diff --git a/How-startups-forecast-recurring-revenue-growth%3A-churn-first.md b/How-startups-forecast-recurring-revenue-growth%3A-churn-first.md
new file mode 100644
index 0000000..12a0d14
--- /dev/null
+++ b/How-startups-forecast-recurring-revenue-growth%3A-churn-first.md
@@ -0,0 +1,152 @@
+
Forecasting recurring revenue growth starts with subtraction before addition. That is the practical answer to how startups forecast recurring revenue growth: take the revenue you already have, subtract what you expect to lose, then add new sales. The order matters because churn sets the baseline. Skip it and the forecast turns into a wish list with a spreadsheet around it.
+
+The churn number sets the ceiling for every growth forecast
+
+
Churn sets the ceiling because lost revenue has to be replaced before anything counts as growth. A hypothetical company billing $40,000 a month that loses $2,000 to cancellations does not start the next month at $40,000. It starts at $38,000, and the first $2,000 of new sales only returns it to where it was. Growth begins after the refill.
+
+
Two companies with identical new sales can finish a year in very different places for this reason. One keeps most of what it sells. The other sells the same amount and keeps less. From the outside they look equally good at selling. From the inside, only one is compounding, and the difference shows up months before it becomes obvious to anyone reading a summary.
+
+
Founders also tend to track churn in the wrong unit. Logo churn counts customers lost. Revenue churn counts dollars lost. A company can lose very few customers and still lose a lot of revenue if the ones leaving are its largest accounts. Watching the customer count alone hides that case completely.
+
+
The order that holds up is base, losses, then target. Write down the recurring revenue you bill each month, list the cancellations and downgrades you expect, subtract them, and treat what is left as the starting point. A growth target applies to that remainder, not to last year's total.
+
+
Reverse the order and the forecast drifts. Starting from a growth target and working backward into sales numbers skips the drag entirely. The plan reads as achievable and quietly assumes churn stays where it was. If churn moves, the plan is wrong before the first month is over, and nobody notices until the quarter closes badly.
+
+
Reading churn monthly keeps the surprise small. Attrition builds slowly in smaller accounts, and by the time it appears in a revenue chart it has usually been running for several months. A number you look at every month is a number you can argue with while there is still time to change something.
+
+Start from cleaned records, not from the invoice total
+
+
Take a hypothetical software company that bills monthly, has been selling for about two years, and keeps its numbers in a spreadsheet. The founder wants a forecast for the next four quarters. The sheet shows invoices, and that is the first thing to fix, because an invoice list and a recurring revenue list are not the same thing.
+
+
Recurring revenue is the part of the billing that repeats: subscriptions, support contracts, retainers. Setup fees, one-time projects and hardware do not belong in it, however large they look on a single invoice. Mixing them inflates the base, and every ratio calculated afterward inherits the error.
+
+
Cleaning the sheet takes an afternoon. Let's get specific: three columns do most of the work, covering customer, amount, and whether the amount repeats. A fourth column for the month a customer started lets the founder group accounts by when they signed. Those groups are cohorts, and they are what make churn visible later instead of hiding inside one average.
+
+
Once the base is clean, the founder can answer a question the sheet could not: how much revenue is already in place for next month if nobody sells anything at all. That figure is the real starting point. It is usually lower than the invoice total and much lower than the pipeline number the founder has been carrying around for months.
+
+
Cleaning also exposes items that were never forecasting problems. Renewals nobody confirmed, accounts still billed after a cancellation notice, discounts that were supposed to end last quarter. Each one distorts the base, and a forecast built on a distorted base fails later in ways that look mysterious from the outside.
+
+
One more pass is worth the time, and it is about money rather than customers. Revenue recognized under ASC 606 and cash collected are related but not identical. A subscription billed annually up front behaves differently in a cash plan than in a revenue plan, so keep both views and label which one you are reading.
+
+Sort renewals, upgrades, and downgrades in the same example
+
+
With two years of records grouped by cohort, the hypothetical company can see the three forces that move recurring revenue after the first sale. Renewals keep it flat. Upgrades add to it. Downgrades and cancellations subtract from it. A forecast is those three forces applied month by month to the current base.
+
+
Renewals are the easiest to predict and the easiest to forget. If a contract runs for twelve months, the renewal date is known well in advance. What is not known is whether the customer will sign again. Grouping renewals by month turns a vague worry into a list of accounts and dates that someone can actually work through.
+
+
Upgrades are the least reliable input of the three. Existing customers who add seats or move to a higher plan do so when their own business pushes them, not when your plan says it is time. A useful habit is to forecast upgrades at a level you have genuinely seen in past cohorts, rather than the level written into the sales plan.
