A growth model is the connective tissue between the bookings plan, the operating budget, and the board narrative. It should tie every revenue dollar back to a channel, a cohort, and a conversion assumption that finance, marketing, and sales can all defend. This guide walks through the full build so revenue leaders can ship a model in one to two weeks, pressure test it against three scenarios, and keep it honest against monthly actuals.
Time: 1-2 weeks
Start by choosing a model shape that fits your business, not the one your last company used. The two primary axes are bottoms-up versus top-down, and cohort-decay versus cumulative. Bottoms-up builds revenue from channel-level inputs like leads, conversion rates, and ARPA, which suits product led and velocity motions where volume math dominates. Top-down starts from a market or capacity ceiling and allocates share down to segments, which suits enterprise motions with low deal count and long cycles. Cohort-decay projects each acquisition cohort forward with its own retention curve, while cumulative treats the installed base as a single pool. Most mature SaaS teams end up blending bottoms-up new business with cohort-decay retention, because the two sides of the P and L behave differently.
The quality of a growth model is set by the inputs, not the formulas. Build an input tab that holds every assumption the projection depends on, grouped by category: ARPA by segment and plan, CAC by channel, gross and net churn by cohort age, and growth rate per acquisition channel. Every input needs a source, an owner, and a last-updated date so the model can be audited later. Separate the inputs tab from the calculations tab so analysts can change assumptions without breaking formulas. Flag any input that is a placeholder in a different color, and track the delta between placeholders and sourced values as a quality metric for the model itself.
Project the model forward on a monthly grid for 12 to 24 months. Twelve months is enough for the operating plan, 18 to 24 months is where board narratives and hiring plans live. The output should show new ARR, expansion ARR, contraction, churn, and ending ARR for every month, with gross and net retention computed from the same cohort tables that drive the retention assumption. Pair the ARR view with a bookings view and a cash view, because the three numbers diverge in interesting ways under annual billing and multi-year deals. Keep formulas horizontal and consistent across the row so a single cell change can be audited by dragging across periods.
A single-point projection is a wish, not a model. Build three scenarios on top of the base case: a downside that stresses the two inputs most likely to disappoint, an upside that funds the two levers most likely to compound, and a stretch that assumes a step function in one specific channel. Run each scenario by cloning the input tab, not by overwriting the base, so the three views can be compared side by side. Tag every scenario with the two or three assumption changes that produce it, because the conversation on the output is really a conversation on which inputs the exec team believes. Keep the deltas transparent so the board can see which lever moves which number.
A growth model only earns trust when it is explained in plain language. Build a one-page summary that leads with the three scenarios, the two or three inputs that drive them, and the hiring plan, cash runway, and net retention each scenario implies. Pair the summary with a short appendix that shows the input tab, the cohort table, and the historical fit of the base case against the last four quarters of actuals. Walk the exec team through the model once before the board sees it, so finance, marketing, sales, and customer success can all stress test the assumptions in their domain. The board conversation should be about which scenario to fund, not about which cell is wrong.
A growth model that lives on three laptops is not a growth model. Pick one location, usually a shared drive folder owned by finance, and declare it the only authoritative version. Version control the file by date in the filename, and keep an archive of every monthly cut so the historical fit tab can be reconstructed. Grant edit access to a short list of named owners and comment access to the broader exec team, because scope creep on who can edit is how models drift. Link to the model from the operating cadence doc, the board materials folder, and the planning tracker so every downstream artifact pulls from the same cell values.
A model that is not refreshed against actuals is a slide, not a tool. Set a monthly cadence where finance drops in the actual period a few business days after close and the model recomputes forward on the latest cohort behavior. Keep the base case assumptions frozen between quarters so the drift between model and actuals is visible, then recalibrate on a quarterly basis as part of the forecast cycle. Treat each monthly refresh as a small variance review: did new ARR, churn, and expansion land where the model said they would, and if not, which input needs an update. Teams that run this loop every month cut their forward error materially within three quarters.
The point of the model is not the projection, it is the learning loop. Every quarter, decompose the variance between forecast and actual into the individual input drivers: did new logo volume miss, did ARPA move, did churn tick up, did expansion lag. Attribute every piece of the variance to a specific input and a specific owner, and feed those tags into the next quarter's input updates and into the sensitivity ranges. Over three to four quarters, the variance decomposition itself becomes a data set that tightens the next model, because you learn which inputs are reliable and which ones need wider bands. This is how modeling maturity compounds instead of resetting every year.
A SaaS growth model is a structured projection of revenue, retention, and cost over a defined period, built from auditable inputs by channel, segment, and cohort. It is used to set the operating plan, size hiring and spend, and anchor the board narrative. A good model ties every output back to a sourced input so leaders can defend the number instead of arguing about the cells.
Twelve months is standard for the operating plan and the current fiscal year. Eighteen to twenty four months is appropriate when the model is feeding a board narrative, a fundraising conversation, or a multi year hiring plan. Beyond two years the uncertainty on inputs usually exceeds the signal, so most teams keep the long range view as a scenario sketch rather than a cell by cell projection.
Finance owns the model as the single source of truth, but revenue operations, marketing, and customer success each own the inputs for their domain. The CFO and the CRO co-sign the base case, because the model is where the sales plan and the finance plan have to reconcile. Any input change requires sign off from the named owner, so a stray edit never silently redirects the forecast.
ARPA, new logo acquisition rate per channel, gross and net churn, and expansion rate per cohort are the four drivers that move almost every SaaS model. CAC and sales cycle length matter for the efficiency view and the hiring plan. If you only have time to source five inputs well, these are the ones that pay back the research effort first.
A cumulative model treats the installed base as a single pool and applies one retention rate to the total. A cohort-decay model tracks each acquisition cohort separately with its own retention curve, which is much more accurate because retention almost always varies by cohort age, segment, and pricing vintage. Cumulative is faster to build, cohort-decay is more defensible once you have four or more quarters of history to anchor it.
Refresh monthly against actuals and recalibrate assumptions quarterly as part of the forecast cycle. Rebuild from scratch once a year, usually during the planning season, so the structure stays aligned with the motion and the segmentation the company is actually running. A model that is still built on last year's channels and segments quietly stops predicting anything useful.
Ship a growth model your board can defend, then pipe the inputs that drive it, pipeline, retention, expansion, straight out of Strkr so the model stays honest against actuals.
No sales call, no migration consultant, no four-month implementation. Enter your card, get 14 days of the full Pro tier, cancel any time before day 14 with zero charge. Spin up a workspace, import your CSV, and have something useful before lunch.
