Lead Quality vs Volume
Date: 2026-09-27
Almost every change that raises form fills lowers the share that turn into customers — shorter forms, gated content, lower bids on broader audiences. The only honest measure is cost per customer, which is months away; the useful compromise is the earliest stage that still predicts it.
Lead quality versus volume is the tradeoff between the number of leads a campaign or form produces and the share of them that become paying customers.
Why they move in opposite directions
Every lever on lead volume works by lowering the bar to submit:
| Lever | Volume | Quality — usually |
|---|---|---|
| Fewer form fields | ↑ | ↓ — less commitment, less information to qualify on |
| Content gated behind a form | ↑↑ | ↓↓ — people wanted the PDF, not a call |
| Broader targeting, cheaper bids | ↑ | ↓ |
| Incentives (“free audit”) | ↑ | ↓ — attracts freebie-seekers |
| Qualifying questions (budget, timeline) | ↓ | ↑ |
| Price shown on the page | ↓ | ↑ — the unaffordable filter themselves out |
Not always true — a confusing form loses good leads and bad alike, and fixing it raises both. But the default expectation for any “more leads” change should be lower quality until shown otherwise — Form Design.
Worked example
Two campaigns, judged on the stage that pays.
CPL LEADS COST WON WIN RATE COST PER CUSTOMER
Campaign A £30 1,000 £30,000 10 1.0% £3,000
Campaign B £90 300 £27,000 15 5.0% £1,800
B is three times the CPL and 40% cheaper per customer. On the dashboard marketing sees, A wins by a mile.
The lag problem
Won deals take months. Early on, the numbers can’t separate the campaigns:
AFTER ONE MONTH leads won rate
Campaign A 1,000 4 0.4%
Campaign B 300 3 1.0%
two-proportion z-test:
pooled rate p = 7 ÷ 1,300 = 0.54%
standard error = √(p(1−p)(1/1000 + 1/300)) = 0.48 percentage points
z = (1.0% − 0.4%) ÷ 0.48% = 1.25
p-value ≈ 0.21
In plain terms: with seven customers in total, a gap this size turns up by chance about one time in five. Not evidence of anything yet — Proportion Tests.
Moving up a stage fixes it. SQLs arrive within weeks and there are many more of them:
AFTER ONE MONTH leads SQLs rate
Campaign A 1,000 90 9%
Campaign B 300 60 20%
z ≈ 5.2, p < 0.0001
In plain terms: a difference this large on this many SQLs would almost never happen by chance. The catch is that it’s only as good as the SQL definition — if SQL-to-won differs by campaign, the proxy misleads.
Working with it
- Pick the earliest stage that still predicts revenue, and check it does — Leading and Lagging Indicators, Lead Funnel Stages
- Feed that stage back to the ad platforms so bidding optimises to it rather than to form fills — Offline Conversion Imports
- Score leads on fit and intent — a lead score adds points for firmographic fit (industry, company size, role) and behaviour (pricing page viewed, repeat visits). Treat it as a routing rule, not a truth; check scores against actual win rates every quarter
- Test form changes on a downstream metric. A form test measured on submissions will pick the worse form — Metric Selection for Tests
- Watch speed of follow-up before blaming lead quality. Leads nobody called look like bad leads — Speed to Lead