The reply to meeting audit
You cannot look up how many of your replies should be turning into meetings, because every band published for it was published by somebody who sells meetings. So here is the ninety minutes that produces your own number: one closed month, your inbox, a calendar, a spreadsheet, and a ranked list of the leaks at the end of it.
By Rahul Bageria, co-founder · Updated August 2026 · 14 min read
Pull one month and count it yourself
Seven steps, about ninety minutes, on one mailbox, one calendar and a spreadsheet. The table is the whole run; the searches and the ranking rule follow it.
| Step | Where it lives | Minutes | What you write down |
|---|---|---|---|
| 1. Pick a month that has finished | The sheet | 5 | The month, what counts as a meeting, and that you count people |
| 2. Pull every reply into rows | The mailbox you sent from | 20 to 30 | Name, company, when their reply landed, their first sentence |
| 3. Tag where you lost each one | The sheet | 15 | Human or machine, then yes, no, or a later date |
| 4. Timestamp your answer and the meeting | Sent mail, then the calendar | 20 | Hours to your answer, the meeting date, whether it happened |
| 5. Count the messages to a slot | Three threads that said yes | 10 | Messages to a confirmed time, and what each one asked for |
| 6. Write the chain in counts | The sheet | 10 | Five numbers, then the size of every gap, in people |
| 7. Rank the leaks and close one | The sheet, then your calendar | 10 | The leaks in order, one circled, the date you run this again |
Sources: Google's own Gmail documentation and Microsoft's own Outlook documentation for the searches in steps 2 and 4, and lemlist's own help center for the campaign export and its reply timestamp, all checked August 2026. Every other number here is one you counted yourself.
Block ninety minutes on a Thursday and run the steps in order. You cannot rank leaks you have not counted.
Start with a month that has finished happening
A reply from last Tuesday has not finished happening. Audit the month you are in and every leak reads worse than it truly is.
A closed month is one that ended at least two weeks ago, so the replies inside it have already gone wherever they were going.
- •Held, meaning it happened, with the person who owns the problem
- •Booked and held are two columns, never one
- •Pin it at the top of the sheet, with today's date
- •If you cannot write it in one sentence, stop and write it now
- •One row per person, however many messages they sent
- •Four replies from one person is one row
- •Two people at one company are two rows
- •A machine answer is a row until step 3 pulls it out
Two sentences at the top of the sheet decide what every number underneath them means. Write them before you count anything.
Get every reply into rows, including the ones you never saw
The highest-yield instruction on this page is twelve characters long, and it goes in the search box of the mailbox you sent from.
Take the number off the dashboard
Campaign performance: 27 replies
- ✕Counts whatever that tool decided a reply is
- ✕You cannot open a number, read it, or tag it
- ✕It never saw anything that skipped the tool
Search the mailbox, spam folder included
in:anywhere to:me after:2026/06/01 before:2026/07/01
- ✓Google's own Gmail documentation puts Spam and Trash inside in:anywhere
- ✓On Microsoft 365 the same move is received: with the two dates
- ✓Every row can be opened, read and tagged
The ones you already answered
One row each: name, company, the time their reply landed, their first sentence in their own words. Keep that last column even in a month when you are only counting.
The ones that already cost you a meeting
Answer anything real in here today, however late, and say plainly that it landed in the wrong folder. Treat it as plumbing rather than copy. Find nothing and you have ruled out the cheapest failure on this page in thirty seconds.
The count that will not match
Write both numbers down, use the mailbox, and refuse to reconcile them: they are counting different objects, which the outbound metrics guide takes tool by tool. That attempt is where most first audits die.
Tag each row with the step where you lost it
Two tags per row, and neither is yours to invent. First, human or machine. Then, for the human ones, the place you lost them.
Our reply management guide names the five places this funnel leaks, so borrow its names. Machine rows are out of office notices, bounces and auto-replies, and from now on the machine handles them.
Rows you had filed as noise
Count the replies from a real person that never got an answer. Then answer every one before you close the sheet today.
Rows where you were slow
Leave this one empty for now. Step 4 fills it with hours, and it is the only tag that needs a clock rather than a judgment call.
Rows where you pitched instead
Count the replies where your next message asked for more than they had agreed to. Rewrite one today as the answer that kind of reply was owed.
Rows that agreed and got no time
A yes is somebody you targeted agreeing to a next step of any size, this quarter. Count the yeses with no confirmed slot; step 5 says why they stalled.
Rows with a slot and no meeting
Count the meetings that did not happen. Nothing marks a no-show for you, so mark them by hand now, and getting them back is a job for after this run.
