A software dashboard shows $4,000 in monthly recurring revenue. The founder feels closer to a durable business. Yet one customer cohort consumes expensive processing, another receives frequent refunds and the founder spends evenings repairing files included in a low-priced plan.
The revenue total describes part of the business, not what remains to support it. Calculate contribution using a stated revenue period and cost boundary, then inspect the cohorts that create or consume it. Contribution is not full profit; fixed expenses, acquisition and other obligations still need their own treatment.
This article in The AI Software Factory owns the reconciliation. AI software costs builds the underlying ledger, while pricing chooses the offer. The existing true ecommerce profit chapter deals with orders and inventory; this chapter examines recurring software delivery.
State the operating definition first
For this decision model, contribution equals retained customer revenue for a defined period minus the attributable delivery costs included in the stated boundary. Contribution percentage equals that amount divided by the same retained revenue.
The words “retained revenue” and “attributable” need definitions. Does the revenue view account for refunds and discounts? Does the cost boundary include payment fees, support and account-level services? Write the answers before comparing products.
Formal accounting classifications may differ from this operating model. Keep the founder’s decision ledger explicit and reconcile it with the business’s actual accounting policies. This article does not establish universal treatment for every cost or provide a particular company’s tax calculation.
The purpose is to understand which customer work leaves resources available for the remaining business. A positive contribution can still coexist with a loss after fixed expenses or insufficient compensation for owner time.
Do not substitute MRR for receipts or contribution
Stripe’s subscription analytics documentation defines its MRR using monthly-normalized active and past-due subscriptions and excludes metered products, taxes and free plans. Discount settings affect that provider-specific calculation.
That metric can help describe a recurring subscription base. It does not show cash collected, all revenue in a usage model or delivery cost. A past-due account can affect the metric differently from the cash available to pay a vendor.
Keep a bridge between commercial records and the operating calculation. Record subscription obligations, invoiced usage, receipts, discounts, refunds and relevant period adjustments separately. Explain which view the contribution calculation uses.
An annual prepayment also needs care. A large receipt improves current cash while creating future delivery obligations. Monthly normalization and actual service-period costs can support analysis, but they should not be confused with a complete accounting revenue-recognition policy.
Match the period and the customer scope
A cohort calculation should compare revenue and cost from the same customer set and relevant period. Mixing all current revenue with last month’s partial cost ledger produces a number without a coherent meaning.
Define the cohort: acquisition month, plan, source format, assistance level or another decision-relevant boundary. Keep the account identities stable enough to trace upgrades, refunds and usage across the period.
Some expenses arrive late. Record the service event and invoice timing so an adjustment can be assigned correctly. Do not declare a cohort highly profitable merely because its vendor bill has not arrived.
State incomplete evidence. If support time or usage attribution is missing, show what is known and the unresolved amount. A qualified estimate can guide a bounded decision, but an exact-looking percentage should not hide missing inputs.
Work through an ordinary hypothetical account
Suppose an account pays $40 for the month after discounts, receives no refund and creates $12 in attributable delivery cost under the defined boundary. Its contribution is $28, or 70 percent of retained revenue. These are hypothetical numbers, not Salars results or vendor prices.
The remaining $28 still needs to support any excluded fixed expenses, acquisition, maintenance and owner compensation. Calling it profit would skip those obligations.
Now suppose delivery cost increases to $20 because the account submits larger supported inputs. Contribution falls to $20, or 50 percent, at the same revenue. The revenue dashboard has not changed, but the delivery relationship has.
This simple comparison gives the founder a specific question: did the offer intend to include that larger workload, and does the price boundary remain sustainable? The response may be a better architecture, a clearer allowance or a different plan rather than a general growth campaign.
Add refunds without erasing the work
Suppose the hypothetical $40 account receives a $10 refund after a failed job. Retained revenue becomes $30. If delivery cost remains $12, contribution is $18, or 60 percent of retained revenue.
The expense of the failed job remains even though part of the charge was returned. A refund changes revenue; it does not undo compute, extraction or human work already consumed.
Keep the reason for the refund. Product errors, unsupported inputs, commercial goodwill and cancellation disputes have different implications. A recurring product error suggests quality work; a misunderstood offer suggests clearer scope or readiness checks.
A refund policy is part of the customer promise. Do not change the contribution definition to make the remedy appear less costly. The ledger should reveal the actual economics of honoring the offer.
