Attribution for a Small SaaS

Attribution tries to answer a simple-sounding question: which marketing activities lead to signups, customers and revenue? Large companies throw data teams and expensive modeling tools at it. Attribution for a small SaaS needs a different approach, one that gives directionally useful answers with tools you already have and a process a busy team will actually maintain.

This article walks through why attribution is hard, a practical stack of methods that work together, how to store and analyze the data, and a worked example of how combining views can change a budget decision.

Why attribution is hard

B2B buyers rarely follow a straight line. A typical path might include reading a blog post found through search, seeing the founder's posts on LinkedIn a few weeks later, hearing a colleague mention the product in a Slack community, and finally searching the brand name and signing up. Every one of those touches mattered, but most analytics tools will only see the last click.

Privacy changes make this harder every year. Browsers limit cookies, users block trackers, and people move between work and personal devices. A meaningful part of any customer journey is simply invisible to your analytics.

B2B purchases add another layer, because the person who signs up is not always the person who found you. A manager might discover the product and send the link to a team member, who becomes the account owner. Long sales cycles stretch the gap between first touch and purchase to weeks or months, by which time the original source is often lost. And word of mouth, which is a major channel for many B2B products, happens in private conversations no tool can see.

The upshot is that no model is perfect, and chasing perfection is a trap for a small team. The realistic goal is directional insight: enough confidence to decide where to spend the next dollar and the next hour.

A practical attribution stack

Rather than relying on one method, a small team can combine several simple ones. Each sees a different part of the picture, and together they cover each other's blind spots.

UTM tracking

UTM parameters tag the links you control, so your analytics can tell a click from a newsletter apart from a click from a LinkedIn post. The value comes entirely from consistency. If one person writes "linkedin" and another writes "LinkedIn" or "li", your reports fragment into meaningless pieces. Agree on a short naming convention for source, medium and campaign, write it down where everyone can find it, and use a simple link builder or spreadsheet so nobody types parameters by hand.

First-touch attribution

First-touch attribution records the earliest source you know about for a user, usually captured from the landing page and referrer on their first visit and stored until they sign up. It tells you which channels create awareness. Content and organic search often look much better through this lens, because they introduce people who convert much later through some other path.

Last-touch attribution

Last-touch attribution records the source of the visit right before the signup or demo request. It shows which channels close the gap between interest and action. Paid search on branded terms and direct visits tend to dominate here, which is worth remembering when you interpret it.

Self-reported attribution

Asking "How did you hear about us?" during signup or onboarding captures influence that tracking cannot. Free text works better than a dropdown for small teams, because people tell you things you would never have listed, like the name of a specific podcast or community. The answers are messy and need a few minutes of manual grouping each month, but that effort is where many of the most surprising insights come from.

CRM notes from sales

If you have any sales conversations at all, the people having them hear how prospects found you. Make it a habit to record that in the CRM, along with who referred them and which competitor they were comparing. These notes are especially valuable for larger deals where the path was long and involved several people.

Store attribution data where revenue lives

Attribution data is only useful if you can connect it to outcomes. When a user signs up, store their first-touch source, last-touch source, the landing page of their first visit and their self-reported answer on the account record. If your product database and CRM are separate, make sure these fields flow into the CRM so they sit next to deal value and customer status.

This step is where many small teams fall short. They have UTMs in their web analytics and revenue in their billing system, and no reliable way to join the two. Capturing the fields at signup and attaching them to the account is the simplest fix, and it is much easier to do from the start than to reconstruct later.

Analyze by revenue, not signups

Signups are the easiest number to count and the least useful on their own. A channel that produces many free signups who never activate can look like a winner while contributing almost nothing to the business. Compare channels by paying customers, by revenue, by retention and by sales cycle length.

It is common to find that a channel with fewer signups produces better customers. Referrals and community mentions, for instance, often bring in people who already trust the product and stay longer. A broad paid campaign might bring volume but weaker fit. You only see this if you follow each source through to the money.