+
+
Downgrades deserve their own line, separate from cancellations. A customer who drops from a larger plan to a smaller one is still a customer, so a headcount measure misses the change entirely while the revenue still falls. In the hypothetical company, separating downgrades from cancellations changed the churn picture considerably.
+
+
Many companies net expansion and contraction into one figure. That figure, often called net revenue retention, is useful at the top level and risky as the only view. A strong net number can sit on top of heavy losses underneath, and those losses are what stall growth once new sales slow down or the sales team has a weak quarter.
+
+
What the sorting produces is a monthly ledger of causes instead of a single churn percentage. Causes can be argued with: this renewal needs a conversation, that downgrade needs a reason, these two cancellations share a pattern worth investigating. A single percentage can only be accepted or denied, and neither response improves the forecast.
+
+Build the forecast in three buckets, not one line
+
+
The hypothetical company now has a clean base, a renewal calendar and a churn history. The temptation is to multiply the base by a growth rate and call the result a forecast. A three-bucket structure survives contact with real months far better than a single line, and it takes about an hour to set up.
+
+
The buckets separate revenue by how much control the founder has over it. Committed revenue is under contract and not up for renewal in the period. Likely revenue depends on renewals and upsells with a track record behind them. Uncertain revenue comes from deals still open or assumptions nobody has tested yet.
+
+
+
+BucketWhat sits inside itHow to treat itMain risk
+
+
+CommittedSigned contracts running past the periodCount at full valueRecord errors and missed cancellation notices
+LikelyRenewals and upgrades with historyApply your observed rate for that cohortRates drifting between quarters without anyone noticing
+UncertainOpen deals and untested assumptionsShow separately, never blended inBeing read as pipeline rather than as a forecast
+
+
+
+
The value of the table is that it makes the argument explicit. When committed revenue covers most of the period, the forecast rests on record quality and contract terms. When uncertain revenue carries the plan, the forecast rests on sales performance, and the honest label for it is a target rather than a projection.
+
+
Cash timing belongs in its own column. Annual contracts billed up front put money in the bank early, while those same contracts leave revenue to recognize later. Monthly billing does the opposite. A plan that mixes the two without saying so will look either too optimistic or too tight, depending on which month you read.
+
+
Revisit the bucket boundaries every month. What was uncertain becomes committed, and what was likely turns into either a renewal or a loss. Moving rows between buckets is the forecast working properly, not the forecast being wrong, and the movement itself tells you where to spend your attention.
+
+Test the churn assumption before you trust the number
+
+
Every forecast carries an assumption about churn, even when nobody writes it down. In the hypothetical company, the built-in assumption was that churn would stay near its recent average. The test worth running is simple: rebuild the plan with churn at a worse rate and see which commitments stop making sense.
+
+
Cohort effects make old averages misleading. Customers who signed during a period of heavy discounting tend to behave differently from those who signed later at full price, and they often renew on different terms. A single blended churn rate hides that difference until the discounted group reaches its renewal dates.
+
+
Volume matters as much as the rate itself. A company with a small number of accounts will see churn jump around from month to month for reasons that have nothing to do with the product. Reporting a precise-looking percentage built on a handful of customers gives a false sense of accuracy.
+
+
Seasonality is easy to miss and easy to check. If most contracts renew in the same two months of the year, an average spread across all twelve months will understate what happens in those two. Splitting the forecast by month, rather than by quarter, usually surfaces this within one [Review part-Time CFO responsibilities with Venture Growth Partners](https://Venturegrowthpartners.com/services/part-time-cfo/).
+
+
Four checks cover most of the weak spots:
+
+
+Does the churn rate come from cohorts of similar age, or from a mix?
+How many customers is the rate actually based on?
+Are downgrades included, or counted separately and forgotten?
+Do the last two quarters look like the average, or noticeably better?
+
+
+
There is a limit to how precise any of this gets. Early-stage companies operate with too few customers and too little history for a tight range, and pretending otherwise leads to decisions that treat a guess as a measurement. The honest output at that stage is a range and a list of the assumptions behind it.
+
+Questions to ask before you rely on any recurring revenue forecast
+
+
The fastest way to find the weak point in a forecast is to ask the person who built it a short list of questions. If the answers come back vague, the number on top is guesswork with nicer formatting. Run through these before the plan goes to a board, a lender or a business partner.
+
+
+Which part of this revenue is under contract, and which part is expected?
+What churn rate is built in, and which months of data produced it?
+Are downgrades counted separately from cancellations?