Two moves come out of this step today: answer everything in the classification column, and suppress the machine rows so next month's pull is shorter.
Timestamp your own answer, and the meeting
Two more columns, twenty minutes. Your mailbox already knows the first one. Nothing records the third step for you at all.
-
1
Hours between their reply and yours
Read your own send time against theirs in the thread, or run the search from Sent Items in Outlook. Write hours, not days: days hide the gap between an afternoon and a weekend.
Do this: if your slowest answers are the ones that went dead, put two fixed passes in the day and an alert on anyone who already said yes.
-
2
The meeting date, from the calendar
Not from memory, and not from the thread. Microsoft's own Outlook documentation carries a separate table of calendar searches, which only work when run from a calendar folder. On Google Calendar the month view is enough at this volume.
Do this: if the meetings that fell through were booked furthest out, offer times inside the next four days and stop offering next week.
-
3
Mark the ones that never happened
No tool marks a no-show for you. Go down the meeting column by hand and mark each one held or not held, with the problem owner in the room.
Do this: a meeting rescheduled once and never rebooked is not held. Put it with the no-shows and stop arguing.
Count the messages between a yes and a slot
Nobody counts this one, and at a founder's volume it is usually the biggest leak your inbox owns. Open three threads where somebody said yes.
Make them do the work
Great, happy to chat. Grab any slot that suits you here: [link]. Quick one first, how many reps are you running?
- ✕Three messages before a time exists
- ✕A qualifying round nobody agreed to
- ✕A link with nothing free for nine days
Put two times in the reply
Thursday 14:00 or Friday 10:30 work my end. Say either and I'll send the invite. Link below if neither fits.
- ✓One message between the yes and a time
- ✓The link is the fallback, not the ask
- ✓Nothing to fill in, nothing to qualify
Write the count for all three threads, and what each message in between asked them to do. One is the number you want. Two is survivable. Three or more is your leak, and it is the cheapest of the five to close: delete a step this week, starting with the form or the qualifying question, whichever you added last.
Want the replies you already earn worked properly, by the two of us?
Book a Fit CheckWrite the chain in counts, not percentages
Five numbers on one line, then the gaps between them in plain words. Do the arithmetic once, in people.
At twenty replies a month a percentage swings on one person, which is why reading your market from replies wants volume before a rate means anything.
Everyone who answered, spam included
If this differs from your dashboard, one of them is wrong. Use this one and leave the dashboard shut.
Humans only, machines taken out
The gap to the first number is your machine mail. Suppress it in the tool rather than reading it again next month.
Agreed to a next step of any size
This gap is usually the widest on the line by a distance. If it is yours, read the boundary section below before you touch your inbox.
A confirmed time, not a link you sent
Everything lost between the yes and here is step 5's problem, and a week of work to fix rather than a quarter.
It happened, with the right person
The only number on the line that pays for anything. The rest are promises somebody made you.
Now write every gap as a sentence about people. Eight said yes, five got a slot, four happened, so you lost three between the yes and the calendar. Circle the biggest: it is the only thing you carry into step 7.
Read the gaps, not the rate. A count survives being recounted; a percentage invites a comparison with a band somebody published to sell meetings.
Rank the leaks and close exactly one
Order decides the month. Close the wrong leak first and the total does not move, however good the work was.
By people lost, not by how bad it felt
Write the five tags in order of how many people each cost you this month. Latency is the leak everybody confesses to, and the count often clears it.
One leak, one change, written in a sentence
Leave the other four alone until the next run. Change three things and next month tells you nothing about any of them.
Only the tightest step changes the total
That is the argument at the center of The Goal, the 1984 book that made "the bottleneck" a business word: effort spent anywhere but the tightest step does not raise output, it feels like progress.
Date it before you close the sheet
Same columns, same two sentences, next month, in half the time. Change what you count and you have thrown away the only comparison you own.
When two leaks cost the same number of people, close the one that needs a routine rather than a rewrite. Latency is the cheaper of almost any pair: a calendar block and an alert, not a copy project.
One month, run end to end
An illustrative walkthrough of the method, not a specific client result. We report real numbers only when they are real.
A solo founder ran one campaign in June: Google Workspace, one sending tool, no CRM, a booking link in her signature. She audits it on the first Thursday of August.
-
1Steps 1 and 2 · 30 minutes
31 rows
Two sat in Spam and she had never seen either. Her tool reported 27 for the same month, so she wrote both down and used 31.