Be patient with the numbers. In a small company, a single large customer can make a channel look spectacular for a quarter. Look at trends over several months before drawing firm conclusions.

Use multiple views on purpose

The first-touch view shows which channels create awareness. The last-touch view shows which channels convert existing interest. The self-reported view shows influence that tracking misses entirely. Looking at the same set of customers through all three is more informative than any single model.

Disagreements between the views are worth studying rather than smoothing over, because they often carry the most useful information. If last-touch says most customers come from paid search, but self-reported answers keep mentioning a podcast appearance, the likely story is that the podcast creates demand and paid search captures it. That tells you something a single model never could.

Track dark social

Much of B2B word of mouth happens in places you cannot measure: private Slack groups, direct messages, internal team chats, conversations at events. This is often called dark social. It shows up in your analytics as direct traffic or branded search, which credits the wrong thing.

Self-reported attribution and sales conversations are the main ways to see it. When several customers in a month mention the same community or the same person recommending you, that is a channel, even if no tracking tool will ever report it. Some teams also watch for spikes in branded search or direct signups after a founder's post or a podcast episode, which gives a rough signal of off-platform influence.

A simple monthly report

Once a month, build a short summary with one row per channel and columns for signups, new customers, new revenue and the number of self-reported mentions. Build it once from first touch and once from last touch, and put the self-reported counts alongside. That gives you three views of the same customers on one page.

Keep the report small enough that it takes less than an hour to produce. A report that requires heroics gets skipped. The value comes from looking at it every month and noticing trends, not from any single month's numbers.

When to adopt advanced tools

Dedicated multi-touch attribution tools and marketing mix models have their place, but they make sense later than most vendors suggest. Consider them when marketing spend is large enough that small improvements in allocation are worth real money, when you are running several paid channels that need to be compared against each other, and when you have enough data volume for the models to be stable. With a few dozen customers a month, sophisticated models mostly produce confident-looking noise.

Common mistakes

The most common mistake is trusting one attribution model completely, usually last-touch because it is the default in most tools. That systematically undervalues the channels that create demand. Ignoring self-reported data is close behind, often because the answers are messy. Messy is fine; a few minutes of grouping each month is a small price.

Other frequent problems are failing to connect attribution to revenue, so channels are judged on signup counts, and letting UTM conventions drift until reports are unreadable. Finally, many teams overcomplicate early, building elaborate dashboards before they have the basics stored reliably. Get the fields captured first; the analysis can come later.

A worked example (illustrative)

Consider a SaaS company adding about 40 new customers a month. The team believed most of its growth came from Google Ads, because last-click attribution in its analytics credited paid search with a large share of signups. The natural plan was to increase ad spend.

Before doing that, the team added two things. The first was a free-text "How did you hear about us?" field in onboarding. The second was a CRM field storing first-touch source alongside the last-touch source they already had.

After two months, a different picture emerged. Many customers whose last touch was paid search wrote answers like "a friend recommended you" or "saw the founder's LinkedIn posts." They had heard about the product elsewhere, searched for the brand name and clicked the first result, which happened to be an ad. First-touch data showed that a meaningful share of customers had first arrived through blog posts, sometimes months before they signed up.

The team concluded that branded paid search was largely capturing demand created by other channels. They reduced spend on branded terms and invested more in the founder's LinkedIn presence and in content. Paid search spend dropped while new customer volume held steady. None of this required new software, only two extra fields and the discipline to look at them together.

A minimal attribution setup

If you do nothing else, put four things in place. Use UTM parameters on every campaign link, with a shared naming convention. Store first-touch and last-touch source on every account. Add a "How did you hear about us?" field, ideally free text. And once a month, review customers and revenue by first touch, last touch and self-reported source side by side.

That setup takes a few days to build and an hour a month to maintain. Three views of the same customers usually reveal more than any single model, and they give a small team enough confidence to move budget and effort toward what is actually working.