Putting a Value on Conversions That Are Not Purchases

A.What an unvalued count actually asserts

A report showing 340 conversions for one channel and 120 for another says the first channel did better. That is only true if the conversions are comparable, and the moment you have more than one conversion type they are not.

The 340 might be newsletter signups. The 120 might be demo requests from companies with budget. Presented as counts, the worse channel wins the report and, if this feeds bidding, wins the budget too.

Leaving the value blank does not abstain from the question. It answers it with "they are all worth one".

B.Where the damage actually lands

Three places, in increasing order of cost.

Reporting. Channel comparisons are wrong in a way that is invisible, because the arithmetic is correct and the denominators are not.

Budget. Somebody reallocates toward the channel producing more of the cheap thing.

Bidding. The worst. An ad platform optimising toward an unvalued conversion will find you enormous numbers of the easiest one to get, which is the same failure described in which events deserve to be conversions. Values are how you tell the machine that not all of them count equally.

A.The formula

The value of a conversion is the rate at which it becomes the next thing, multiplied by the value of that next thing. Apply it repeatedly until you reach money.

A demo request: 22% become opportunities, 30% of those close, average first-year contract $5,200. That is 0.22 × 0.30 × 5,200, so roughly $340.

A newsletter signup: 4% become trials, 18% of trials convert, average $1,700 first year. About $12.

Both numbers are approximations built from rates you already report. That is the point: this does not need a model, it needs three figures and ten minutes.

B.Approximately right beats silent

The standard objection is that the close rate varies by segment and the contract value has a wide distribution, so any single number is wrong.

True, and the alternative is not a better number, it is the value 1, which is wronger by a much larger margin. Being within a factor of two is enough to fix every decision listed above.

Use the median rather than the mean where the distribution is skewed, which for contract values it almost always is. One enterprise deal will otherwise set a demo value that no demo can live up to.

C.Recompute it, or it becomes a lie

A value derived once and left in place decays. Close rates move, contract values move, and the mix of who requests a demo moves with your marketing.

Quarterly is enough for most, and the recomputation takes the same ten minutes. Put it in the same review where you look at the pipeline, since the inputs are already on the table.

A.What becomes possible

Channel comparison on a single axis, so paid search and content can be ranked against each other despite producing different conversion types. Campaign ROI without waiting for deals to close, which for a long sales cycle is the difference between quarterly optimisation and annual.

And bidding that optimises toward the valuable outcome rather than the frequent one, which is where most of the money is. See channel mix optimisation.

B.The two ways it makes things worse

Valuing something you cannot influence. Attaching a value to a step makes it a target, and any step can be inflated without producing more of what comes after it. Doubling pricing page views is a nav change.

Circular values. Deriving the value of a conversion from the behaviour of people who converted, then optimising acquisition toward it, changes the mix of who converts, which changes the rate the value was derived from. Nobody notices because the number is now stale in a direction that flatters the change.

The guard for both is the same: reconcile against actual closed revenue, quarterly. If estimated conversion value is drifting away from real revenue, one of these two is happening.

C.Getting to the real number instead of the estimate

Every value here is a proxy for something you would rather measure directly: what the people who took this action actually went on to pay.

Kissmetrics keeps the conversion and the eventual payments on the same person record, so the value of a demo request can be computed from the demo requesters who later paid rather than from a chain of averaged rates. Then the estimate becomes a check on the measurement instead of a substitute for it.

The honest caveat: that only works once enough of a cohort has had time to close. For a nine-month sales cycle you are estimating for the first three quarters no matter what tooling you have, which is why the formula above is worth doing today.