Show the Assumptions That Make the Revenue Forecast Move

A revenue line that climbs from $60,000 a month to $70,000 over six months is a conclusion. It is the last item in a chain of events that have to happen first, and the line itself says nothing about whether they will.

That is the problem a reader has when they meet a smooth growth curve in a deck. They can see the endpoint. They cannot see the business behavior underneath it — which customers start paying, when, at what price, and what has to be available for that to be possible. Your job is not to show every cell in the model. It is to expose the few inputs whose movement would actually change the line, and to be honest about which of those inputs you know and which you have chosen.

Start with the commercial event behind revenue

Revenue arrives at a moment that a finance owner picks, and that moment sits downstream of several things that are easy to blur together.

A lead is someone who might buy. Nothing is earned, nothing is due. A signed agreement is a commitment with terms — for a subscription, usually a contract that says the customer owes a recurring amount over a period. An active account is one that can actually use the product: implementation is finished, the users are live, the service is doing what the customer pays for. An invoice is the bill. Collection is the cash.

These can land in four different months for the same customer. A practice signs in March, goes live in May, gets billed in May, pays in June. Sales reports a win in March. Revenue lands in May and onward. Cash arrives in June. One customer, three calendar positions, and a reader looking at your deck has no way to tell which one your growth line is describing.

So decide, with whoever owns the numbers, what the line represents. If it is recognized revenue, use the activation and contract dates that finance uses. If it is bookings or contract value, label it that way, because a bookings line and a revenue line will diverge exactly when your delivery is slower than your selling — which is often the interesting part of the story, not a detail to smooth over.

The temptation to describe a signature as a paying customer is strong, because it makes the chain shorter and the growth faster. Resist it. A reader who later discovers that the March win was really a May activation has learned something about your forecast that you did not intend to teach.

Show the minimum causal chain

For a subscription business, the whole thing can be said in one direction of travel:

signatures → activation queue → active accounts → revenue

Active accounts multiplied by revenue per account gives you the period's revenue. The active base changes when additions arrive and when departures leave. Additions don't appear the moment someone signs; they appear when delivery can start.

That's four or five moving parts. Everything else — brand awareness, hiring, marketing spend, customer acquisition cost — matters to the business, but it belongs in the forecast only if it changes one of those parts. A driver list of twenty adjustable cells is not more rigorous than a list of four. It just moves the hard question out of sight.

Consider a small company, invented here for the exercise. Call it Harborline. It sells shift-scheduling software to multi-location dental groups at $500 per practice per month. On January 1 it has 120 active practices, and 18 practices that have signed but haven't gone live yet. Sales is signing about 9 practices a month. The implementation team can bring 6 practices live a month. Two practices leave each month.

One convention has to be stated before any number means anything, and it has to be stated separately for each quantity in the table. Activations and departures take effect at month end: a practice that goes live in January enters the active base on February 1, not January 1. The revenue column is recognized revenue, and recognition for a practice begins in the month it first appears in the base — so that January activation is recognized in February. Cash is a third date: it arrives one month after the month it is recognized, which puts that same practice's first payment in March.

That convention is chosen, not discovered. Change it and every monthly figure below changes without a single customer doing anything differently.

The line rises $2,000 a month, which is four net additions at $500. Quarter one totals $186,000; quarter two totals $204,000.

Two things in that table deserve more attention than the revenue column. First, the queue grows from 18 to 36 across six months while the revenue line stays perfectly smooth. The line is silent about a constraint that is accumulating underneath it. Second, look at what the assumptions require over a longer horizon. The queue cannot grow forever; no company signs nine and delivers six indefinitely. And departures are stated as a flat count of two, which means the churn rate quietly falls from about 1.7% of the base in January to about 1.4% by June, purely as a side effect of growing. Neither of those is a fatal flaw. Both are things a sharp reader should be able to ask about, and things you should be able to answer without opening the model.

Separate evidence from chosen assumptions

Every row in that table rests on something. The useful discipline is naming what, for each one, because a number that is measured and a number that is desired look identical once they're typed into a cell.

There are roughly four kinds of basis, and they deserve different language in a deck:

• Recorded behavior. The 120 active practices and the 18 in the queue are counts from a system, on a stated date. These are the strongest rows you have. Keep the date attached — "120 active as of January 1" is a fact; "120 active" invites the reader to guess how stale it is.
• An agreement. The $500 monthly price comes from signed contracts. It is still an average across plans, so a shift in what customers buy will move it. Say that, or someone will treat it as fixed.
• An observed rate over a window. The two departures a month is the weakest observed row, because a small base produces lumpy churn and because the window matters. Two departures a month over a quarter is not the same evidence as two a month over two years.
• A chosen assumption or a target. The nine monthly signatures is the sales leader's plan. The six activations a month is an operations plan resting on two specialists each bringing three practices live. These are legitimate inputs. They are not evidence that the future will cooperate.

The failure mode to watch for is missing evidence acquiring authority through precision. A conversion rate of 3.7% looks measured; if it came from a sales target divided by a pipeline count, say so. An industry churn benchmark is not your churn. A published growth rate for your category is not your growth rate — it describes a population, and your business is one member of it, possibly an unusual one.

US small-business planning guidance makes roughly the same move: the SBA's "Plan your business" material connects marketing and sales choices to financial projections and asks for the assumptions to be explained (reviewed 20 September 2026). That is guidance about what to explain. It is US planning material for small businesses, and it does not validate any particular formula, driver, or rate — so it can support the shape of your disclosure, not the numbers inside it.

