Draft measurement system
The measurement system I would stand up in the first 60 days (funnel architecture, KPI trees by function, unit economics by path, and a driver-based three-year forecast), built so that swapping placeholders for facts is a data exercise, not a rebuild.
This is the measurement system I’d stand up in the first 60 days, not a finished plan. Every number below is illustrative and labeled: the structure is the deliverable; the inputs get replaced with real data on day one. I’ve built it so the assumptions live in one place and everything downstream recalculates from them. Companion to the June 1 commercial-motion memo.
How to read this document
Treat the figures as placeholders, not claims. The funnel definitions, the KPI trees, the by-path columns, and the driver-based forecast are the work; the illustrative inputs simply demonstrate that the machine computes. On day one the blue input cells get real CAC, real cost-to-serve, and the real self-serve/assisted split, and the whole model recalculates.
Pairs with the commercial case (the memo argues where the leverage is; this model measures whether we capture it) and the secret-shop teardown that surfaced the assisted-motion question this model instruments.
Five rules the model is built on, each of them a reaction to something specific about Ambiki.
The website implies self-serve (30-day trial, “set up immediately”); the win stories are assisted (billing team phones payers, weekly account touch). Until we know the real split, every funnel and unit-economic line is reported by path: self-serve vs. assisted. Blending them hides the only number that decides how many reps and CSMs we hire. This is the open question from §2 of the June 1 memo, made into a permanent column rather than a one-time study.
ARR is a scoreboard, not a steering wheel. The model elevates the predictive metric per stage: trial→activation, not just trial→paid; activation→renewal, not just renewal rate after the fact.
A PMaaS, retention-led business at this stage lives or dies on net revenue retention. The forecast is driven as much by the expansion/churn engine as by new-logo acquisition.
Marketing and Sales report off the same stage definitions and the same source-of-truth. A “lead” means the same thing in both dashboards or the whole model is fiction.
The forecast is built from drivers (trials × conversion × ACV, capacity × ramp × quota). The TAM check (~$70B SLP/OT/PT services market, pediatric the fastest segment) only confirms the bottoms-up isn’t capped by market size; it never drives the number.
One shared funnel, two paths through it. Stage definitions are deliberately observable: a system event, not a human’s opinion.
| Stage | Definition (the observable event) | Self-serve path | Assisted path |
|---|---|---|---|
| Visitor → Trial start | Account created, trial clock starts | Primary | Primary |
| Trial → Activated | Hits the activation milestone (§4) inside the window | Self-driven | CS-assisted |
| Activated → Paid | First successful charge | Mostly automatic | Rep-closed |
| Paid → Onboarded | Defined onboarding milestone complete | Light-touch | High-touch |
| Onboarded → Expanded | Seat add, Tenalog attach, or module attach | Rare | Land-and-expand |
| Any → Churned | Cancellation or downgrade | Higher | Lower |
The bet I’d want to confirm fast (June 1 memo §2): most paid conversions are assisted even when they begin in the trial, i.e., the trial is a lead source, not a true self-serve motion. If true, marketing spend, rep count, and CS staffing all change shape. The funnel above lets us read that off the data within a quarter instead of arguing about it.
Each function gets one outcomemetric (what it’s accountable for) and a short list of leading metrics (what predicts the outcome early enough to act). Resisting the urge to track everything is half the point.
| Metric | Why it’s here | Illustrative target |
|---|---|---|
| Trial starts / mo (by path) | Top-of-funnel volume, split by motion | – |
| Trial-source mix | Organic / ASHA partner / referral / paid: diversification away from a thin footprint | ≥50% non-paid |
| Trial → SQL rate (assisted) | Lead quality, not just quantity | 30–40% |
| Marketing-sourced pipeline % | How much of sales’ pipeline marketing actually creates | ≥50% |
| Pipeline coverage | Open pipeline ÷ quarter quota | ≥3.0× |
| Blended CAC (by path) | The efficiency ceiling on growth | see §5 |
Ambiki-specific watch item:the ASHA Corporate Partnership (Nov 2025) is a credibility asset that doesn’t yet look metabolized into demand. I’d instrument it as its own source so we can see whether it produces trials, not just logos on a page.
