Cost per Lead, and the Funnel Arithmetic Behind It

The number is a division; the meaning is in the denominator

Cost per lead is spend ÷ leads. Nobody disputes the numerator: it is what you paid, and if you are careful it includes the agency retainer and the creative cost, not only the media. The denominator is where the metric quietly stops being a measurement. A lead is not a natural object. It is a row in a table that somebody decided qualified, and the qualification rule is written by whoever benefits from the number.

That is why CPL, quoted alone, is one of the least informative numbers in a marketing report. It has no units of business value in it. A CPL of $10 is not good and a CPL of $80 is not bad; both are answers to a question that has not been asked yet, which is what those leads turn into. The metric that carries business meaning is at the other end of the chain — cost to acquire a paying customer — and the whole point of this article is the arithmetic that connects the two.

The chain, worked on one campaign

Give two channels $30,000 each and follow the money all the way down. Channel A is paid search on broad commercial terms. It returns 3,000 leads, so its CPL is $30,000 ÷ 3,000 = $10. Of those, 20% pass marketing qualification, giving 600 MQLs. Sales accepts 25% of the MQLs as real opportunities, giving 150. It closes 20% of opportunities, giving 30 customers. The CAC is $30,000 ÷ 30 = $1,000.

Channel B is a placement on an industry review site, where readers arrive already comparing vendors. It returns only 1,500 leads for the same money, so its CPL is $30,000 ÷ 1,500 = $20 — twice as expensive per lead, and on a CPL dashboard it looks twice as bad. But 40% of those leads pass marketing qualification, giving the same 600 MQLs from half the raw volume. Sales accepts 40% of them, giving 240 opportunities, and closes 25%, giving 60 customers. The CAC is $30,000 ÷ 60 = $500. Twice the cost per lead, half the cost per customer.

The identity that makes this inevitable rather than surprising is CAC = CPL ÷ lead-to-customer rate. For A: $10 ÷ 0.0100 = $1,000. For B: $20 ÷ 0.0400 = $500. Channel B's CPL is worse by a factor of 2; its lead-to-customer rate is better by a factor of 4; four beats two, so its CAC comes out half. That is the whole article in one line. It also gives you the breakeven directly: at a $20 CPL, channel B would need only a 2.00% lead-to-customer rate to match channel A's $1,000 CAC, and it is running at double that. From the other side, channel A would have to get its CPL down to $5 to reach B's CAC without improving a single conversion rate.

Notice what the table does that a CPL comparison cannot. Both channels produce exactly 600 MQLs. If your dashboard stops at MQL — and many do, because that is the last stage marketing controls — the two channels look identical, and the $10 CPL channel looks like the efficient one. Everything that separates them happens after the handover to sales, in stages that marketing reports on but does not own. That structural blind spot, not bad maths, is why CPL survives as a headline metric.

The definitional trap, quantified

Now change nothing about the campaigns and change only the word. Suppose the leads above were counted as form fills of any kind — a whitepaper download, a newsletter signup, a pricing enquiry. Redefine a lead as a demo request, which in this model is the marketing-qualified stage. Channel A now reports 600 leads on $30,000, a CPL of $50 instead of $10 — a fivefold increase. Channel B also reports 600 leads on $30,000, a CPL of $50 instead of $20 — a 2.5-fold increase. And the customers, the revenue and the CAC have not moved by one cent: still 30 and 60 customers, still $1,000 and $500.

Read the three pictures side by side, because the comparison is the point. Under the form-fill definition, A looks twice as efficient as B. Under the demo-request definition, the two look identical. Under the only definition that touches money, B is twice as efficient as A. Three answers from one dataset, and two of them are artefacts of a word. Any benchmark that quotes an industry CPL without publishing the lead definition it used is reporting the third of those three numbers and calling it the first.

The operational consequence is unglamorous and unavoidable. Write the lead definition down as a query, not as a sentence — the exact event, the exact filter, the exact exclusion of internal traffic and existing customers — and version it with a date. When the definition changes, and it will, restate the historical series under the new definition before you compare anything. A CPL time series across a definition change is not a trend; it is two different metrics plotted on one axis.

The lag problem: CPL is today, CAC is next quarter

The reason CPL refuses to die is not that anybody believes in it. It is that CPL is knowable this afternoon and CAC is not knowable until the deals close. If leads take a month to reach opportunity and opportunities take three months to close, then the CAC of this month's spend is a fact about November, and by November you will have committed four more months of budget on numbers you could actually see. A metric available now will always beat a better metric available later, unless you build something that stands in for it.

That something is a stage-weighted pipeline value: give every record the expected revenue of the stage it has reached, which is the product of the conversion rates still ahead of it times the average contract value. Take a $12,000 ACV. In channel A a lead has a 1.00% chance of closing, so it is worth $120; an MQL has 25% × 20% = 5.00%, so it is worth $600; an opportunity has 20%, so it is worth $2,400. In channel B a lead is worth 4.00% × $12,000 = $480, an MQL 40% × 25% = 10.00% and so $1,200, an opportunity 25% and so $3,000.

Now do the thing the lag was preventing. On day zero, when all you have are leads, channel A has created 3,000 × $120 = $360,000 of expected pipeline on $30,000 of spend, a ratio of 12.00. Channel B has created 1,500 × $480 = $720,000 on the same spend, a ratio of 24.00. That is the CAC ranking, available on day zero instead of in November, and it is exact rather than approximate: expected value per unit of spend works out to ACV ÷ CAC by construction, which is $12,000 ÷ $1,000 = 12.00 and $12,000 ÷ $500 = 24.00. The weighted metric is not a proxy for the CAC ranking. It is the CAC ranking, expressed in a quantity you can see today.

