Calculating marketing ROI when the payback takes months

The textbook formula is:

ROI = (revenue − cost) ÷ cost

Spend $1,000, earn $3,000, ROI is 200%. Fine for a one-off transaction. Wrong for anything with a subscription attached, for three reasons.

Why it breaks

Revenue arrives over time. A customer acquired for $200 paying $50 a month has produced $50 by the end of the first month. The naive calculation says you lost 75% and it is a disaster. By month twelve the same customer looks spectacular. Both readings are artefacts of when you looked.

Revenue is not margin. $50 of revenue at 75% gross margin is $37.50 of contribution. Using revenue overstates every channel by whatever your hosting, payment fees and support cost.

Spend and result are in different periods. Content published in January produces customers in June. Dividing June revenue by June spend measures nothing.

What to use instead

Contribution over a fixed window

Pick a window - 12 months is standard - and compute:

Contribution = customers acquired
             × monthly revenue
             × gross margin %
             × expected months retained (capped at the window)

Then (contribution − spend) ÷ spend.

The cap matters. Uncapped, you are projecting years of revenue from customers you acquired last week, and the projection is dominated by a churn estimate you do not have enough data for.

CAC payback period

CAC ÷ (monthly revenue × gross margin %)

$200 CAC, $50 revenue, 75% margin → 5.3 months until that customer has paid for themselves.

For a business funding growth from revenue, this is the number that decides whether you can spend more. Under 12 months is healthy for SMB self-serve; over 18 means you need capital to grow at all.

A channel with excellent long-run ROI and a 20-month payback can still bankrupt you before the returns arrive.

Attribution you cannot resolve

Someone reads a blog post, forgets it, sees a mention two months later, searches your brand name and signs up. Last-click attributes it entirely to branded search. First-click attributes it entirely to the post. Both are wrong.

Rather than buying an attribution model you will not trust:

Segment by channel and compare rates, not totals. Branded search converting at 20% and cold organic at 2% is a real, usable difference regardless of the model.

Watch the branded-search trend. Brand searches rise when awareness work is landing. It is the cleanest available proxy for the effect of everything unattributable.

Run holdouts where you can. Pause a channel for a month and watch what happens to the total. Blunt, and more honest than any model.

Tagging every link consistently is the precondition for any of this. Without it you are reasoning about a "direct" bucket that contains everything.

A worked example

$2,000 on content in Q1. By Q4, 40 customers attributable to organic search, $40/month each, 80% gross margin, average 14 months retained.

Contribution = 40 × 40 × 0.80 × 12 (capped) = $15,360
ROI = (15,360 − 2,000) ÷ 2,000 = 668%
Payback = 2,000 ÷ (40 × 40 × 0.80) = 1.6 months

Both numbers are needed. The ROI says the channel works. The payback says you can afford to do it again immediately, which is the decision you actually have to make.