How to measure marketing ROI for SaaS, channel by channel

To measure marketing ROI for SaaS, take a channel's gross profit from billing-verified subscription revenue over a fixed window, subtract the channel's fully loaded cost, and divide by that cost.

The revenue comes from Stripe, Paddle Billing, Polar, Lemon Squeezy, Creem or Dodo Payments, because an ad dashboard only counts the conversions it claims.

A customer who cost $250 to acquire and pays $49/month at an 80% gross margin repays that cost in 6.4 months, and the same arithmetic works for every channel once the inputs are right.

The SaaS marketing ROI formula

A subscription keeps paying for months, shrinks through refunds and cancellations and grows through upgrades, so retail ROI math needs a window and a margin.

1loaded cost   = media + fees + creative + tools + commissions + salary share2CAC           = loaded cost / new paying customers3gross profit  = billing revenue in window, after refunds x gross margin4ROI           = (gross profit - loaded cost) / loaded cost5ROAS          = billing revenue in window / loaded cost6payback       = CAC / (monthly revenue per customer x gross margin)

ROAS stops at revenue. ROI and payback apply gross margin as well, which is how Bessemer Venture Partners measures CAC payback.

The window matters as much as the margin. On a $49/month plan the first invoice is $49, while a customer who stays 12 months pays $588, so dividing by first-payment revenue undercounts every channel that brings customers who stay.

A lifetime window overshoots, since nobody knows yet what this quarter's customers will pay in year 3. A fixed 12 months sits between the two.

How to measure marketing ROI for SaaS in 6 steps

Step 1: Total the full cost of each channel

For each channel add media spend, agency or freelancer fees, creative production, tools used only for that channel, affiliate commissions paid, and the share of salary spent on it.

A marketer on a $90,000 salary who gives 8 of 40 weekly hours to one channel costs it $18,000 a year, or $1,500 monthly.

Check the grand total against a published median. The SaaS Capital 2026 spending benchmarks, from a survey of more than 1,000 private B2B SaaS companies, put median spending at 8% of annual recurring revenue on marketing and 15% on sales.

At $500,000 of ARR, 8% is $40,000 a year, and channel costs that sum to far less deserve a second look for missing lines such as salaries.

Step 2: Count paying customers and revenue from billing

In Stripe a customer is paying when invoice.paid fires and the subscription status is active, as the subscription webhook guide advises. Refunds and disputes arrive as charge.refunded and charge.dispute.created.

Stripe retries a failing webhook for up to 3 days in live mode, so check the stream against the invoice list each month.

Leave the customer count out of the ad platform's report. Google Ads, for example, credits a conversion for 30 days after a click and 1 day after a view by default, and the click window can stretch to 90 days.

Every platform counts on its own, so 2 platforms can claim the same buyer. In the step 5 example the paid search dashboard reports 55 conversions against 40 paying customers in billing, which moves CAC from $182 to $250.

Step 3: Carry the source from the first click to the invoice

Without a source on the billing record there is nothing to attribute revenue with.

Tag every campaign link with UTM parameters (the free UTM builder keeps the spelling consistent), then store them and any click ID from your own server when the visitor lands.

A script-written cookie is the weak point: Safari's Intelligent Tracking Prevention deletes cookies created in JavaScript after 7 days without a visit and caps them at 24 hours when the landing link carries click IDs as URL parameters, so one can vanish before a 21-day trial converts.

At checkout, write the source onto the Stripe objects. Stripe metadata holds up to 50 key-value pairs per object, with keys up to 40 characters and values up to 500, enough for utm_source, utm_medium, utm_campaign and a click ID.

It does not copy between objects by itself. Pass subscription_data.metadata when you create the Checkout Session and it lands on the Subscription, which Stripe snapshots onto each invoice at parent.subscription_details.metadata. The Stripe metadata attribution guide has the full setup.

Step 4: Choose the revenue window and the attribution model

Set the revenue window to the payback you can afford. Bessemer's target is under 12 months for SMB-focused cloud companies, so a self-serve product at $49/month should read 12 months of revenue per customer.

Label the figure "to date" until every customer in the cohort reaches month 12. Keep the click-to-payment window longer than the median days from first click to first payment in billing.

Google Ads defaults to 30 days, which cuts off a sales-led deal that takes 90.

The model decides which channel gets the customer. Last-touch credits the final click, first-touch the first, and linear splits the payment evenly.