-
2Step 3 · 15 minutes
22 real replies, 8 yeses
Nine rows were machines: six out of office notices, three bounces. Of the 22 left, six said stop, four gave her a date in Q3, four said not for us.
-
3Step 4 · 20 minutes
5 got a slot, 4 held
Most answers landed inside six hours, and her slowest took four days and still booked. The meeting that fell through was booked eleven days out.
-
4Step 5 · 10 minutes
Four, three and two messages
Both threads that died had a form in them, and one carried a qualifying question nobody had agreed to.
-
5Steps 6 and 7 · 20 minutes
31, 22, 8, 5, 4
The gaps, in people: 14 between a real reply and a yes, 3 between a yes and a slot, 1 between a slot and a meeting. Inside the inbox: booking 2, Spam 1, the show 1, her speed nothing.
So she circles the booking step, deletes the form, and starts replying with two named times and the link underneath. One change, written in a sentence, re-run booked for September. Both people she lost there had already said yes, so this recovers meetings she earned.
Know when it is not an inbox problem
The biggest gap in the chain is usually the one this page cannot fix, and an audit that never says so costs you a month of the wrong work.
"Your replies are not turning into meetings, so tighten the handling: answer faster, template the answers, put the booking link everywhere."
- ✕Assumes the leak sits after the yes
- ✕Rebuilds a routine when the message was at fault
- ✕Moves a number that was never the problem
"Most of the people who answered never agreed to anything. No answering routine touches that gap."
- ✓That gap is the offer and the ask, not the inbox
- ✓Fix what you asked for before how fast you answered
- ✓The three inbox leaks are still worth an afternoon, second
When that gap is your biggest, close the sheet and go and work on the ask. What actually earns a reply is a different job from what happens once one arrives, and it pays for the rest.
One boundary worth naming: setting a meetings target and working backwards to the contacts it needs is planning, done before you send. This reads the same chain backwards from a month that happened.
I have watched founders rebuild a whole reply routine off a number whose biggest gap sat further upstream than the inbox reaches. The tell is the same every time: answers got faster, templates got tighter, the calendar did not move. Count first, then pick the job.
Where the audit goes wrong
Four ways to finish a run holding a number nobody should act on, and one that destroys the comparison you are building.
Half of those replies have not finished happening, so every leak reads worse than it is and next month reads better than it was. Wait two weeks.
The dashboard and the mailbox count different objects, so they will never match. Write both, use the mailbox, move on.
Warmth is not a tag. What they committed to is. A friendly maybe and a flat yes belong in different columns.
Everybody confesses to answering slowly, so latency gets fixed first and the total does not move. Fix whatever your own count put at the top.
Changing what you count between two runs
There is no honest outside number to measure yourself against here, so your own last month is the entire basis for judging anything you change. Redefine a column mid-quarter and you hold two unrelated measurements and no way to tell whether the work paid.
What to do on Thursday
The audit's output is a decision, not a dashboard. Finish with five charts and you have not finished.
- 1 Audit a month that ended two weeks ago, never the one you are in.
- 2 Pull the replies from the mailbox with the spam folder in, one row per person.
- 3 Write the chain in counts and name every gap in people, not in percent.
- 4 Close one leak, write the change in a sentence, and date the next run.
On a first month of sending, read the counts as a starting point rather than a verdict, because outbound takes longer than that to show its shape. When two runs in a row put classification and latency at the top, the fix is a routine, and reply automation is the version of it we run for teams.
Questions founders ask
What is a good reply to meeting conversion rate?
How do I calculate reply to meeting conversion?
Which month should I audit?
Can I run the audit without a CRM?
My sending tool and my inbox report different numbers. Which do I use?
What if replies are landing in spam?
Co-founder of Real Good GTM. He has been the first business hire and Chief of Staff at seed-stage B2B startups, building outbound pipeline before any playbook existed. This is the count he runs on a shared inbox whenever a founder says the replies are fine but the calendar is empty.
Connect on LinkedInOnce you know which leak you have
The guide behind the five names, the rule for a yes, and what the same month cost.
Reply management
The five leaks explained, and the four decisions behind every reply.
Read the guideWhat counts as a yes
The three gates a positive reply clears, and eight hard cases worked.
Read the postCost per meeting
The money side: what those held meetings cost once your hours are in.
Read the postWant your last closed month counted properly?
Book a fit check. We'll run this count on your last closed month, tell you which leak is costing you the most people, and say straight if the biggest gap is your offer rather than your inbox.
Book a Fit CheckNo hard sell. No fake numbers. Real good work speaks for itself.