Cadence Red line Owner
New installs COUNT(first_launch) per week, by channel PostHog weekly < 30/wk by M3 Founder
Install-completion rate permission_granted ÷ download_started PostHog weekly < 50% = friction Founder
CAC (blended) (S&M spend + sponsor $) ÷ net-new paid Stripe/Polar + ledger monthly > $37 Founder
Activation 3
Metric Formula Source of truth Cadence Red line Owner
Activation rate WK users w/ first dual-track session AND first exported card by D7 ÷ installs PostHog weekly < 40% (≤40% even hand-held = kill) Founder
Time-to-activation median minutes: first_launch → first export PostHog weekly > 1 session of use Founder
Eval-gate pass rate latency p95 <700ms · WER <12% · MT adequacy ≥4 · hallucination <2% babelio-evals per deploy any metric regresses = no merge Eng
Retention 4
Metric Formula Source of truth Cadence Red line Owner
North Star WK median sessions (>5 min, kept on) per active user per week PostHog + ledger weekly < 2/wk for 2 wks Founder
D1 / D30 retention cohort active on day 1 / day 30 ÷ cohort size PostHog weekly D1 <30% · D30 <12% Founder
Monthly logo churn WK paid cancels in month ÷ paid at month start (watch weekly ≈1.7%) Stripe/Polar weekly > 8%/mo Founder
Sean-Ellis PMF score % "very disappointed" if Babelio went away (day-14 survey, ≥30 hands) PMF survey monthly < 40% = no PMF (do not claim) Founder
Revenue 4
Metric Formula Source of truth Cadence Red line Owner
MRR Σ active paid subscriptions × ARPU ($12) Stripe/Polar weekly growth <10% MoM ×2 mo Founder
Net-new paid WK paid adds − paid cancels, per week Stripe/Polar weekly flat/neg 2 wks Founder
Free → paid conversion paid conversions ÷ reverse-trial expiries (cohort) Stripe/Polar + PostHog monthly < 6% (base 8%) Founder
NDR (net dollar retention) (start MRR + expansion − churn − contraction) ÷ start MRR Stripe/Polar monthly < 95% Founder
Efficiency 2
Metric Formula Source of truth Cadence Red line Owner
LTV : CAC [ARPU × GM% × (1 ÷ monthly churn)] ÷ CAC — cap reported at ~3:1 Stripe/Polar + ledger monthly < 3:1 Founder
CAC payback CAC ÷ (ARPU × GM%) Stripe/Polar + ledger monthly > 18 mo Founder
AI Cost 2
Metric Formula Source of truth Cadence Red line Owner
COGS per active-hour (STT + MT + TTS spend) ÷ active-hours, split dub vs subtitle usage ledger weekly dub > $0.50 · sub > $0.31 Eng
Gross margin (blended) (revenue − COGS) ÷ revenue ledger + Stripe/Polar monthly < 50% Founder
Growth Loop 2
Metric Formula Source of truth Cadence Red line Owner
Loop factor K WK (shares per activated user) × (installs per shared artifact) PostHog weekly < 0.15 = demote loop Founder
Artifacts shared externally COUNT(artifact_shared_external) per week PostHog weekly < 1 per 3 activated users Founder
Bar for good A new hire can open this page Monday morning and self-serve "are we on track?" without asking the founder — because every metric names its formula, its source, and the exact threshold that means trouble. No metric is here without a red line; a number with no escalation rule is a vanity number.

Escalation rules · when a red line is crossed

A red line is not a chart colour — it pages a named person and triggers a specific meeting. P0 = halt-and-rethink (the thesis is at risk). P1 = scheduled review inside the week. Owners and meeting bodies are defined in 15-decisions-raci.html and 13-operating-cadence.html.

Tripwire Sev What it triggers — who is paged
Loop factor K < 0.15 (Week-4 reading) P0 Loop is dead. Demote the loop, re-run LTV:CAC on real paid/community CAC. If community CAC also > $37 → emergency strategy session (founder + board); growth becomes paid-only and 06/07/10 are rewritten before any spend.
Activation < 40% even hand-held P0 The product doesn't deliver its core value. Stop scaling, founder runs 5 onboarding watch-sessions this week, fix the first-export flow before any channel test (growth.md §7 W3 kill-criterion).
Gross margin < 50% (dub-mix shift) P0 Heavy dub usage is eating margin. Eng pages founder → tighten the dub allotment / overage, recompute pricing in 05-pricing.html. Defense already exists ($0.06/min overage, 15-min free dub cap).
Monthly churn > 8% P0 The #1 modeled risk fires. Founder runs 5 churn-interview calls this week, ships the "you understood X% on your own" reframe + meter-before-overage nudge (growth.md §9). Re-forecast 10-financial.
Install-completion < 50% P1 Gatekeeper/antivirus friction is the bottleneck. Founder + eng: ship the first-run trust playbook (product.md §5.5) — notarized build, "why these permissions" copy — before scaling spend.
MRR growth < 10% MoM ×2 mo P1 Growth stalling. Monthly business review escalates to a focused growth retro: which funnel stage broke (acquisition / activation / conversion)? Owner: founder, with growth lead once hired.
Sean-Ellis PMF < 40% P1 No PMF — do not claim it in any investor update (18-investor-update). Stay in the validation loop; segment the "very disappointed" cohort and double down on what they share in common.
Eval-gate regression P1 Any deploy that regresses latency/WER/adequacy/hallucination is blocked from merge automatically (Promptfoo CI). Eng owns; PM reviews promoted failures in Braintrust weekly.
The two readings that decide everything

This is a prototype with zero users, zero revenue, zero LOIs. Every threshold above is a hypothesis until the §7 field-work runs. Two weekly numbers carry the whole thesis: loop factor K (does the product spread, or is it single-player?) and the North Star / activation pair (does the habit form?). If both fail in the Week-4 channel test, Babelio is single-player and growth is paid-only — re-run the unit economics before any spend.

The binding constraint is not AI unit cost. COGS at $0.50/dub-hr and $0.31/subtitle-hr is already margin-viable. The two things that can kill this are consumer churn staying above 8%/mo and the unvalidated WTP-ceiling-vs-dub-COGS gap — which is exactly why churn and gross-margin sit on the weekly read, not the quarterly one.

Cross-references
  • 12-okrs.htmlThis dashboard is the union of the OKR key results — every KR appears here as a tracked metric with a source.
  • 08-retention.htmlActivation event (first dual-track + first export), aha-moment threshold, D1/D30 curve and churn signals.
  • 10-financial-model.htmlSource of the revenue, COGS, LTV:CAC and CAC-payback formulas and their tripwires.
  • 07-growth-loop.htmlThe artifact loop being measured by K and external-share count.
  • 13-operating-cadence.htmlWhere each cadence happens — Monday top-5, monthly business review (dashboard read).
  • 16-risk-register.htmlChurn, WTP-vs-COGS and loop-failure as P0 risks; the tripwires here are their early-warning metrics.
Babelio · KPI Dashboard · Playbook 14/24 20 metrics · refresh weekly (top-5) & monthly (rest)