GTM fundamentals for founders · Episode 03

Conversion rate, decoded: the numbers between the steps

Count how far each lead got, not where it is now.

Jamie Sullivan · Arc 1: Who you sell to ·

Prefer YouTube? Watch it there

The video moves quickly through a lot of numbers. This page takes its lessons one at a time, so you can play one part, then go through the arithmetic at your own pace. Every example here is made up.

Conversion rate, decoded

Conversion rate is the share of people at one step who make it to the next.

10 ÷ 25 = 40%25 people reply and 10 of them turn into real deals. A made-up example.

You’ll hear MQL and SQL too: marketing-qualified and sales-qualified lead. They’re labels for two of those steps. Lots of early teams don’t need them yet.

Why it matters now

When the pipeline feels thin, the reflex is more leads. Conversion tells you where they’re actually leaking. Sometimes it’s not volume at all. It’s one step where almost nobody gets through.

Count “got at least this far”

The episode uses five steps:

  1. Imported
  2. Contacted
  3. Engaged
  4. Replied
  5. Deal

Engaged means opened or clicked, and anyone who replied or became a deal counts as engaged too.

The denominator, the number you divide by, is where it goes wrong. When we first built our own funnel numbers, we counted where each lead was right now. So every lead that had moved on looked like it had dropped out. A lead that has already replied is no longer “contacted” today, but it obviously got through contacted.

Counting “got at least this far” fixed it, and the leak moved to a completely different step.

Where each lead is right now

Imported200
Contacted7070 ÷ 20035%
Engaged4040 ÷ 7057%
Replied2828 ÷ 4070%
Deal1212 ÷ 2843%

Got at least this far

Imported200
Contacted150150 ÷ 20075%
Engaged8080 ÷ 15053%
Replied4040 ÷ 8050%
Deal1212 ÷ 4030%
The same 200 made-up leads, counted two ways. The lowest rate is highlighted, and it’s a different step each time.

Here’s how the right-hand side is built. Each step is everyone who’s there now, plus everyone who went further. Start at the bottom and work up:

  • Deal: 12
  • Replied: 28 + 12 = 40
  • Engaged: 40 + 28 + 12 = 80
  • Contacted: 70 + 40 + 28 + 12 = 150. The other 50 were never contacted.

Counted by where leads are now, it looks like the leak is at contacted, at 35%. Counted properly, the biggest drop is reply to deal, at 30%.

Below about 20, the rate is noise

If a step has three leads in it, I don’t trust the percentage.

2 of 3 isn’t a 67% stepIf one of those leads goes the other way, the same step reads 33%.

It’s “not enough yet”. Say so out loud. Below about 20 leads at a step, or about 5 closed deals, the rate is noise. Those are rules of thumb. I just write “not enough yet” instead of a number.

Worked example: Tallybird

Tallybird is a made-up company with 50 leads. Try each step before you open the answer.

Your turn · made-up example

Tallybird’s 50 leads, by the furthest step each one reached:

Imported only6
Contacted14
Engaged12
Replied10
Deal8
  1. 1. Ten stopped at replied. How many got at least as far as replied?

    Show the answer

    The 10 who stopped there, plus the 8 who went on to a deal.

    10 + 8 = 18

  2. 2. Up one step. How many got at least as far as engaged?

    Show the answer

    The 12 who stopped at engaged, plus the 18 you just worked out.

    12 + 18 = 30

  3. 3. And contacted?

    Show the answer

    The 14 who stopped at contacted, plus the 30.

    14 + 30 = 44

  4. 4. Last one, and it’s a check. Does imported come back to 50?

    Show the answer

    The 6 who were imported and never contacted, plus the 44.

    6 + 44 = 50

  5. 5. Now you have 50, 44, 30, 18 and 8. Divide each by the one above it. What are the step rates?

