Funnel conversions in B2B SaaS 2026: why there are no other people’s standards and how to build your own
The words “demo” and “pilot” mean different things to different companies, so market averages for these stages are useless. I show where external reference points work, how to calculate a funnel by cohorts, and which stage to fix first.
The question about conversion rates in B2B SaaS comes in the same form: “it takes us so much from demo to pilot, is this normal?” The honest answer is inconvenient: there is nothing to compare with. But there is a way to build norms that really work - your own.
Total lead-to-deal conversion is a useless number. It doesn't tell you where you're losing money, which is exactly what you need to know.
1. Why there are no other people’s norms
The problem is not that there is little data. The problem is that behind the same words there are different things.
For one company, a “demo” is a thirty-minute call with a presentation; for another, it’s an independent launch of a sandbox. “Pilot” - sometimes a free month, sometimes a paid implementation for a quarter. “Application” - both registration for a trial from a corporate email, and a request from a buyer from a company with five hundred employees.
Hence the result: two companies with the same “conversion from demo to deal 22%” figure can operate in completely different economies. And most of the public averages for this stage are taken from foreign studies, where both the sales structure and the receipt are different.
Therefore, what follows is a division: two stages, where the external reference point is meaningful, and all the rest, where you need to consider yourself.
2. Top of the funnel: there are guidelines here
Landing conversion and lead cost do not depend on how your sales department is structured. They can be compared with the market.
| Indicator | Okay | Median | Weak |
|---|---|---|---|
| Conversion of SaaS landing page into trial or demo | 5,5% | 3,0% | 1,1% |
| CPL on demo-request, B2B SaaS | 2 800 ₽ | 6 000 ₽ | 14 000 ₽ |
The numbers are from landing page conversion benchmarks and CPL benchmarks in B2B. They should be used for exactly one thing: to understand whether your problem is in traffic and landing or further down the funnel. If the landing page conversion rate is near the median, but there are no deals, there is no point in fixing the top of the funnel.
3. Count by cohorts, not by months
The main arithmetic error in B2B. The transaction cycle is longer than a month, so September transactions grew from July orders. Dividing September trades by September bids correlates different people and gives a number that jumps around on its own.
That's right: applications are grouped by month of receipt, and then it looks what happened to this group in a month, two, three. This is the same mechanism as in cohort retention analysis, only the event is not a repeat purchase, but a transition to the next stage.
A side but valuable result is that you find out the real length of the cycle: the month in which the cohort gets most of the deals. It is impossible to plan revenue without this figure.
4. Breakdown by stages: what does failure mean at each stage?
| Transition | What does the drawdown mean? | What to fix |
|---|---|---|
| Application → qualification | The wrong ones come | Targeting, offer, application form |
| Qualification → demo | We take a long time to respond or it’s unclear why we’re meeting | First touch speed, invitation wording |
| Demo → commercial | The demo does not fit into the client's task | Preparing for the meeting, demo script |
| Commercial → pilot | The price does not agree or the scope of work is unclear | Sentence structure, pilot boundaries |
| Pilot → contract | The pilot did not show the result or there is nothing to measure it with | Pilot success criteria recorded before the start |
The last line is the most expensive and most common. The pilot is launched without written criteria for success, and two months later both sides give different answers to the question of whether it worked. One line in the agreements at the start covers this loss entirely.
5. Which stage to fix first?
Instinct tells you to take the biggest drop in percentage. This is a mistake: interest costs different amounts of money at different stages.
You need to count in rubles. You take the number of transactions that are lost on each transition, multiply by the average check - and you get the price of the stage. It often turns out that a neat plus ten percent on the transition “commercial → pilot” gives more than doubling the conversion at the entrance, because only selected and expensive deals reach this stage.
Next, the figure from this calculation goes to unit economics: It shows how much you can pay per application so that the funnel converges with current conversions.
6. SDR and PLG: two approaches people ask about
SDR needed to separate qualification from sales. The meaning is purely economic: the manager’s time is more valuable, and dealing with irrelevant requests is a direct loss for them. But with a small flow, a separate role does not pay off: qualification is done by regulations and a couple of required fields in the form.
PLG reverses the order: a person first uses the product and only then, if necessary, talks to the seller. The metrics become different - instead of conversion from the demo, the key indicator is activation within the product, and attraction is compared not with the receipt of one transaction, but with the accumulated revenue of the cohort.
You can mix two models in one funnel, but you need to count them separately: the average conversion for both means nothing.
7. How to build your standards in an evening
- Upload transactions for 6-12 months with transition dates by stages.
- Group by month the application was received.
- Calculate the conversion rate of each click within the cohort.
- Taking the median across cohorts is your norm.
- Note the variation: if the conversion of one stage jumps by half from cohort to cohort, the process is not standardized, and the norm does not yet make sense.
The fifth point is more important than the rest. A stable spread is a sign that the stage is manageable. Jumping - that the result depends on who exactly conducted the transaction, and it is the process that needs to be fixed, not the conversion.
8. Summary
External benchmarks in B2B SaaS work only at the top of the funnel: landing page conversion is about 3% at the median and CPL is about 6,000 ₽. Everything below the demo is compared exclusively to itself.
You need to count by cohorts, step by step and in rubles. Then the question “is 22% from demo to deal normal?” turns into another, to which there is an answer: “at the stage of the commercial offer we are losing 1.4 million per quarter, and here’s why.”
Related materials: B2B SaaS marketing, lead generation in B2B, cohort analysis, LTV/CAC benchmarks, analysis of the funnel by numbers, checking the landing page using a checklist, landing page structure.
If there is a funnel in CRM, but the norms from it are not considered - write to Telegram or through form, we'll sort it out at your unloading.