SaaS Growth Marketing

Most SaaS teams run tactics against a funnel they never quantified. That’s the whole diagnosis. You can tell within ten minutes of a growth review: ask which input is most constrained right now and watch whether anyone answers with a number.

Growth is a model. Five terms, multiplied. Every tactic, every experiment and every hire attaches to exactly one of them or it doesn’t belong on the roadmap.

The growth equation and why multiplication changes everything

Growth in SaaS is new users, times activation rate, times retention, times expansion, times a loop multiplier. Write it down with real numbers for your business and the priorities become obvious within an afternoon.

The reason it matters that these multiply rather than add: a 20 percent improvement to any single term produces the same result as a 20 percent improvement to any other, but the cost of getting it differs by an order of magnitude. Doubling traffic might cost 400,000 dollars in paid spend. Moving activation from 22 percent to 26 percent might cost one engineer three weeks.

Activation is almost always the cheapest term to move and almost always the last one anyone touches, because it lives in the product and marketing teams don’t feel ownership. That ownership gap is the single most expensive org design problem in SaaS growth.

Build the model in a spreadsheet, not a BI tool. You want to be able to change a cell and watch the forecast move. How to build a SaaS growth model walks through the construction, including how to handle cohorts properly instead of averaging them into mush.

The four loop types and how each one compounds

A loop returns its own output to its own input. That’s the only definition that matters, and it’s why loops produce compounding growth while funnels produce linear growth that stops when spend stops.

Viral loops. A user’s normal use of the product exposes a non user to it. Calendly is the cleanest example in B2B: booking a meeting sends a branded page to someone who doesn’t have an account. Loom works the same way through video shares. The critical number is the viral coefficient, users generated per existing user per cycle, and anything above roughly 0.4 is materially reducing your paid acquisition need.

Content loops. Users or the product generate pages that rank, which attract users, who generate more pages. Canva’s template library does this. So does any product where user generated artefacts are publicly indexable. It’s slower than a viral loop and far more durable, because the pages don’t churn.

Paid loops. Revenue from acquired customers funds more acquisition. This compounds only if payback is short enough that you can recycle cash faster than you burn it. At a 16 month payback, a paid loop isn’t a loop, it’s a financing decision.

Sales loops. Customers generate referrals, references and case studies that shorten the next sales cycle and raise win rate. Underrated and badly instrumented almost everywhere, because nobody tracks which deals cited a reference and closed faster.

Loop type	Time to build	Compounding speed	Breaks when
Viral	2 to 6 months of product work	Fast, cycle measured in days	The shared artefact stops being seen by non users
Content	9 to 18 months	Slow but durable	Pages decay or search behaviour shifts
Paid	Weeks	Only with payback under 12 months	CAC rises or cash runs short
Sales	6 to 12 months	Moderate	Reference customers churn or stop agreeing

Pick one loop and build it properly before starting a second. Teams that attempt three simultaneously finish none, because each loop needs instrumentation, a dedicated owner and at least two quarters of iteration before it turns. Growth loops for SaaS covers the construction mechanics, and 12 SaaS growth loop examples shows real ones with the numbers attached.

PLG mechanics from signup friction to PQL

Product led growth means the product does the selling. It works when one person can reach value alone, in one sitting, without talking to anyone. Everything else in PLG is downstream of that test.

Start with signup friction. Every field you add costs you signups, and the standard trade is between volume and qualification. A three field signup (email, password, company) produces materially more signups than a nine field one; whether those extra signups are worth having depends entirely on whether your activation flow can qualify them for you.

Then activation. Define one event, measured within seven days. The right event is whichever one most sharply separates users still present at day 30 from those who left. Run that analysis in Amplitude or PostHog rather than picking an event because it feels important. Most teams guess, and most guesses are one step too early in the flow.

On format choice: freemium versus free trial is the recurring argument and the numbers favour freemium on raw volume. Per 1,000 site visitors, freemium tends to produce around 90 signups and 5 paying customers, while a standard free trial produces roughly 45 signups and 3.6 customers. Freemium wins on customers. It also costs more to serve, attracts users who’ll never pay, and makes support load unpredictable.