A channel that ranks first under last-touch and last under first-touch is harvesting demand others created, and the attribution model comparison shows how rankings shift. Run all 3 on the same customers.

Google Analytics 4 cannot run that comparison. Its attribution reports offer 3 models: data-driven, paid and organic last click, and Google paid channels last click.

First click, linear, time decay and position-based were removed in November 2023, and every remaining model leaves direct visits without credit unless the whole path was direct. First-touch or linear credit has to be computed outside those reports.

Step 5: Calculate ROI, CAC and payback for each channel

Take a hypothetical product at $49/month with an 80% gross margin, so each active customer yields $39.20 of gross profit monthly. One quarter of costs and billing results looks like this.

The blended row looks acceptable: a 10.3% return and a payback inside 12 months. The channel rows disagree. Newsletter sponsorships return 84.9% and repay in 6.5 months, while paid social loses 40.0% and would need 20.0 months.

Paid search shows a ROAS of 1.44, or $14,400 of revenue on $10,000, yet its return on gross profit is 15.2%, because 20% of every dollar goes to delivering the product.

Step 6: Reconcile to billing and set decision rules

Sum the revenue credited to every channel and compare it with all new-customer revenue in billing for the same period. If billing shows $12,000 and channels explain $9,000, the other $3,000 (25%) has no source.

Direct visits and word of mouth fill part of that bucket, and my rule is to fix capture before comparing channels whenever it passes 20%. Set the payback and lifetime value thresholds before looking at results.

Against the 12-month SMB line, newsletter sponsorships (6.5 months) and content (10.0) pass, paid search at 10.4 is close, and paid social misses by 8 months, so its budget is the first to cut or rebuild.

Lifetime value needs more than 12 months of data, because it estimates revenue you have not seen. Compute it per channel, as in subscription LTV attribution, before applying the 3-to-1 rule from Skok's SaaS metrics guidelines.

Channels that need a different cost column

Affiliate and referral programs

The cost is the commission, and a recurring commission keeps adding cost for as long as the customer pays. A 20% recurring commission on a $49/month plan is $9.80 monthly, or $117.60 over a 12-month window, per customer.

Refunds reverse the commission along with the revenue, so the 2 numbers should move together; affiliate refund clawbacks covers the mechanics.

Content and SEO

Customers arrive months after the writing is paid for, so group the 18 customers in the example by the month of their first tracked click and read each cohort after its 12 months.

First-touch credit suits this channel, since an article seldom closes a sale on the same visit.

Newsletter, podcast and video sponsorships

A flat fee buys one placement, so give each placement its own tracked link. A $2,700 sponsorship that produced 12 paying customers has a CAC of $225. Tracking newsletter links to revenue covers the setup.

Why Google Analytics alone cannot close the loop

Google Analytics 4 records browser sessions and whatever events a tag or the Measurement Protocol sends it, while Stripe files payments under customer and invoice IDs.

Closing the loop means carrying the GA4 client ID into Stripe metadata and posting a purchase event from a webhook.

Google says such events should arrive within 48 hours of the browser event, or they may not be processed as expected for conversion attribution.

A first payment after a 14-day trial lands 336 hours after the click and a month-2 renewal about 720, so neither can reliably credit the campaign. The GA4 and Stripe build shows where each piece breaks.

What TrackRev covers and what stays in your sheet

TrackRev is SaaS affiliate software with revenue attribution built in, and that handles the revenue half of the sheet.

It connects to Stripe (with a restricted, read-only key), Paddle Billing, Polar, Lemon Squeezy, Creem and Dodo Payments, and it matches each payment to the tracked click that preceded it by customer email, so there is no checkout code to change.

The Channels page ranks every channel by clicks, conversions and attributed revenue with refunds and chargebacks netted out, and a second table shows all-time revenue per channel, renewals included.

Last-touch is the default, and first-touch or linear is one setting away.

It does not read ad spend. TrackRev has no ad-account connection, so cost, CAC and payback stay in your spreadsheet, divided into the revenue column TrackRev supplies.

Revenue figures need a paid plan from $29/month; the free plan tracks 50 links and 1,000 events per month and hides revenue. The Revenue Attribution page lists the details.

Pick something else if you need ad-account spend pulled in for campaign-level ROAS, or if your billing runs through a processor TrackRev does not connect to.

A team that wants only standalone attribution, with no partner program, may also prefer a dedicated attribution tool.