Most unit economics spreadsheets get built the week before a fundraise, to show a comfortable LTV-to-CAC ratio and a short payback, and are never opened again. That wastes the one tool able to show which assumption is carrying the whole model and where the business breaks first.
A unit economics calculator usually gets sold as a planning tool, but its sharpest use is diagnostic. It shows where profit leaks, which assumptions are brittle, and which interventions actually move the outcome. What separates an audit from a forecast is the opening question. A forecast asks "what is our CAC?"; an audit asks "where does profitability break, and why there?"
Passing the audit means three things hold at once. Each unit is structurally profitable, or can be made so with levers you could realistically pull. Scaling adds profit faster than it adds risk: cash exposure, cost volatility, support load, refunds.
And the model survives stress: a shift in channel mix, a price change, a cost spike, a dip in retention. When one of those claims fails, tuning won't fix it. You redesign the thing that broke: the unit definition, the pricing, the packaging, the channel mix, or the cost architecture.
Inspect the denominator before anything else
Most broken models are broken at the unit itself, because the unit is fuzzy. A good unit can be counted without debate, earns money, carries its own variable costs and means the same thing in every channel and segment.
Take a B2B workflow platform that reports "active users" while its contracts are priced per workspace. Support cost scales per workspace, not per user; marketing reports CAC per signup; finance talks LTV per account. Three denominators, so the model never reconciles. The way out is one unit for everyone: the paid workspace-month. It starts with the first successful invoice. It ends with cancellation, or with non-payment after a stated grace period. CAC is then measured per paid workspace created. That one alignment reshapes decisions downstream, because cost, revenue, and acquisition finally sit under the same unit.
Audit net revenue, because gross numbers never pay bills
Revenue is usually overstated because the model reaches for list price or gross transaction value. An audit runs on net revenue: what is left after the leaks. Subtract refunds and credits, chargebacks and disputes, discounts and sales concessions, the taxes you collect and pass on (they were never yours), and partner shares or app-store fees.
An online events platform shows the trap plainly. It reports strong GMV per attendee, but it only keeps a service fee, and some event categories refund heavily. Model GMV and you will approve acquisition spend that is mathematically impossible to recover. The unit revenue has to become the service fee collected per completed attendance, net of refunds and payout adjustments, with the refund rate segmented by event category and channel source. That is the number that tells you whether the unit actually makes money.
A cost that grows with units is variable, whatever the ledger says
This is where businesses quietly fail: they treat a cost as fixed right up until scale forces it to grow, and then the model collapses. The audit rule is blunt. If a cost must increase when units increase within your planning horizon, it is variable for the purposes of the audit. The costs that hide tend to be the human and metered ones: onboarding hours per customer, customer-success touches per account, support tickets per active unit, infrastructure per usage band, and third-party per-action fees for things like verification, messaging, or enrichment.
A B2B contract-management product makes the point. It sells annual contracts, and a standard onboarding runs 8-12 hours of specialist time. At small scale, founders absorb that work and onboarding looks fixed; at larger scale it becomes a hiring driver. Booked honestly as a front-loaded variable cost per unit, it changes the payback picture (margin may only begin once onboarding is complete) and it often reveals that growth is capped by how many units the team can actually deliver, whatever demand looks like.
Walk contribution margin down five layers
Rather than jump to LTV, the audit walks contribution margin down one layer at a time:
- Net revenue per unit
- minus transaction leakage (fees, refunds, disputes)
- minus variable infrastructure (usage-tied compute/storage/bandwidth)
- minus vendor fees (per action)
- minus human-variable costs (support/onboarding scaled per unit)
= contribution margin per unit
Run it once with numbers. For an illustrative workspace-month: net revenue $100, leakage $6, variable infrastructure $9, vendor fees $12, human-variable cost $18. Contribution margin = 100 less 45 = $55, or 55%. With CAC at $660, payback = CAC / monthly contribution margin = 660 / 55 = 12 months, exactly on a 12-month boundary. Raise CAC 20% to $792 and payback stretches to 14.4 months; add 25% to vendor fees as well and margin drops to $52, pushing payback past 15 months. Two ordinary shocks, and the unit that passed now fails.
The walk matters because it isolates which layer dominates. A messaging-heavy support platform is the classic case.
It integrates SMS and WhatsApp notifications, customers configure alerts freely, and usage varies wildly: some customers trigger 10× the message volume with no matching revenue. Walk the margin and vendor messaging fees turn out to drive most of the variance, while a blended "gross margin" quietly averages away the negative-margin cohorts.
The response follows the driver: introduce usage bands or credits, add in-product controls and defaults that curb message spam, and align pricing with message volume. When you cannot explain margin variance with a small handful of drivers, the model is missing a cost attachment.
Retention is a survival curve, and payback has a schedule
Retention has to be read as cohort survival over time; a single churn figure flattens the lifecycle. A project-based SaaS shows why the average lies. Customers churn once a project wraps, but new signups keep replacing them, so monthly churn reads acceptable, even as LTV stays capped because the usage is episodic by nature.
The useful cohort cut is by project type and frequency, and the real question is what share of customers start a second project within X days. If the product cannot make multi-project behavior natural, acquisition has to be constrained, because both payback and the LTV ceiling are structurally limited, and the roadmap's job becomes making repeat use easy through templates, duplication, and cross-project dashboards. That is a different conclusion from "improve churn": here churn is built into how the product gets used.
