B2B marketing metrics: connect lead quality to pipeline

Lead volume is up and cost per lead is down. The marketing report looks good. Then the sales director says pipeline is flat and half the leads are not worth a call. Both can be true at the same time, and the gap between them is where many B2B measurement arguments live.

The fix is not a better dashboard. It is agreeing what each funnel stage means, defining every metric precisely, and comparing groups of leads that have had enough time to progress. Then look at the whole route to a qualified opportunity, including sales response and the handoff, before deciding a channel is good or bad. Cost per lead tells you what a lead costs. Cost per qualified opportunity tells you much more about what it is worth.

Agree what each stage means

Many disagreements between marketing and sales numbers start with words. “Lead” means a form fill to one person and a vetted prospect to another. Before comparing anything, write down local definitions for each stage and who decides when a record moves.

Most CRMs come with default stage names. The names matter less than the written criteria behind them. If two people would classify the same record differently, the stage is not defined yet. Your fit criteria should come from your ideal customer profile, not from whatever fields the form happens to collect.

Define every metric precisely

A metric without a definition is an argument waiting to happen. For each one, write down:

• Numerator: what is counted. Qualified opportunities created from this channel’s enquiries.
• Denominator: what it is divided by. Enquiries from this channel in the period.
• Source: which system, which report.
• Period: the date range, and whether it counts by enquiry date or by stage date.
• Exclusions: duplicates, existing customers, spam, internal tests, job applicants.

The common metrics, with their plain formulas:

• Cost per lead (CPL) = channel spend ÷ enquiries
• Qualification rate = qualified leads ÷ enquiries
• Acceptance rate = sales-accepted leads ÷ qualified leads
• Opportunity rate = qualified opportunities ÷ enquiries
• Cost per qualified opportunity = channel spend ÷ qualified opportunities

“By enquiry date or by stage date” is the definition that causes the most confusion. If you count this month’s opportunities against this month’s enquiries, you are comparing different groups of people. Use cohorts instead.

Compare cohorts that have had time to mature

A cohort is a group of enquiries that arrived in the same period, followed forward through the stages. Enquiries from last week have not had time to become opportunities; enquiries from four months ago mostly have. Comparing a fresh cohort with a mature one makes the fresh one look worse, whatever its quality.

Three rules keep cohort comparisons fair:

1. Wait roughly one sales cycle before judging a cohort’s opportunity rate. If deals usually take ten weeks to reach opportunity, a cohort younger than that is incomplete.
2. Compare cohorts of the same age. January’s enquiries measured at ninety days against February’s measured at ninety days.
3. Deduplicate at the level that matters. Three people from one company filling in forms is one account. Decide whether you count people or accounts, and stay consistent.

A worked example: cheaper leads, more expensive pipeline

This example is illustrative. The channels, spend and counts are invented to show the arithmetic.

Two channels each receive £2,000 in the same month. Both cohorts are measured ninety days later, after a typical sales cycle has passed.

Judged on cost per lead, channel A is two and a half times better. Judged on cost per qualified opportunity, channel B is cheaper. A report that shows only the first row of costs would move budget in the wrong direction.

But be careful with the conclusion. This comparison describes a trade-off in one month’s cohorts; it does not prove that channel B is always better. Before moving budget, check:

• Opportunity value. Eight small opportunities may be worth less than five large ones.
• Closed outcomes. Opportunities are not revenue. Follow the cohort to won and lost.
• Sample size. Five against eight is a small difference in small numbers. Look at several months before acting on it.
• Capacity. Channel A’s volume may be useful if sales has spare time, or harmful if it buries the good leads.

Before blaming the channel, check the handoff

A channel can look poor because of what happens after the lead arrives. Before you cut spend, check:

• Response time. How long until a person responds, and does it differ by channel or by time of day?
• Routing. Do leads reach the right person, or sit in a shared inbox?
• Message and offer fit. Does the ad promise something the follow-up does not deliver?
• Sales capacity and incentives. Are some leads deprioritised because of how sales is measured?
• CRM hygiene. Missing source fields, stages skipped, records merged incorrectly.

Each of these can turn good leads into “bad leads” in the report. Fixing a handoff problem is often cheaper than finding a new channel.

Why marketing and sales numbers disagree

When the two teams report different numbers, the cause is usually one of four things: different stage definitions, different time windows (enquiry date against close date), different attribution rules (first touch, last touch or the CRM’s source field), or duplicates counted differently. Put the definitions side by side before arguing about performance. Much of the disagreement often disappears once both teams use the same cohort view.

Use AI to generate explanations, not a verdict

AI is useful here in a specific way. Give it the cohort table, the stage definitions and a description of recent changes, and ask for competing explanations:

Keep the counts auditable: exported from the CRM with the filters written down, not typed from memory. Check any calculation the model makes. And treat its output as a list of questions to investigate, not a diagnosis. Correlation in a funnel table is not a cause, and a model cannot recover attribution data that was never recorded or infer buying intent from a score. For a regular, structured view of these numbers, see how to write a weekly marketing report with AI. If the question is whether a lead score helps, how to evaluate AI lead scoring covers it.

Attribution is not causation, and pipeline is not revenue

Two distinctions keep reports honest. Attribution assigns credit for a conversion among the touchpoints that preceded it; it does not tell you what would have happened without a channel. Finding that out needs a comparison, such as a holdout or a structured test (marketing experiments covers the design). And pipeline is an estimate of future revenue, not revenue. Report both, label them clearly and follow cohorts through to closed outcomes.

Funnel diagnostic worksheet

Copy this into a spreadsheet. One row per stage transition, per channel, for cohorts of the same age.

The stage with the largest unexplained drop is the place to investigate first. Name the most plausible explanation, the evidence that would test it and the one change you would make, and resist the temptation to fix everything at once.