| Metric | Why it’s here | Illustrative target |
|---|---|---|
| SQL → Win rate (by segment) | Core conversion efficiency | Core 25–30% |
| Sales cycle length (by path) | Cash-flow and capacity planning | Assisted 45–60 days |
| New-logo ACV (by segment) | The denominator under everything | Core ~$20k / SS ~$2.2k |
| Win rate by competitor | Where we beat Fusion/Ensora vs. SimplePractice | – |
| Quota attainment / ramp | Capacity model reality-check | 70%+ ramped reps |
| Slippage rate | Forecast integrity | <20% |
| Metric | Why it’s here | Illustrative target |
|---|---|---|
| Activation rate | The leading indicator of retention (§4) | ≥70% in-window |
| Time-to-first-value (TTFV) | Days to activation milestone | ≤30 days |
| NRR (by activation cohort) | Proves PMaaS is a managed number, not anecdote | Core ≥110% |
| Gross revenue retention (GRR) | Floor under the business | Core ≥92% |
| Expansion ARR (seats / Tenalog / modules) | The land-and-expand engine | – |
| Cost-to-serve (support hrs / account) | PMaaS margin discipline | trend ↓ |
| Net health score | Early churn warning | – |
The CS table is where this role’s whole thesis gets proven or disproven: the hands-on support that wins the reviews should show up as a measurable NRR gap between activated and stalled cohorts. If it doesn’t, PMaaS is masking an onboarding problem rather than being a moat, and I’d rather learn that from a dashboard than a lost renewal year.
The leading indicator everything hangs on: the operating change from §3 of the memo, specified.
Activation milestone: instrument it, don’t change it for two quarters
Definition: pick one and hold it fixed:
My default is Afor the insurance-billing core ICP, because it’s the exact anxiety that drives switching. B may be the better milestone for smaller/self-serve. We test both and let the NRR gap decide.
30–45 days from paid.
1. Share of new accounts hitting the milestone in-window (rising = onboarding is improving as a system). 2. NRR(activated) − NRR(stalled): a real, persistent gap = activation is the right thing to manage.
How I’d know I was wrong:if activated and stalled cohorts retain at the same rate, the milestone is wrong: activation isn’t what drives retention here, learned for the price of a dashboard.
The whole point. Illustrative figures; the structure is what matters, and these get replaced with real CAC and cost-to-serve in week one.
| Metric | Formula | Self-serve (illustrative) | Assisted / Core (illustrative) |
|---|---|---|---|
| ACV | – | $2,200 | $20,000 |
| Gross margin | (rev − cost-to-serve) ÷ rev | ~85% | ~70%* |
| Gross logo churn | annual | 25% | 8% |
| NRR | (start + exp − churn − contr) ÷ start | ~80% | ~110% |
| CAC | fully-loaded S&M ÷ new logos | ~$1,000 | ~$9,000 |
| CAC payback | CAC ÷ (ACV × GM ÷ 12) | ~6.4 mo | ~7.7 mo |
| LTV | ACV × GM ÷ churn | ~$7,500 | ~$175,000 |
| LTV:CAC | – | ~7.5× | ~19× |
*Core margin carries the PMaaS service cost-to-serve; that’s the line I’d most want to watch as we scale, because “software + service” can quietly become a margin drag if cost-to-serve doesn’t trend down per account (the CS table tracks exactly this).
Read of the illustrative numbers:the core ICP is where the model wants to live: higher LTV:CAC, NRR above 100%, retention that compounds. Self-serve is cheaper to land but leaks (NRR <100%). This is the quantified version of the ICP argument in memo §2: win the 10–30-therapist insurance-billing practice even if smaller accounts are easier to close. If the real numbers invert this, the ICP thesis is wrong and I want to know in Q1.
