I looked at a prospect’s funnel a few weeks ago and the first number they handed me was a trial-to-paid rate of about 10%. They offered it the way you offer a number you aren’t worried about. The industry average for a free trial sits around 18%, so 10% reads as “room to improve” — an uncomfortable-but-survivable place to be.
It wasn’t 10% of the same thing. Their trial required a credit card and ran on a clock. Card-gated trials convert at roughly 49% on average, nearly three times an opt-in trial, because the gate has already removed everyone who wasn’t most of the way to buying before they ever opened the product. Read against the gate that produced it, 10% isn’t slightly below average. It’s bad.
Then we looked one step up. About 350 visitors a month were producing 20 to 25 signups. That was the number actually costing them money — and the same gate that flattered the trial rate was hiding it.
A benchmark is a number attached to a population
Here is the set most people are comparing themselves against:
- Free trial, no credit card: 18% average, 25%+ best in class
- Freemium: 3% average (2–5%; B2B SMB can reach 6–10%)
- Reverse trial: 5% average, 15–30% best in class
- Free trial, credit card required: 49% average
These are not four scores on the same test. They are four different populations wearing the same word. A freemium company at 4% is beating its benchmark. A card-gated trial at 30% is losing badly against its own. Both would look identical on a slide that just says “conversion rate.”
The mistake isn’t using benchmarks. It’s using one without the gate that produced it.
The gate flatters what’s below it and hides what’s above it
A gate is a filter that runs before your number does. Everything measured downstream of it improves. Everything measured upstream of it gets worse. That’s not a subtlety — it’s the entire mechanism, and it works in both directions at once.
A credit card requirement moves qualification upstream. Nearly 3× the conversion rate, and a fraction of the funnel volume. Nobody argues with the first half. What gets missed is that the same move quietly makes visitor-to-signup the number that decides your business, and that number is now the one nobody is watching.
So read conversion rates in pairs. The rate you’re proud of, and the rate directly above it.
Run the arithmetic on a thousand visitors:
- Freemium: ~9% of visitors sign up → 90 signups × 3% → about 2.7 paying customers.
- Opt-in free trial: ~5% of visitors sign up → 50 signups × 18% → about 9 paying customers.
Freemium’s signup rate is nearly double, and it produces roughly a third of the customers. Neither of those percentages tells you that on its own. The pair does. This is the same logic behind choosing between a free trial and freemium in the first place — the model isn’t a taste question, it’s an arithmetic one.
Where the gate belongs
Once you accept that the gate sets the number, gate placement stops being a philosophical debate and becomes a design decision with a rule of thumb attached.
On the card: under roughly $500 in annual contract value, skip it. The funnel volume you lose costs more than the conversion rate you gain. Above about $10K, it’s worth testing — you’re trading reach for qualification, and at that price point qualification is the scarcer thing. In between is a genuine trade, and the tiebreaker is which of the two numbers you can actually move with the team you have.
On the paywall, there’s a cleaner instrument. Patrick Campbell’s Goldilocks Rule asks one question: what percentage of your free users hit the limit? Under 10% and the fence is too high — nobody is reaching it, so it isn’t doing any work. Over 50% and you have a friction problem — the limit is landing before the value does. The zone is 20–30%, where hitting the wall feels like the natural next step rather than a punishment. That’s the same measurement discipline as the math behind your value gate, pointed at the paywall instead of the pricing page.
One thing I’ve watched play out in client sessions recently, worth saying plainly: at the moment a user hits the gate, they don’t need the tier structure explained to them. They need to want what’s behind it. A blurred result the user just asked for outperforms a well-designed comparison table every time, because the first one creates a feeling and the second one creates a decision task. Teach the ladder later, on a pricing page, to someone who has already decided they want in.
Like this? I send one idea like it most weeks. One PLG idea a week → subscribe
The gate you didn’t build
There’s a third gate in every product, and almost nobody designs it: activation.
Median PLG activation runs 25–35%. Bottom quartile is under 20%. Top quartile clears 60%. Activated users convert at 5–10× unactivated ones, and moving activation ten percentage points has been worth a 20–30% improvement in 30-day paid conversion.
Time-to-first-value is the lever underneath it. Under an hour for simple tools, under a day for complex ones, and anything past three days is a danger zone regardless of category. Roughly 70% of users abandon an onboarding that runs longer than twenty minutes.
Put those together and an unfinished First Value Moment is a gate you didn’t design and can’t see on any dashboard. It filters harder than a credit card form, and it filters out exactly the wrong people — the ones who would have paid if they had gotten in the door. A card gate at least removes the uninterested. A broken first session removes whoever ran out of patience first, which is not a segment you’d have chosen.
That’s why activation is the number I’d fix before any of the others: it’s the one gate where tightening the funnel and widening it are the same move.
Write the gate next to the number
The practical version of all of this fits in three columns.
For every conversion rate you track, write down the gate that produced it and the rate immediately above it. Then compare only to a benchmark with the same gate. A card-gated trial gets measured against 49%, not 18%. A freemium product gets measured against 3%, not against the trial company down the street that raised on a better-sounding number.
And when you go to prove that something you changed worked, compare the number to your own baseline from day one rather than to anybody’s benchmark. That’s the Delta-Only Attribution Model, and it exists for exactly this reason: an industry average can tell you whether you’re unusual. It can never tell you whether you improved.
Most of the funnels I open in an engagement have healthy-looking numbers somewhere in them. The healthy-looking ones are usually where the gate is doing the work, and the problem is one step up, unmeasured.
Running a trial you’re not happy with? Book a call and I’ll walk your funnel live.