+Does the base include or exclude one-time work?
+What happens to the plan if churn comes in worse than the average?
+Does the cash view line up with the revenue view, or do the two tell different stories?
+
+
+
The first question does the most work. A forecast that blends signed contracts with hopeful renewals cannot be tested, because nobody can tell which [cfo part time](https://Venturegrowthpartners.com/services/part-time-cfo/) moved when a month goes wrong. Splitting the two makes the plan arguable, and an arguable plan improves over time.
+
+
The second question catches a common shortcut. A churn rate pulled from one strong quarter will flatter the whole plan. A rate pulled from an average that includes the first year, when most customers were new and still engaged, can understate what is coming as those accounts age.
+
+
Question six matters more than most founders expect, since revenue and cash rarely move together. A large annual invoice paid up front can make a single month look permanent. Monthly billing can make a healthy year look tight in the middle of it, even when the underlying business is fine.
+
+
There is a real difference between a forecast and a budget, and mixing them causes confusion. A budget says what you intend to spend. A forecast says what you expect to happen. Keep them in separate columns and compare them at month end, rather than blending the two into one number that answers neither question.
+
+Where part-time finance help changes a forecast, and where it does not
+
+
Outside finance support usually enters the picture when the forecast stops being believable. The founder is busy with customers, the sheet has grown into something only one person understands, and the next renewal conversation keeps getting pushed to Friday afternoon for the third week running.
+
+
What good support does first is unglamorous. It rebuilds the base, sets up cohort tracking, and keeps a renewal calendar that somebody actually works. Forecasts improve after that work, not before it, because the inputs are the forecast, and cleaner inputs change what the output can support.
+
+
From there, the scope generally covers a handful of connected jobs.
+
+
+Record quality and a monthly close, so the numbers feeding the forecast are trustworthy.
+Management reporting that separates committed, likely and uncertain revenue.
+Cash flow forecasting, including working capital and the timing of large invoices.
+Budgeting and planning tied to hiring and the cost side of the plan.
+Investor reporting and preparation when a financing, an audit or a transaction is on the horizon.
+
+
+
None of that removes uncertainty. Churn depends on customers, competitors and markets, and no reporting model changes that reality. A better forecast narrows the range of outcomes you are prepared for. It does not make the favorable end of that range certain, and anyone suggesting otherwise is selling something.
+
+
Stage decides how much machinery is worth having. A company with a handful of customers needs a plain sheet, a renewal list and one honest churn rate. One with outside investors, annual contracts and audit obligations needs more structure, because the same numbers now go to people who will ask detailed questions about them.
+
+
Cost is a real consideration, and scope drives it more than anything else. A narrow engagement around reporting and forecasting costs less than one that also covers bookkeeping, cash planning and investor materials. The useful question is which of those jobs is failing right now, not which package sounds most complete on paper.
+
+
Continuity is the part founders underestimate. Forecasts improve when the same person sees twelve months of misses and adjusts the assumptions accordingly. A rotating set of helpers restarts that learning each time, which is why the handover plan deserves as much attention as the initial setup.
+
+Three tests to run before you act on your own forecast
+
+
Before you commit to a hiring plan, a spending increase or a financing conversation on the strength of one number, put the forecast through three tests. Each takes a few minutes, and each fails often enough to be worth running every quarter rather than once a year.
+
+
+Does the base include only revenue that repeats?
+Can you name the churn rate, where it came from, and which customers it covers?
+Does the plan still hold if churn comes in worse than the average?
+
+
+
The first test is about discipline. One-time work drifts into the recurring base, especially during a strong month, and each time it does the growth rate looks better for a reason that will not repeat. Keeping those lines apart is the difference between a forecast and a highlight reel.
+
+
The second test is about evidence. If the churn rate cannot be traced to specific cohorts and a specific number of customers, treat it as a guess. Guesses are allowed in a forecast, provided they are labeled as guesses and checked against what actually happens in the following months.
+
+
The third test is the one that changes decisions. Run the plan with churn worse than your recent average and see which commitments still stand. If none of them do, the plan is leaning on an assumption rather than a margin of safety, and the assumption is doing work that a reserve should be doing.
+
+
Failing a test is not a reason to abandon the exercise. It is a reason to fix the input, write the weaker assumption into the plan, and look at the number again next month. A forecast you argue with every month is worth more than one you file away and trust without revisiting.
+
+
The first practical step is smaller than it sounds. Open last month's billing, mark which lines repeat, and total only those lines. That figure is the one worth forecasting from, and a better answer to how startups forecast recurring revenue growth starts there rather than with a growth target.
\ No newline at end of file