On this page
The short version
Funnel analysis is the practice of following one group of leads through each stage of the buying process and measuring where volume is lost, so a team can find the stage where pipeline leaks. At RevSure, we hand that check to our Digital Marketing Analyst, a Digital Worker our GTM engineers run on top of the context layer, while people on the customer's team approve every change.
In a September working session with our team, the marketing operations lead at an engineering information company looked back at July's new-logo funnel and found a shape that should not exist. The team had 342 new-logo MQLs that month and 457 sales accepted leads. More people had been accepted by sales than had qualified through marketing, so the funnel got wider at its second stage.
The extra sales accepted leads were not a sign of a strong month. Some had been accepted without ever being marked as MQLs, and others had qualified in one month and been accepted in the next. Once the team followed one group of MQLs forward instead of counting each stage by calendar month, the leak showed up a stage lower, between sales accepted and sales qualified.
What is funnel analysis?
Funnel analysis measures how many records move from each stage of a buying process to the next, how long each move takes and where records stop, so a team can locate the stage losing the most pipeline and look for the cause. In a B2B funnel the stages usually run from lead to MQL, sales accepted lead (SAL), sales qualified lead (SQL), pipeline and closed won. Our MQL vs SQL guide defines each stage and the handoff between marketing and sales, and our sales pipeline guide covers the stages after qualification.
Cohort. A fixed group of records that entered the same stage in the same period, such as every lead that became an MQL in the second quarter. Funnel analysis follows the cohort forward and never adds new records to it.
Funnel conversion rate. The share of a cohort that reaches a later stage. It can be measured from the previous stage or from the stage where the cohort started, and the two answer different questions.
Leakage. Records that leave the funnel at a stage because they were rejected, disqualified, closed lost or left to stall. A stage leaks when it loses more than its own history says it should.
Why does a B2B funnel get wider instead of narrower?
A funnel report that shows more records at a later stage than an earlier one is counting two different populations. Three causes account for most of it.
Records that skip a stage. A rep creates a lead and marks it sales accepted directly, or an inbound request routes straight to sales. Those records appear in the SAL count with no MQL behind them. The engineering information company's ops lead asked on the call to "filter out SALs that had not MQL'd" for this reason.
Calendar counting. A stage count by month includes whatever entered that stage during the month, wherever it started. A lead that qualifies in July and is accepted in August adds to July's MQLs and August's SALs. Our MQL vs SQL guide works through the cohort and period methods with numbers.
Re-entry. A lead recycled to nurture that qualifies again can enter a stage twice. Unless the report counts the person once, the later stage grows.
If any stage-to-stage rate in a report is above 100%, the report is mixing populations, and none of its rates can locate a leak.
How do you calculate funnel conversion rates?
Every funnel conversion rate divides a count at a later stage by a count at an earlier stage, with both counts drawn from the same cohort. The choice is which earlier stage sits underneath.
From the previous stage. SQLs divided by SALs, pipeline divided by SQLs, and so on down the funnel. This is the rate most teams ask for, and it shows how well each handoff works. Stage skips distort it, because a record can count at the later stage without ever being in the earlier one.
From the cohort's starting stage. SALs, SQLs and pipeline each divided by the MQLs the cohort began with. Stage skips cannot inflate it, because every record is measured against the same starting count, and the last rate in the chain is the cohort's overall yield.
In the engineering information company's session, our customer success lead put both versions on screen for the same cohort. When the two disagreed, the ops lead read the reason straight away: "So some stage skipped." Use the previous-stage rate to judge a handoff and the cohort-start rate to judge the funnel as a whole.
Worked example: how to find the leaking stage in your own funnel
The method below runs on an export from any CRM. It needs one record per person with the date each stage was reached.
- Pick a cohort old enough to have finished, such as the MQLs created in one quarter, read at least one sales cycle later. A cohort read too early shows open records as leaks.
- Count the cohort at every later stage, counting each person once, and set aside records that reached a stage with no MQL date.
- Calculate both rates for each stage: from the previous stage and from the MQL count.
- Put the previous-stage rates of the four prior quarterly cohorts next to the current ones. Your own history is the benchmark, because it uses your definitions.
- For each stage, work out how many closed-won deals the cohort would have produced if that one stage had converted at its historical rate. The stage with the largest gap is the leak.
- At that stage, split the records that did not move into still open, dropped and moved backward, then read ten of them by name.
Take an illustrative company whose second-quarter cohort holds 1,000 MQLs, read in early October.
The biggest drop by volume is from sales accepted to sales qualified, where 550 records stop, but volume alone points at the hardest stage, which is not always the one that changed. What marks it as the leak is the comparison with history: 34.5% against 41%. Stage rates multiply, so restoring one stage to its old rate lifts closed won by the same factor. At 41%, the stage would convert 1.19 times as many records, and 44 deals become about 52, or 8.3 more from the same 1,000 MQLs.
Closed won has the lowest previous-stage rate in the table, at 22%. Against its own history it is one point behind and costs two deals. A low rate on its own often marks the hardest stage in a funnel rather than a broken one, which is why each stage is compared with its own past.
Step six is where the cause comes from. Records still open at the leaking stage point at speed or capacity in sales. Records dropped with a rejection reason point at lead quality or a definition. Records that moved backward point at a stage being used for something other than what its definition says.
Where did the engineering information company's funnel leak?