The catch is honest and worth stating: the weights are your historical conversion rates, so the method assumes the future resembles the past. It reads a channel correctly as long as its stage rates are stable, and it misreads a channel that is changing — a new placement, a new audience, a new sales team. Recompute the weights from a rolling window rather than a fixed one, hold them per channel rather than site-wide, and treat a channel with fewer than a few dozen closed deals behind its weights as unmeasured rather than as good or bad.

What to do with the answer, and where it stops

The blended view of the two campaigns is $60,000 of spend, 4,500 leads and 90 customers: a blended CPL of $13.33 and a blended CAC of $666.67. Shifting the entire $60,000 into channel B, if it held its efficiency, would give 3,000 leads and 120 customers at a $500 CAC — thirty more customers for the same money. That is the size of the prize, and it is also where the arithmetic hands off to judgement, because the assumption inside it is the fragile one: channel B is a placement on a finite site with a finite audience, and a channel that returns $500 CAC at $30,000 rarely returns $500 CAC at $60,000. Spend beyond the natural depth of an audience buys progressively worse leads, and the stage rates that made B look good are the first thing to slip.

Two more limits worth naming. First, the CAC computed here is a paid CAC — media spend over customers won — and it excludes the salaries, tooling and sales time that a full CAC contains; a companion piece in this series works through CAC payback and why the fully loaded figure is the one that decides whether growth funds itself. Second, CAC without a lifetime value beside it ranks channels but does not tell you whether any of them is worth running. A $500 CAC is excellent against a $12,000 contract and ruinous against a $300 one, and nothing in the funnel arithmetic knows which you are selling.

One last piece of sensitivity worth carrying, because it changes where you spend your effort. Channel A closing 25% of opportunities instead of 20% would produce 37.5 customers rather than 30 and drop its CAC from $1,000 to $800 — a 20% improvement from five points at the last stage, with no change to CPL at all. Late-stage rates are levered hardest because everything upstream has already been paid for. If you are choosing between negotiating a lower CPL and fixing a leak between opportunity and close, the arithmetic almost always points at the leak.

Frequently asked questions

How do I calculate cost per lead?

Divide total spend by the number of leads that spend produced. $30,000 across a campaign that returned 3,000 leads is a $10 CPL. Two details decide whether the answer is useful. Include everything you paid, not only the media — agency fees, creative production, the tool subscriptions attributable to that campaign — because omitting them flatters the channel that carries the most overhead. And attribute the leads to the spend that produced them over the same window, which means a campaign whose leads arrive over six weeks cannot be divided by one week of spend.

Can a channel with a higher CPL really be cheaper overall?

Routinely, and the condition is exact. CAC = CPL ÷ lead-to-customer rate, so a channel survives a higher CPL whenever its lead-to-customer rate is better by a larger factor. In the worked example, channel B's CPL is worse by 2× while its lead-to-customer rate is better by 4×, so its CAC comes out at half. Turned into a decision rule: before rejecting an expensive channel, compute the lead-to-customer rate it would need to match your current CAC. If that rate is well below what the channel already does, the CPL is a distraction.

What is a good cost per lead in my industry?

The question cannot be answered across companies, and the reason is not that the data is scarce. Two firms in the same industry can differ tenfold on CPL purely because one counts newsletter signups and the other counts demo requests — the worked example shows a fivefold move from that redefinition alone, with the CAC unchanged. Published industry CPL figures are distributions with very long right tails and no shared denominator, so their averages describe nothing you can act on. Benchmark yourself against your own CAC target instead: divide the CAC you can afford by your lead-to-customer rate, and that is the CPL you can afford.

Should I optimise the top of the funnel or the bottom?

Usually the bottom, and the arithmetic says why. Every stage multiplies, but late stages act on records you have already paid for, so a point gained there is pure margin. In the worked example, moving channel A's close rate from 20% to 25% — five points at the very last stage — takes it from 30 customers to 37.5 and cuts its CAC from $1,000 to $800, a 20% improvement with no change to CPL at all. Getting the same 20% from the top would mean negotiating the CPL from $10 to $8, which is usually harder and always temporary.

How do I judge a channel before the deals close?

Weight each record by the conversion still ahead of it and multiply by your average contract value. At a $12,000 ACV, a channel A lead is worth 1.00% × $12,000 = $120 and a channel B lead is worth 4.00% × $12,000 = $480. Sum across the leads a campaign produced and divide by its spend: A returns $360,000 on $30,000, a ratio of 12.00; B returns $720,000 on the same, a ratio of 24.00. That ratio is exactly ACV ÷ CAC, so it reproduces the eventual CAC ranking on day zero. The weights are your own historical rates, so recompute them on a rolling window and per channel, and distrust them for a channel with too few closed deals behind it.

Does the CAC in this article include salaries?

No. Everything above uses a paid CAC — campaign spend divided by customers won — which is the right measure for comparing two channels against each other, because the salaries and tooling are shared and would cancel. It is the wrong measure for asking whether acquisition is affordable at all. A fully loaded CAC adds the sales and marketing payroll, the tooling and the overhead attributable to acquisition, and it is typically several times the paid figure. Use the paid CAC to rank channels; use the fully loaded CAC, against lifetime value and payback period, to decide how much to spend in total.