Conversion rate calculator

Conversions per what, exactly

The numerator is usually agreed and the denominator almost never is. Two analysts can report rates that differ by a third from the same week of data, and both be computing correctly, because they divided by different things.

• Sessions is the common default in web analytics. Somebody who visits three times before buying contributes three to the denominator and one to the numerator, which pushes the rate down.
• Users, or unique visitors, divides by people instead. The same buyer counts once, so the rate is higher and answers a different question: what share of the people who came ended up buying.
• Entrances to one page is the right denominator for judging that page. Using site-wide sessions to judge a landing page silently includes everybody who never reached it.
• Paid traffic deserves its own denominator. A site-wide rate blends in returning customers and direct traffic, and will always make paid acquisition look better than it is.

The exact sums this page runs

Three arrangements of one division. The third is the one worth reaching for in a planning meeting, because it converts a conversion target into the traffic that target actually requires.

• Conversion rate = (Conversions ÷ Visitors) × 100. The default. Note that the answer depends entirely on which denominator you picked.
• Conversions to expect = (Conversion rate × Visitors) ÷ 100. Forecasting from traffic you can already see in analytics.
• Visitors needed = (Conversions × 100) ÷ Conversion rate. The traffic target behind a conversion target, and the honest reply to an unrealistic goal.

Define the conversion before you measure it

A rate is only as solid as the event underneath it, and that event is a decision somebody made in a tag manager months ago. Check it before you defend a number built on it.

1. 1Write down the single action you are counting. Mixing purchases, newsletter signups and contact form submissions into one numerator produces a figure nobody can act on.
2. 2Check for double counting. A thank-you page that fires on refresh, or an event on both a button click and the confirmation screen, inflates conversions quietly and permanently.
3. 3Decide how you handle refunds and cancellations, and apply it consistently. Counting a cancelled order as a conversion is defensible for judging a landing page and indefensible for judging a channel.
4. 4Confirm the denominator survives consent. If analytics only records visitors who accepted cookies while conversions are recorded server side, the rate is inflated by however many people declined.
5. 5Choose the attribution basis for the numerator, last click or otherwise, and use the same one for every channel you compare.
6. 6Note the date range and denominator alongside the result. An unlabelled conversion rate cannot be reproduced three months later, and it will be questioned.

Rate movements that are not improvements

Conversion rate is a ratio, so it can be improved by shrinking the denominator. Several of the most common apparent wins are exactly that, and they reduce the number of sales while the dashboard turns green.

• Cutting broad top-of-funnel traffic raises the rate immediately and lowers total conversions. If the rate improved and the sales did not, this is usually why.
• Traffic mix shifts do the same thing invisibly. A week with more branded search and less display will post a better rate with no change to the site at all.
• Tracking losses move it in the other direction. A broken tag, a consent banner change or an ad blocker update can knock a chunk off recorded conversions and look like a genuine decline.
• Seasonality and promotions move both sides. Comparing a sale week against a normal week measures the discount, not the funnel.