Driver-based and bottoms-up. Below: the assumption block, then the annual roll-forward, then the Year-1 quarterly ramp. Change an assumption and everything recalculates; in the spreadsheet version, these are the only blue input cells.
| Driver | Core | Self-serve |
|---|---|---|
| Starting accounts (Y0) | 45 | 135 |
| Starting ARR | $0.90M | $0.30M |
| New logos added: Y1 / Y2 / Y3 | 55 / 95 / 150 | 300 / 450 / 650 |
| ACV | $20,000 | $2,200 |
| Gross logo churn (annual) | 8% | 25% |
| Gross expansion (annual) | +18% | +5% |
| ⇒ Implied NRR | ~110% | ~80% |
| ($M) | Y0 (base) | Y1 | Y2 | Y3 |
|---|---|---|---|---|
| Beginning ARR | 1.20 | 1.20 | 2.90 | 5.60 |
| + New-logo ARR | – | 1.76 | 2.89 | 4.43 |
| + Expansion ARR | – | 0.18 | 0.34 | 0.62 |
| − Churned ARR | – | (0.24) | (0.53) | (0.95) |
| Blended NRR | – | ~96% | ~100% | ~105% |
| YoY growth | – | 142% | 93% | 88% |
| New logos (core / SS) | – | 55 / 300 | 95 / 450 | 150 / 650 |
| Ending ARR | 1.20 | 2.90 | 5.60 | 10.50 |
That’s the sub-$1M-trajectory → $10M+ in three yearsthe role is hiring for. Note the blended NRR climbing toward 105%: that’s the model mix-shifting toward the core ICP, which is the strategy expressed as a number. If we grow on self-serve instead, ending ARR is similar but NRR stays under 100% and the business is far less valuable. The forecast and the ICP thesis are the same decision.
| ($M) | Q1 | Q2 | Q3 | Q4 |
|---|---|---|---|---|
| New-logo ARR | 0.28 | 0.40 | 0.50 | 0.58 |
| Expansion | 0.03 | 0.04 | 0.05 | 0.06 |
| Churn | (0.05) | (0.06) | (0.06) | (0.07) |
| Net new ARR | 0.26 | 0.38 | 0.49 | 0.57 |
| Reps (ramped-equiv) | 1.5 | 2.0 | 3.0 | 3.5 |
| New-logo ARR / ramped rep | 0.19 | 0.20 | 0.17 | 0.17 |
| Ending ARR | 1.46 | 1.84 | 2.33 | 2.90 |
The per-rep line is the capacity reality-check: if a ramped rep can’t carry ~$0.7–0.8M of new-logo ARR/year on a $20k ACV with a 45–60 day cycle, the hiring plan is wrong and the forecast is fantasy. This is where the model catches an over-optimistic plan before the budget gets committed.
The forecast sets the targets; these ratios keep the spend that produces them honest. Stage-appropriate, not enterprise-grade.
| Guardrail | Formula | Stage-appropriate range |
|---|---|---|
| Magic number | Net-new ARR ÷ prior-Q S&M | 0.6–1.0 building, >1.0 efficient |
| S&M as % of new ARR | – | trend ↓ as motion matures |
| Burn multiple | Net burn ÷ net-new ARR | <2.0 |
| CAC payback (blended) | §5 | <12 mo |
| Quota capacity coverage | Σ rep quotas ÷ plan | ≥1.2× |
Budgeting flows fromthe forecast: headcount is derived from the capacity model (§6c), not negotiated top-down. That’s the discipline: every dollar of S&M traces to a driver in the assumption block.
How the model gets used, not just built. A dashboard nobody reviews is decoration. The rhythm:
And what changes when I have it, honest about the unknowns, same as the memo.
Rewrites §2 and §5; it’s the single highest-leverage input and I’d nail it first.
Turns §5 from illustrative to real; the PMaaS margin line especially.
Tells me whether §4 is a net-new instrument or a formalization of something already there.
If core ACV is meaningfully higher (Tenalog + services attach), the whole forecast gets easier and the ICP case gets stronger.
Until then this is the scaffold I’d walk in with on day one and pressure-test against reality by day thirty, built so that swapping placeholders for facts is a data exercise, not a rebuild.
Pairs with the June 1 commercial-motion memo. The memo argues where the commercial leverage is; this model is how I’d measure whether we’re capturing it.