Include support under an explicit boundary
If an account requires thirty minutes of attributable support, the operating review should record it. A planning valuation of $30 per hour would represent $15 of time value in a hypothetical scenario.
Keep the time value separate from cash expenses when the founder provides the work without a corresponding payment. The combined decision view can be useful, but it should not imply that an invoice was paid.
For the earlier $40 account with $12 cash delivery cost, adding the illustrative $15 support valuation leaves $13 of contribution under the expanded boundary. That is 32.5 percent of retained revenue. A favorable compute-only margin has become a different relationship.
Profit per human hour examines owner time more directly. The contribution review should disclose whether it includes, excludes or separately presents that time rather than allowing the boundary to change between comparisons.
Aggregate cohorts without hiding the mix
Consider a hypothetical month with eighty ordinary accounts contributing $28 each and twenty heavy accounts contributing negative $10 each. Total contribution is $2,040: $2,240 from the ordinary group minus $200 from the heavy group.
At $40 retained revenue per account, total revenue is $4,000 and aggregate contribution is 51 percent. The total is positive, but the heavy cohort consumes contribution rather than adding it.
Growing the heavy cohort at unchanged terms worsens the relationship. Doubling only those twenty accounts adds revenue while reducing total contribution by another $200. This is why revenue growth alone cannot establish better economics.
Investigate what distinguishes the cohort. Larger inputs, repeated errors, unsuitable customer expectations or a broad support promise can each create the pattern. The ledger identifies where to look; it does not establish the cause without further evidence.
Compare marginal and fully allocated views
Adding one small account may create little additional shared expense. That marginal view helps answer whether the next account contributes under current capacity. It does not show whether the product can support all fixed obligations over time.
A fully allocated operating view assigns shared costs using a stated rule. It can help compare products or estimate total sustainability. The rule may be useful without representing a directly caused expense for each account.
Keep the views side by side when they answer different decisions. A product can have positive marginal contribution and insufficient aggregate contribution to cover its dedicated fixed platform cost. Another can look unattractive under a broad allocation while still helping use existing capacity.
Do not select whichever definition supports a preferred conclusion. State the decision, choose the relevant view and preserve the alternative limitations. Consistent definitions make the review useful across periods.
Reconcile acquisition separately
Customer acquisition can require advertising, sales time, commissions or onboarding effort. Those obligations matter, but the contribution calculation should explain whether they are included or assessed separately.
A product with positive recurring contribution may take too long to recover acquisition spending. A hypothetical $100 acquisition expense and $20 monthly contribution suggests five months of contribution to recover that expense under simplified assumptions.
That scenario ignores churn, delayed collection, additional support and the time value of money. It is a planning illustration, not a forecast or a complete investment calculation. Actual cohort retention changes the result.
Separate one-time acquisition from recurring delivery so the founder can see both mechanisms. Hiding acquisition inside a generic cost average can make it difficult to choose between improving distribution and improving the product’s delivery economics.
Include cancellations and downgrades in cohort history
A cohort’s contribution can change as customers leave, reduce scope or move to a different plan. Preserve the original cohort identity while recording the current commercial state.
An account that cancels after expensive onboarding may leave a negative relationship even if its final month looks cheap. A successful account that upgrades may contribute more while creating additional support obligations. Both patterns need the full relevant history.
Do not remove canceled accounts from the calculation when evaluating acquisition cohorts. That would compare retained revenue from survivors with costs incurred for the original group and hide the failed relationships.
State the observation window. Young cohorts have not yet revealed their long-term retention or maintenance exposure. A promising first month is evidence about that month, not proof of lifetime value.
Recognize revenue that carries future obligations
Prepaid credits and annual contracts can collect cash before delivery. The operating review should identify the work still owed and the cost exposure associated with it.
A large balance of unused credits is not necessarily a favorable contribution result. Customers may consume the credits later, and the actual terms may create refund or expiration obligations. The model needs a stated treatment and professional review where accounting rules matter.
For planning, compare expected delivery scenarios and preserve the assumptions. A temporary low-consumption period should not justify spending all receipts as if no future usage will occur.
Pricing architecture explains credit and subscription tradeoffs. Contribution analysis makes their economic consequences visible without claiming a universal accounting treatment.
Stress-test the cohort before expanding it
A growth plan should examine changes in usage, failure rate, vendor cost, support and customer mix. Hold other assumptions steady initially so the team can identify the mechanism creating exposure.
