Conversion rate calculator

Clicks or sessions is the traffic figure, and the page asks for it by name rather than calling it "traffic" because the choice decides the answer. Clicks come from the ad platform and count one person twice if they click twice. Sessions come from analytics and end on a timer, so one person can produce several in an afternoon. Unique visitors count a person once. Same conversions, three denominators, three different rates.

Conversions is the count of the action the campaign exists to produce, on one definition, over the same period and the same rows as the traffic above it.

Target conversion rate is optional. It is not a benchmark this page supplies; it is your own trailing rate, or the rate the page held before it changed, and supplying it turns a percentage into a number of conversions.

A rate is an abstraction, and abstractions lose budget arguments.

On 4,800 clicks, a rate of 3 per cent is 144 conversions. Lifting it to 4 per cent is 192 — forty-eight more, from traffic that was already bought and paid for. Whether that is worth a fortnight of work on the page is now a question somebody can answer, because it is stated in conversions rather than in decimal points.

The same line works in the other direction, and this is the version worth taking to a meeting. Forty-eight conversions, at whatever a conversion is worth, is the cost of leaving the page alone — and it is nearly always larger than the saving available from another round of bid adjustments.

It also reframes what improving the rate is worth against buying more traffic. Forty-eight extra conversions from a better page cost whatever the work costs, once, and then keep arriving on every visit afterwards. The same forty-eight bought as additional clicks cost the click price every month, for as long as the campaign runs. The rate compounds and the media does not, which is the whole argument for spending on the page before spending on the auction.

Most people treat conversion rate as a landing-page metric. It is a property of everything between the impression and the outcome, and the page is only the last part of it.

The promise. An ad that overstates what it links to buys clicks from people who leave immediately. The rate falls and nothing about the page changed.

The intent behind the traffic. Somebody who typed the problem into a search box is closer to a decision than somebody who scrolled past a video, so the same page converts at very different rates by source. An account-wide rate averages those and describes none of them.

The ask. A form with nine fields converts worse than one with three, a checkout with a required account converts worse than one without, and both are decisions somebody made for a reason that may no longer apply.

The page itself. Speed on a mid-range phone, a headline that repeats the ad's promise, a price visible without hunting, and the absence of anything that interrupts before the visitor has read a sentence.

What happens after. For a lead-generation business the rate that matters is not the form submission; it is the share that qualifies. Optimising the first without watching the second reliably makes the second worse.

Blended across sources. Branded search and cold prospecting have no business being in one average, and the average will always be dragged toward whichever one has the most traffic.

Blended across devices. A phone rate and a desktop rate usually differ enough that a shift in device mix moves the blended figure with no change in either.

Blended across events. Counting a newsletter signup and a purchase together produces a rate that improves whenever the cheap event grows.

Read too early. At 144 conversions, an honest week-to-week wobble is ordinary noise. Treating it as a result is how a working page gets rewritten. The sample-size calculator says how much traffic a difference of a given size actually needs.

Read as a goal. A rate can be raised by cutting the traffic that converts least, which raises the metric and shrinks the business. The cost per outcome is the arbiter; this rate is the diagnosis.

Work in the order of what is cheapest to move.

Start with the match between the ad and the page. The single most common cause of a poor rate is an ad that promised something the page does not immediately deliver, and the fix costs a headline rather than a redesign.

Then the ask. Remove fields nobody reads, defer the account creation, and require only what the transaction genuinely needs.

Then speed, measured on a mid-range phone on a normal connection rather than on the machine the page was built on.

Then the traffic. Excluding the placements, search terms and audiences that have produced clicks and no conversions raises the rate without touching the page at all, and it lowers the cost of an outcome at the same time.

Then test, rather than argue. A change worth making is worth sizing first, because a page rewritten on the strength of a fortnight's noise is a page that will be rewritten again.