Home · Guides · Revenue build by driver

Guide · ~8 min read

Building revenue by driver & the ARR bridge

The fastest way to lose a room is to type "revenue grows 100% a year" into a cell. An investor's next question is "from what?" — and if the answer is "from a hard-coded growth rate," the model is decoration. A driver-based build answers the question before it's asked.

Drivers, not outcomes

A driver-based revenue build starts from things you can defend with a hiring plan and a sales pipeline, and lets the growth rate fall out as a result:

Change any one of these and the whole model recomputes. Nobody can accuse you of a hockey stick, because the hockey stick — if there is one — is visibly the output of a sales-capacity assumption you can point to.

Step 1 — the customer roll-forward

Every year: begin with last year's customers, add new logos, subtract churn, end with a new count. Using the fictional sample company (14 starting logos, 46 new in year one, 1.2% monthly logo churn), with new adds treated as arriving mid-year so they only churn for half the year:

Customer roll-forward202620272028
Beginning customers14.054.9113.5
+ New logos4671110
− Churned5.112.423.0
Ending customers54.9113.5200.5

New logos grow 55% a year here (46 → 71 → 110), reflecting a sales team that scales — not an arbitrary revenue percentage. Fractional customers are an artifact of applying a monthly churn rate over the year; they round in the presentation layer.

Step 2 — revenue from customers × ARPA

Recognized revenue for a year is roughly the average customer count (not the ending count — customers who joined mid-year weren't paying all year) times ARPA times 12:

Worked — year 1

Average customers = (14.0 + 54.9) / 2 = 34.45. Revenue = 34.45 × $2,400/mo × 12 = $992,126. Using the ending count instead would overstate year-one revenue by ~60% — a classic beginner error investors catch instantly.

Step 3 — the ARR bridge investors ask for by name

Annual recurring revenue (ARR) is the run-rate: customers × ARPA × 12 at a point in time. The ARR bridge is the one exhibit a growth investor asks for by name, because it separates the four forces moving your revenue:

Beginning ARR  +  New ARR  +  Expansion  −  Churn  =  Ending ARR
ARR bridge202620272028
Beginning ARR$403,200$1,802,508$4,303,084
+ New ARR (new logos)$1,324,800$2,167,488$3,559,565
+ Expansion (upsell on base)$120,318$537,880$1,284,069
− Churned ARR$45,810$204,792$488,895
Ending ARR$1,802,508$4,303,084$8,657,823

Read year one across: $403,200 beginning + $1,324,800 new + $120,318 expansion − $45,810 churn = $1,802,508 ending. It closes exactly. This is what "the bridge ties" means, and it's the source of your retention metrics — NRR and GRR are just the expansion and churn lines above, divided by beginning ARR (see the metrics guide).

Why the bridge is worth more than the total

Two companies can both grow ARR from $1.8M to $4.3M and be completely different businesses. One does it with a fat expansion line and small churn — a strong, retentive product. The other does it by pouring on new logos to outrun heavy churn — a leaky bucket that gets more expensive every year. The total hides this; the bridge reveals it. That's exactly why investors ask for the bridge and not just the ARR line.

The honest caveat

A driver-based build makes your assumptions legible — it doesn't make them right. If your "new logos grow 55%/yr" isn't backed by a hiring and pipeline plan, the model is still fiction, just well-organized fiction. The value is that it forces you to name the assumption an investor can then pressure-test with you, instead of hiding it inside a growth percentage.

Build your own bridge in minutes.

ModelKit's revenue tab is fully driver-based: enter starting logos, adds, churn, ARPA and expansion, and it produces your customer roll-forward and a closing ARR bridge — the same exhibit investors ask for by name.

Request the kit — $149 See the ARR bridge

Educational material only — not investment, legal, financial, accounting, or tax advice, and not a valuation opinion. Figures are illustrative outputs of a deterministic model.

← All guides · Next: model mistakes in diligence →