IGNITE EXTENDED / LESSON 19
How does the operating process generate the forecast?
A model map linking hiring, units, subscribers, inventory and cash.
The supplied case describes connected shoes sold for $400 and a related subscription priced at $15 per month. Customers begin a subscription when they buy shoes. A productive salesperson sells 100 pairs per month after three months of onboarding.
These facts imply separate model schedules: hiring and productive sales capacity, hardware units, subscriber cohorts, production and inventory, and operating expenses. A single revenue growth percentage would hide the training lag and the difference between a one-time shoe purchase and recurring subscription receipts.
A new salesperson incurs salary before reaching the stated productive capacity. The case uses $150,000 annual salary and a $40 commission per pair. At full productivity, 100 monthly pairs generate $40,000 hardware revenue and $4,000 commission.
Model each hiring cohort so that productivity begins after its training period. Track departures and salary changes separately. If the hiring date shifts, revenue can move by several months while payroll begins immediately. This is one reason a cash model must follow the operating process.
An operating model translates how the company works into quantities, timing and cash. Each important total should be traceable to a visible driver.
Describe what must happen: hire salespeople, train them, acquire customers and deliver the product. Hiring and productive capacity are separate events when training takes time.
Calculate productive sellers × units per seller × price. Model subscriptions as cohorts with an explicit start, billing and churn convention, rather than attaching growth to an unexplained percentage.
Add payroll from the hiring date, commissions under their rule and delivery costs when incurred. Then distinguish recognized expense from when the obligation is actually paid.
Reconcile opening cash, receipts, payments and financing each period. Locate the lowest balance and identify the assumption or action that could change the funding requirement.
The one-time versus recurring distinction below matters because a sale creates a unit once while a subscription produces a stream that depends on continued participation.
Illustrative month: 10 × 100 × 5% = 50 sales. At $200 collected per sale, monthly receipts are $10,000.
A product sale and a subscription create different revenue patterns. A forecast must define when a customer joins, pays and leaves.
Sales in the period × price, with explicit collection timing.
Existing customers retained + new customers, multiplied by the applicable price; account for churn, timing and service costs.
Illustrative application: 200 device sales create device revenue once. If those buyers begin $15 monthly subscriptions, future subscription revenue depends on activation, billing timing and survival. Adding 200 every month while ignoring churn overstates the subscriber base. New customers can also conceal decay in an older cohort.
Keep device units, subscriber balances and cohort survival as separate schedules linked by explicit assumptions.
A timeline starting at 'early stage' where investors weigh the founder's story, then early metrics split into transaction-focused and customer-focused measures.
Before revenue is meaningful, the model's job is to make the story's operating claims measurable. For Relay that means fill rate, clinics live, churn and CAC rather than a revenue line extrapolated at a growth rate.
The continuation of the timeline: operating-unit and company-level metrics, then full financials and valuation as a company matures.
The model's emphasis shifts along this line. An early model that leads with a five-year P&L and a valuation skips the metrics investors actually use to judge an early company.
Defines free cash flow as cash from operations net of capital expenditure; mature firms use positive FCF to pay investors, while startups run negative FCF, called the burn rate, funded by outside capital.
The model's bottom line for a venture is FCF by month, because that is what the next financing must cover. Relay's late-FY1 burn of about $60,000 a month is negative FCF, not the operating loss.
A direct method (cash received minus cash paid for employees, inventory, other operations and capex) marked wrong, and an indirect method (EBIT minus taxes plus non-cash items minus working-capital change minus capex) marked right.
Both reach the same number if done correctly; the indirect route is preferred in a model because it ties FCF to the income statement and makes working-capital effects, such as Relay's prepaid deferred revenue, explicit.
Asks students to explain in one or two sentences how the case model produces hardware revenue, subscription revenue, expenses and free cash flow.
If you cannot describe a block in two sentences of operating logic (who does what, how often, at what price), the block is not yet a driver-based model.
The case initially assumes 10% monthly churn and later 15%. A model must state whether churn applies to opening subscribers, new subscribers or both, and when subscription receipts begin. Different conventions can produce different monthly totals.
Under a simple convention that churns opening subscribers and then adds new buyers, closing subscribers equal opening subscribers multiplied by retention plus new subscribers. Subscription revenue and the $3 monthly service cost must use an explicitly chosen active-subscriber measure.
Victoria purchases electronics one month before a shoe production run. Shoe manufacturing has quantity-based prices. The business therefore pays for some inputs before it records the associated shoe sale and cost of sales.
Track units ordered, units received, units sold and units remaining. Link the electronics payment to the preceding month. A larger production batch may reduce the assembly price while increasing inventory cash requirements and the risk of unsold stock. A cheaper unit cost can consequently increase the amount the company needs to raise.
Illustrative model. The driver assumptions need evidence; arithmetic alone does not validate them.
Identify the key processes, forecast their non-financial metrics first, then refine as understanding improves.
Dollars come last. For Relay the non-financial metrics are sellers, clinics added, churn, messages sent and headcount; revenue and cost follow from them.
Shoes sold per month equals the number of productive sales reps multiplied by shoes per rep per month, with the prompt to refine it.
The word 'productive' carries the hiring lag. Headcount and productive capacity are different series in the model.
Expands shoes sold into first visits and follow-on visits to running clubs, members per location, purchase rates, retention and conversion, all multiplied from productive reps.
Each box is a testable assumption with its own evidence. Note the tree separates first-time conversion from repeat purchase, the same distinction Relay needs between new clinics and expansion within multi-location groups.
Selected reading: PDF pages 1, 2, 3, 4
The fictional case builds a financing decision from a connected operating forecast. Hardware sales, subscriptions, hiring and inventory create different timing patterns that must reconcile into cash.