    Show the answer
    StepGot at least this farStep rate
    Imported50(start)
    Contacted4444 ÷ 50 = 88%
    Engaged3030 ÷ 44 = 68%
    Replied1818 ÷ 30 = 60%
    Deal88 of 18: not enough yet

The wrong read: counted where they stopped, only 10 of 50 are at replied, so 20% reply. That ignores the 8 who went on to a deal. So 18 got at least as far as replied, not 10.

Why the last rate is “not enough yet”: only 18 leads reached the step before. A step rate depends on how many went into it, so the 20-lead floor is the one that decides.

Time lag and cohorts

Leads imported last week haven’t had time to reply. So compare groups that started in the same period, which is a cohort, or give each step a fair window.

One number, or four?

Side by side: the bad version is “our conversion is 2%”, a made-up number. One number, end to end, counted by where leads are today, old and new leads mixed together. The good version is four step rates, counted “got at least this far”, one cohort at a time, with “not enough yet” under about 20.

Bad looks like“Our conversion is 2%.” One number, end to end.

Good works likeFour step rates: imported → contacted → engaged → replied → deal.

Why it mattersOne number can’t tell you which step to fix; four can.

Bad looks likeCounting where each lead is right now.

Good works likeCounting “got at least this far”.

Why it mattersCurrent status makes every lead that moved on look like it dropped out, so the leak shows up at the wrong step.

Bad looks likeLast week’s leads mixed in with last quarter’s.

Good works likeOne cohort at a time, or a fair window per step.

Why it mattersNew leads haven’t had time to reply, so mixing them drags every rate down.

Bad looks like“2 of 3 converted, so it’s a 67% step.”

Good works like“Not enough yet” under about 20 leads at a step.

Why it mattersA confident percentage on three leads misleads everyone who reads it.

Bad looks likeComparing your rate to a number from a blog post.

Good works likeComparing this cohort to your own last cohort, counted the same way.

Why it mattersYour own baseline is counted your way; a borrowed one almost never is.

I don’t measure it against published benchmarks either. They vary too much by source and stage: every vendor defines the steps differently, and most count end to end. Your own last cohort, counted the same way, is the one to beat.

Try it this week · Workbook

Try this week

Your last 50 leads: where each one stopped, who got at least that far, and four step rates.

  1. 01Write down your steps and what counts as reaching each one. Keep the definitions fixed.
  2. 02Take your last 50 leads from one period (one cohort). Mark the furthest step each one reached.
  3. 03For each step, count everyone who got at least that far, including the ones who went further.
  4. 04Work out the four step rates. Under about 20 at a step, write “not enough yet” instead.
  5. 05Repeat for the next cohort, counted the same way, and compare the two.

Your funnel, one cohort

Your answers stay in this browser. Nothing is sent anywhere.

StepStopped hereGot at least this farStep rate
––
––
––
––
––

The biggest drop with enough leads in it is where to look first.

Watch out for

  • Blended rates. One rate across inbound and outbound, or across very different deal sizes, can hide a step that works well for one group and badly for another (episode 8).
  • Time lag. A cohort that started last week will look worse than it is. Give each step a fair window before you judge it.
  • Small samples. Below about 20 at a step, the percentage is noise. Say “not enough yet” out loud.
  • Borrowed benchmarks. Vendor and survey numbers use their own step definitions and their own customers’ data. They rarely match how you count.

Decoded this episode

The glossary

Conversion rate
How many make it to the next step ÷ how many were at this one.
Funnel / steps
The steps a lead goes through on the way to becoming a customer.
Engaged
Opened or clicked. Anyone who replied or became a deal counts too.
MQL / SQL
Marketing-qualified / sales-qualified lead: labels for two of those steps.
Sample size
How many you’re counting. Too few and the percentage is noise.
Cohort
A group that started in the same period, so you compare like with like.

Your turn

Which step leaks most for you right now: getting a reply, or turning a reply into a real deal?

Answer in the YouTube comments