Include realistic adverse cases. A provider outage can create retries and refunds. A new source format can increase manual correction. A popular feature can attract customers whose workload differs from the pilot.
Preserve customer acceptance and safety criteria. Reducing support by making errors harder to report does not establish better delivery. Passing costs to customers through unclear charges may improve a spreadsheet while weakening the offer.
Use the result to choose a bounded next step. A cohort that remains favorable under tested scenarios may deserve a limited expansion. A cohort with unresolved expensive exceptions may need a scope correction before acquisition spending increases.
Turn a margin problem into a mechanism question
A falling percentage can result from lower price, more cost, different mix or incomplete records. Separate those possibilities before selecting an action.
Trace a few representative accounts from accepted work to billing and cost. This review can reveal duplicate resource use, a misconfigured discount or support hidden outside the ledger. It can also reveal that the product simply promises too much for the price.
Choose an improvement with independent success criteria. If the hypothesis is that earlier validation reduces failed processing, compare failed-job cost while preserving valid-job completion. If the hypothesis concerns expensive manual exceptions, define the supported input boundary and evaluate affected customers.
Record proposed and executed changes separately. A model predicting higher contribution after a price change is a scenario until customers respond and actual delivery is observed. The business needs that uncertainty visible while deciding how much to commit.
Reconcile a cohort whose billing model changes
A customer moving from a flat plan to usage pricing can change the revenue pattern even if the underlying work stays similar. Preserve the plan version and effective period so the team can compare the old and new obligations accurately.
Do not treat all usage income as recurring subscription revenue merely to keep the dashboard familiar. Record the actual unit, accepted work and adjustments. The comparison should use the same relevant service period and cost boundary.
A proposed plan migration review would examine a light account, an ordinary account and a heavy account. It would show their expected bills under both architectures, then compare actual accepted work and cost after the change. Customer response remains uncertain until observed.
If the migration creates confusion or additional billing support, include that burden. A technically successful meter can still create a commercially expensive relationship. The contribution review should capture the new obligation rather than celebrate higher invoices in isolation.
Use currency consistently in the calculation
Revenue and expenses may arrive in different currencies or at different times. A decision model needs a stated conversion approach and a consistent period so exchange-rate movement does not masquerade as a product improvement.
Keep the original amount and currency available for reconciliation. Record the rate source or accounting record used for the operating view. Do not combine raw numbers from unlike currencies simply because the spreadsheet cells look identical.
The actual accounting treatment belongs to the business’s established policy and professional review. For a founder’s bounded planning scenario, the necessary discipline is to label the assumption and test sensitivity when currency exposure is material.
Know when the estimate is too incomplete to guide release
If a product lacks reliable usage attribution and support records, a polished margin percentage can create unjustified confidence. Identify the missing mechanism and collect a limited set of representative evidence before committing to broad growth.
The team can still make progress. It can reconcile a small pilot cohort, time specific exception categories and inspect invoices. It should state what those observations support and what remains unknown.
A stopping condition might be repeated inability to reconcile charges with work, or delivery costs that exceed the bounded pilot budget. These are proposed decision rules, not claims that a particular Salars product has met them. The point is to keep uncertainty from becoming an invisible unlimited obligation.
What Would We Do at Salars?
For a proposed merchant report app, we would publish an internal operating definition before reviewing the first cohort. Retained revenue, cash delivery costs and owner-time valuation would be explicit, with acquisition and shared expenses shown under their own stated boundaries.
We would reconcile each account’s accepted work, retries, credits, refunds and support. A proposed review would compare ordinary and heavy workloads without excluding canceled or failed accounts. The calculations in this article remain hypothetical; they do not report an existing Salars software margin.
If a heavy cohort consumed contribution, we would investigate the mechanism before buying more traffic. A clearer input scope, improved parser, revised allowance or separate setup service could each be appropriate under different evidence.
We would keep receipts and the billing provider’s recurring metric separate from the operating contribution view. A growing dashboard would prompt a review of what remains after delivery, not automatically justify expansion.
Contribution gives the founder a more useful question than revenue alone: does this customer relationship leave resources to sustain the rest of the business? The broader AI collection connects that question to acquisition, owner time and portfolio decisions.
Sources
- Stripe: Subscription analytics — provider-specific MRR, discount and subscription-state definitions; not a profit or cash-receipt measure.
Loading comments…