Salespeople require three months of onboarding before the assumed monthly sales productivity begins. Payroll and productive capacity therefore start at different times.
A shoe purchase creates hardware revenue and starts a subscription. Continued software revenue depends on the population billed and the churn convention.
Electronics are purchased ahead of production and shoe costs depend on run quantity. Inventory timing and unit costs affect the financing requirement.
The case links the raise to shoe sales and a subscriber base retained for a stated period. A credible funding amount follows the monthly path to those conditions.
Selected reading: PDF pages 1, 2, 3, 4
Sales productivity and churn are model inputs in this fictional case. Label them accordingly when interpreting the forecast.
Define when hiring, training, sales, billing and churn occur. A one-month shift can change both the forecast and its minimum cash balance.
The supplied tax and capital assumptions belong to the exercise. A full venture forecast requires the actual obligations, rights and operating schedules.
Bessemer's State of AI 2025 reports fast-scaling AI application companies averaging about 25% gross margin and steadier ones about 60%, compared with the 75%+ often expected of SaaS. The difference is largely model inference that scales with usage.
For Relay Scheduler, inference cost belongs in its own schedule: conversations per clinic per month, tokens per conversation and price per token, alongside SMS fees. This lets the team test how much routing, caching or a smaller model protects the bible's 72% to 80% margin range.
Sources: Bessemer Venture Partners: The State of AI 2025 · SaaStr: Have AI gross margins really turned the corner?
Use the case’s $150,000 annual salesperson salary, three-month training lag, 100 monthly pairs at full productivity, $400 shoe price, $40 commission, $15 monthly subscription and $3 monthly service cost. Illustrative timing assumptions: two hires begin together, complete three full months of training, and new subscriptions begin the first productive month. Ignore raises, churn and all other costs for this isolated first productive month.
Calculate training-period base payroll, first productive-month shoe units, hardware revenue, commission, subscription revenue and subscription contribution. Explain which payments the simplified calculation leaves out.
Relay starts FY2 with 60 clinics paying $399 a month at 80% gross margin (v1; Relay Scheduler's inference cost is not yet in the model), 3% monthly logo churn and $924,000 of cash. Late-FY1 burn is about $60,000 a month with 9 people. The plan is to grow to 22 people by the Series A, one hire a month from month 2, at an illustrative fully loaded $12,500 per new hire per month. Two sellers close about 5 clinics a month at $2,400 CAC. A PT-association referral channel at $1,500 CAC, with no revenue share, may go live later.
Build a monthly model (24 months) with separate schedules for headcount, clinic cohorts (opening, churned, new, closing), revenue, cost of revenue, operating expenses and cash. Every total must trace to a named driver. Reproduce months 1 to 4 and state your timing conventions.
Payroll begins at hire; productive selling begins after ramp. A model that collapses the two pulls receipts forward and hides the cash low point.
Whether churn hits opening or closing clinics, and whether new clinics bill in their first month, changes monthly revenue. State the convention on the assumptions sheet and use it everywhere.
Annual prepayments put a year of cash into one month. Keep a deferred-revenue schedule so that recognized revenue, cash receipts and the liability reconcile each month.
A 24-month driver-based model for Relay with headcount, clinic cohort, revenue, cost and cash schedules; month 1 burn of $59,848 reconciling to the bible's ~$60,000; and a list of missing schedules to add.
Next lesson: Scenarios, sensitivity & cash requirements
Check your understanding
Question 1
Relay opens a month with 60 clinics. Churn of 3% applies to opening clinics and 5 new clinics are added at month end. How many clinics open the next month?
60 x 0.97 = 58.2 retained, plus 5 new = 63.2. Adding 5 and ignoring churn (65) overstates the base; applying churn to the new clinics as well changes the convention and must be stated if chosen.
Question 2
Relay hires an account executive at a fully loaded $12,500 a month. After a three-month ramp, the AE closes 2.5 clinics a month at $399 and 80% gross margin. What is spent on the AE before the first closed clinic, and what monthly gross profit does each month's batch add?
Three ramp months cost 3 x 12,500 = $37,500 with no sales. Each later month's 2.5 clinics add 2.5 x 399 x 0.80 = $798 of monthly gross profit. Using revenue ($997.50) instead of gross profit overstates the contribution.
Question 3
Twenty clinics prepay the annual plan of $4,070 at the start of the year. In the month they pay, how much cash comes in, and how much revenue is recognized that month?
Cash is 20 x 4,070 = $81,400; revenue is earned monthly, 81,400 / 12 = $6,783. The difference becomes deferred revenue. A model that books the cash as revenue overstates the month's income and understates later burn.
Question 4
A founder's model shows revenue growing 12% a month from a single growth input. What is the strongest objection?
The problem is not the number but that it has no mechanism. A driver model makes the 12% an output of sellers, productivity, churn and price, each of which can be checked against evidence and stressed in a scenario.
Question 5
Relay Scheduler adds LLM inference cost. If gross margin per clinic falls from 80% to 72% on $399, by how much does monthly gross profit per clinic fall, and where should the model book it?
399 x 0.80 = 319.20 versus 399 x 0.72 = 287.28, a $31.92 drop. Inference is incurred to deliver the service to each clinic, so it is a variable cost of revenue. Modeling it per conversation lets the team test caching or smaller models.
Question 6
A model treats each new hire as fully productive from the hiring month. What is the most likely error in the cash forecast?
Salary starts on day one while sales capacity arrives after the ramp. Pulling revenue forward hides the months in which payroll runs ahead of receipts, which is exactly where the minimum cash balance tends to fall.
Workbook
Module output: A linked financial analysis, operating forecast, valuation and staged-investment decision.
Sources
Original pages and full documents are in the Library edition.