In the September session, once SALs without an MQL date were filtered out, the team read its new-logo MQLs as one cohort. The cohort view showed the largest drop by volume at sales accepted, 291 records, and the ops lead recognized it at once: "That's where the big drop off is, is between sales accepted and sales qualified." Looking at both sets of rates a few minutes earlier, he had put it as "So we have two places where we need to improve, it looks like."
The demand generation leader, who runs the company's paid programs, read the same view as a brief for his own work: "hey, we've got a gap between sales accepted, sales qualified, what's going on there". A drop like that is hard to see on a monthly report with more SALs than MQLs, because the SAL count it starts from is already inflated by records that never qualified through marketing.
Why do stage timestamps belong in the check?
Six weeks later, with the new-logo funnel close to the team's own report, the same ops lead found a second shape that could not be a funnel. A view of every lead that had touched an event campaign in the previous 365 days showed far more sales accepted leads than MQLs. In his words: "how could I get 629 SALs out of 52?" Requiring an MQL type on every record did not close the gap.
The answer was in one lead's history, opened on screen. The lead had become an MQL about an hour before it was tied to the event campaign. The ops lead walked through why that happens every time in the company's own process: "It's the upload then that triggers the MQL. Then it MQLs, the MQL is the trigger to push to Salesforce. Now it's in Salesforce. Now we can tie it to the campaign, so like it happens in phases." Because the campaign association always lands after the MQL, event leads appear to enter the funnel at sales accepted. His proposed fix was a one-day grace period on the association.
For any team running funnel analysis, a stage count is only as accurate as the timestamp behind it, and the order of operations between marketing automation and the CRM sets that timestamp. Reading stage dates in sequence for a sample of records each week catches this kind of skip before a quarterly report does.
What should a weekly funnel check cover?
Neither leak needed new data. Both needed the same cohort read the same way every week by someone who noticed when a shape stopped making sense. Five checks cover it.
Our earlier post on real funnel leaks lists where pipeline tends to drain, and our post on automated funnel analysis covers the case for shorter reporting cycles. The checks above are the routine that tells a team which of those leaks it has this quarter.
How does a Digital Marketing Analyst run the weekly check?
The checks repeat every week and touch every stage, which makes them the kind of work we built our Digital Workers for. The Digital Marketing Analyst runs a Funnel Analytics workstream that analyzes conversion, velocity and leakage from lead through booking, refreshed daily, and keeps one job with people: root-causing structural funnel changes. Its weekly workstreams compare cohorts and segments and explain what changed, with recommended actions.
The Digital Marketing Analyst reads from our context layer, built on the Full Funnel Data Graph, where leads, contacts and accounts from CRM and marketing records are matched so stage counts trace back to named people. Both leaks above depended on that. A stage skip or an out-of-order timestamp shows up only when a record's stage dates and campaign associations sit together.
Our GTM engineers run the worker as an embedded GTM pod alongside the customer's team. With Digital Workers, the customer's team sets the strategy and approves exceptions, while our GTM engineers manage the digital worker day to day and are accountable for the completion and quality of the work it delivers. In the funnel check, that leaves people with the calls in the right-hand column of the table: excluding a group of records, changing a stage definition, approving a rule such as a one-day grace period and choosing which leak to work first. A digital worker is a managed AI role that owns a repeatable GTM job end to end, which is what sets it apart from an agent or a bot.
What changes when the funnel narrows the way a funnel should
A monthly funnel report says how many records reached each stage. The engineering information company's July numbers did that and still pointed the team at the wrong stage, because 457 sales accepted leads against 342 MQLs made the handoff into sales look generous. Read as one cohort, compared with its own history and checked record by record for timestamps, the same data put the leak at qualification, where leads sales had accepted were not becoming qualified opportunities. That finding belongs to both teams at once, and a weekly check keeps it in front of them before the quarter closes.
FAQs
What is funnel analysis?
Funnel analysis follows a fixed group of leads, called a cohort, through each stage of the buying process and measures how many move to the next stage, how long they take and where they stop. B2B teams use it to find the stage where pipeline leaks, usually between MQL, sales accepted, sales qualified, pipeline and closed won.
How do you calculate funnel conversion rates?
Divide the number of records that reached a stage by the number at an earlier stage, with both counts taken from the same cohort. Measured from the previous stage, the rate shows how well one handoff works. Measured from the cohort's starting stage, it is protected from stage skips and shows the funnel's overall yield.
What is a good funnel conversion rate for a B2B funnel?
The most useful benchmark is your own history. Compare each stage's rate for the current cohort with the same stage over the previous four cohorts, using the same definitions. Published averages rarely state how stages are defined or counted, so a rate that is low against your own past says more than one that is low against an industry figure.
Why does my funnel show more SALs than MQLs?
The report is counting two different groups of people. Usually some sales accepted leads never became MQLs because reps created them or they routed straight to sales, or some MQLs from earlier months were accepted this month. Counting one cohort of MQLs forward removes the effect.
How often should a B2B team run funnel analysis?
Weekly, on the most recent cohort that has had time to mature, with a fuller quarterly review. A weekly check catches stage skips, out-of-order timestamps and definition changes while they affect a few records, and the quarterly review compares complete cohorts to decide which leak to fix first.
What does RevSure's Digital Marketing Analyst do in funnel analysis?
The Digital Marketing Analyst is a Digital Worker whose Funnel Analytics workstream analyzes conversion, velocity and leakage from lead through booking, refreshed daily, on our context layer. Our GTM engineers manage it day to day. People on the customer's team keep root-causing structural funnel changes and approve any change to definitions or rules.