An assumption table does the disclosure work compactly. Four columns are enough.

The owner column is not decoration. It tells the reader who can update a driver when reality contradicts it, and it prevents the common situation where a number changes in the deck and nobody can say who decided.

Trace a changed assumption to its consequence

A forecast becomes a usable document when a reader can see what happens to the output when one input moves. So move one.

In the baseline, two implementation specialists are in place from January, which is where the six-a-month capacity comes from. Now suppose the second specialist's start date slips from January 1 to April 1. Capacity is three a month for January, February, and March, and six a month from April. Hold everything else fixed: the same nine signatures a month, the same two departures, the same price, the same convention. This is a delivery hypothesis changing. It is not a new hire creating demand, and nothing in the exercise assumes that it does.

January is identical, because January's revenue comes from the active base at the start of the month and the delay hasn't touched it yet. The gap opens in February and widens through April, then stops widening. By the end of June the delayed case has 135 active practices against the baseline's 144 — nine fewer, which is $4,500 a month, and the same $4,500 in May and June. Quarter one is $4,500 short; quarter two is $13,500 short; and if the horizon ran to a third quarter with the same inputs, it would be $13,500 short again rather than worse.

That detail matters, because it's easy to write about a delay as though it compounds. Here it doesn't. Once capacity recovers, both cases add four net practices a month, so the level stays nine practices apart while the level keeps rising. A reader who understands that will not over-read a slow quarter as a business in decline.

Those nine practices did not vanish. They are sitting in the queue, which in the delayed case reaches 45 by the end of June instead of 36. That is the consistency check to run whenever you change an activation number: signatures minus activations has to show up somewhere. If you reduce activations and leave the queue unchanged, you have quietly deleted customers.

Two other checks are worth naming. Make sure the inputs that depend on a changed driver actually moved — in this case activation dates feed the start-of-month base, which feeds revenue, which feeds the quarterly totals, while price, departures, and signatures stay where they were. And confirm that finance is still applying the same recognition convention, because a timing change is exactly when a convention gets applied inconsistently across periods.

One risk sits outside the exercise rather than inside it. The delayed version holds signatures at nine a month, so it assumes every practice that signs is still willing to wait until April. If a longer wait costs you some of those signatures, that is a separate assumption, it needs its own evidence, and it should be added deliberately rather than smuggled in as a discount on the delayed case.

What the reader should be able to say back

By the end, someone reading your deck should be able to state the conditions in plain words: we have this many paying customers today, each paying about this much; we add customers when they sign and when implementation can bring them live; some leave; the line assumes these particular numbers for each. And then, without opening the spreadsheet, they should be able to name what would change it.

For this illustration, the driver that deserves the most scrutiny is also the least financial-looking: the date the second implementation specialist starts, because it moves nine practices and $4,500 a month without touching the sales story at all. The next most useful evidence is the recorded counts with their dates — the active base and the queue — since those are the only rows that are genuinely facts rather than plans. After that, a departure rate measured over a stated window, in place of a flat count; a signature figure that is labeled as a target or as an average over a specific period; and a confirmation that the price is an average across a mix that could shift.

A revenue forecast is a conclusion. The assumptions behind it are the argument, and an argument is the part a reader can question, test, and believe.

Harborline and all figures in this article are invented for illustration. A real forecast needs its inputs read from current records and its dependency logic checked by the person who owns the numbers.

Frequently asked questions

Why can a smooth revenue line mislead a reader?

Revenue arrives at a moment that sits downstream of several events. A lead, a signed agreement, an active account, an invoice, and collection can land in four different months for the same customer. A signature is not a paying customer, so decide with the numbers owner what the line represents: recognized revenue, bookings, or contract value, and label it accordingly.

What is the minimum causal chain for a subscription forecast, and what convention does the example use?

The chain runs signatures to activation queue to active accounts to revenue; active accounts multiplied by revenue per account gives the period's revenue. The example states a chosen convention: activations and departures take effect at month end, recognition begins in the month a practice first appears in the base, and cash arrives one month after recognition. That convention is chosen, not discovered, and changing it changes every monthly figure without any customer behaving differently.

What happens in the example when the second implementation specialist starts late?

If the start slips from January 1 to April 1, capacity is three a month for January through March and six a month from April, with everything else held fixed. January is identical, the gap opens in February and widens through April, then stops widening. By the end of June the delayed case has 135 active practices against the baseline's 144, nine fewer and $4,500 a month; quarter one is $4,500 short, quarter two is $13,500 short, and a third quarter with the same inputs would be $13,500 short again rather than worse. The nine practices sit in the queue, which reaches 45 instead of 36. Whenever activation changes, signatures minus activations has to show up somewhere.

How should evidence be separated from chosen assumptions?

Use four kinds of basis: recorded behavior, an agreement, an observed rate over a window, and a chosen assumption or target. An assumption table with Input, Basis, Affected output, and Owner does the disclosure compactly. Watch for missing evidence acquiring authority through precision, and remember that an industry benchmark is not your churn and a category growth rate is not your growth rate. SBA planning guidance supports the shape of the disclosure, not the numbers inside it.

What are the main limits of the Harborline example?

Harborline and all figures are invented for illustration. A real forecast needs its inputs read from current records and its dependency logic checked by the person who owns the numbers. The delayed version holds signatures at nine a month, so it assumes every practice that signs is still willing to wait until April; if a longer wait costs signatures, that is a separate assumption needing its own evidence. The flat departure count also means churn quietly falls from about 1.7% to about 1.4% as the base grows, and an observed rate from a small base or a short window is the weakest observed row.