Timing matters as much as survival. Two businesses can carry the same margin-based LTV and live in completely different cash realities, so the audit wants timing curves. When is CAC paid: immediately, weekly, across a sales cycle? When does margin begin: after onboarding, after delivery, after first renewal? And how does it accumulate: monthly, usage-based, seasonal?
An implementation-heavy analytics tool illustrates the gap. It sells annual contracts collected upfront but needs integrations, so cost is front-loaded and renewals hinge on adoption. Early profit looks high on the strength of that upfront cash, renewal risk climbs when adoption is slow, and an honest payback view adds a time-to-adoption leading indicator. The moves that follow (pricing implementation explicitly so cost becomes revenue, steering sales toward profiles that integrate quickly, and requiring minimum contract terms) all exist to stop upfront cash from being mistaken for durable unit economics.
Blended numbers are not evidence
An audit fails the moment it leans on blended metrics while the segments underneath behave differently. The cuts that usually matter are acquisition channel, for intent and cohort quality; plan or tier, for support and feature mix; customer size and complexity in B2B; usage intensity for usage-based products; and geography, which drags in fees, compliance, tax, and dispute rates.
A consumer subscription that grows through both search and influencer campaigns makes the failure concrete. Search brings high intent and high retention; influencers bring volume and early churn. Blended CAC and blended churn look fine, but split them and the influencer cohorts are underwater against the payback boundary. The response is cohort-specific: tighten the refund-prone influencer offers, rework onboarding to lift early retention for that group, and cap spend until the cohort clears its margin floor and payback threshold. A model that cannot produce segment-level pass/fail verdicts can tell a story, but it cannot audit anything.
Does the model survive a realistic shock?
You do not need dozens of scenarios, only the handful that actually kill businesses. A workable set has five shocks. CAC up 20%, as ad auctions and competitors push prices. Refunds or disputes up 10%, after a policy or channel change. Variable infrastructure up 25%, after a workload or vendor change. Support tickets per unit up 15%, as the product gets more complex. And an early dip in retention, as if a product change made users wait longer for value.
An expense-management SaaS shows what the battery catches. A new integration lifts adoption but also support tickets; without the shock, the team simply celebrates a successful feature. Run it, and support-driven margin erosion pushes payback past the boundary at the planned growth rate. The response is to ship self-serve diagnostics and integration health checks before scaling, put the integration behind tiers that fund its support load, and slow acquisition until the support driver returns within cap. Framed this way, growth has to earn permission by staying inside audited constraints. Re-running that battery by hand every quarter is tedious in a spreadsheet, which is why some teams park their segment assumptions in a standing unit economics model and just fire the shocks at it again.
Three verdicts: pass, conditional pass, fail
An audit should end in a short set of verdicts and remediations. A segment passes when it clears the margin floor and payback boundary under both the base case and mild shocks. It earns a conditional pass when the base case holds but a single shock breaks it, which points to targeted remediation. It fails when the unit is negative-margin, or when payback sits beyond the boundary even in the base case.
Each remediation names a driver and a lever. Pricing and packaging work aligns revenue to cost drivers through bands and restructured tiers. Channel work caps spend, changes targeting, or shifts creative to pre-qualify cohorts. Product work improves time-to-value and pulls down early churn and ticket volume. Operations work automates support, standardizes onboarding, and cuts implementation hours. Vendor work renegotiates fees, reduces per-action usage, and adds caching or batching.
Three redesigns show what those levers look like in practice; all three are composites of common patterns. A B2B signature-workflow tool ran a "free forever" tier that generated heavy support usage without revenue, while its healthy-margin paid tier suffered because support was overwhelmed and retention slipped. The redesign constrained the free tier by usage and support access, added self-serve onboarding and documentation, and introduced a mid-tier that monetized high-usage free users without enterprise overhead.
A consumer marketplace watched partner revenue-share creep erode net revenue per unit; acquisition looked efficient until payback failed once the shares were subtracted. The fix renegotiated shares against volume thresholds, moved growth toward direct channels for the better-margin segments, and added partner-exclusive tiers priced to protect the margin floor.
A developer platform depended on a third-party API whose per-action fees dominated margin, with a small subset of users triggering excessive vendor calls. The redesign introduced caching and batching, aligned pricing to action volume, and added monitoring alerts so customers could see and throttle their own usage. Each one starts from a verdict and ends at a driver-aligned intervention.
Three questions come up whenever the calculator is reviewed.
Which metric is the most important audit output? Segmented contribution margin and payback boundary compliance. Together they decide whether scaling is safe and whether you can afford acquisition for each segment.
Why does the audit emphasize net revenue instead of list price or GMV? Because you cannot pay bills with gross numbers. Refunds, disputes, partner shares, and fees materially change unit profitability, and leaving them out invites false scaling decisions.
How often should a unit economics audit be run? Keep it lightweight and regular (monthly for fast-moving businesses, quarterly for slower ones) and re-run it after any major change: pricing, new channels, major feature releases, vendor cost changes, or policy changes.
Start the audit at the denominator
Start by checking the unit definition; inconsistent denominators can make otherwise accurate revenue, cost, and acquisition figures impossible to reconcile. Before you touch CAC, LTV, or a single scenario, force one unambiguous unit (the thing that carries both revenue and variable cost) and make acquisition, finance, and support all report against it. That one act reconciles arguments that otherwise run for quarters.
From there the sequence is not negotiable: net the revenue, attach the costs that actually grow with volume, walk the margin down, and only then look at cohorts, timing, and shocks. If a segment fails, resist the urge to "optimize": scale only multiplies a negative unit. What the audit buys you is the discipline to say which growth you have earned the right to fund, and which you have not.