I’d pick freemium when marginal serving cost is near zero and the product has a natural usage ceiling. I’d pick a 14 day trial when the product costs real money to run per account, or when value takes a team to demonstrate. Notion and Figma made the first choice. Plenty of infrastructure products correctly made the second.

The experiment operating system

Experiment velocity is the most misreported metric in growth. Teams count tests run. The number that matters is cumulative validated lift shipped, and you can run 40 tests a year while moving nothing at all.

Win rates for well run B2B SaaS growth teams sit somewhere between 15 and 30 percent. If you’re claiming 60 percent, either your traffic is too low for significance and you’re reading noise, or you’re only testing things you already knew would work, which means you’re shipping, not experimenting.

The discipline that raises the win rate is not statistical. It’s the requirement that every test names the model input it moves and states the prior. Write it in one sentence before the test starts: we believe changing X will move activation from 24 to 28 percent because of Y evidence, and we’ll call it at 3,000 users per arm.

• Every test names one equation term. Tests that move nothing measurable get cancelled at the proposal stage, not after four weeks.
• Every test states a minimum detectable effect before it runs, which forces an honest conversation about whether you have the traffic.
• Every result gets written down in one place, wins and losses both. The losses are worth more, because they stop the same test being rerun by the next hire.
• Anything under 1,000 users per arm is not a test, it’s a decision. Make it, ship it, move on.

B2B traffic volumes are the real constraint here. Most Series A companies cannot run a valid test on a pricing page that gets 900 visits a month. In that situation, run qualitative research and painted door tests instead of pretending you have statistical power. Pretending is how teams ship changes that are net negative and never find out.

The stage map, from pre PMF to Series C

Growth work looks completely different at each stage, and applying Series B tactics pre product market fit is the most common way early teams waste a year.

Pre product market fit, there is no growth marketing. There’s retention analysis and there are conversations with users. If week four retention is falling toward zero, pouring traffic in makes the problem louder and no easier to solve. The ICP template for SaaS is more useful at that stage than any channel guide, because the fastest route to retention is usually selling to a narrower set of people.

Between 1 and 5 million ARR, the failure mode flips. Teams start four channels because a board deck said diversify, and none of them get enough attention to produce a signal. Pick one, run it for two quarters, and only add a second once the first is genuinely repeatable. SaaS performance marketing covers what running one properly looks like on the paid side, and B2B SaaS growth strategy by stage covers the full sequence.

Once sales enters the picture, the handoff between marketing and sales becomes a growth constraint in its own right. Write it down. The sales and marketing SLA template exists because verbal agreements about lead follow up time decay inside a quarter, and a PQL that sits for three days converts like a cold lead.

What the compounding companies did differently

Look at product led growth examples and the pattern is consistent. They built one loop, instrumented it obsessively, and were patient enough to let it turn several times before adding anything.

Dropbox’s referral program is famous because it doubled signups, but the part worth copying is that the reward was product currency rather than cash, which meant the loop cost almost nothing per cycle. Calendly’s loop is structural: using the product requires exposing it to someone else. Canva’s content loop took years before template pages dominated design search, and it now produces acquisition at a marginal cost approaching zero.

The honest tradeoff with loops: they’re slow to build and they can stop working for reasons entirely outside your control. A content loop depends on how a search engine behaves. A viral loop depends on the shared artefact staying visible, and a single product redesign can quietly kill it. Anyone who tells you loops are a permanent moat hasn’t watched one break.

Where to start this quarter

If you have no model, build it. One spreadsheet, five terms, real numbers, segmented by motion. Two days of work and it will reorder your roadmap.

If you have a model, find the constrained term and put your best person on it exclusively for a quarter. Not a portfolio of improvements across five areas. One term, one owner, one quarter.

If the constrained term is activation, which it usually is, start with step level drop off data and remove something. If it’s new users, pick one channel. If it’s retention, stop reading growth content and go talk to twenty churned accounts. B2B SaaS growth goes deeper on the diagnostic sequence, but the order above will not steer you wrong.

The teams that compound aren’t cleverer. They just decided what the model was, picked the constrained input, and refused to work on anything else until it moved.