Conversion Rate as a Misleading Number

A ratio has two halves

Conversion rate is enquiries divided by visitors, and it moves when either changes.

Which means it can improve for a bad reason: a drop in traffic that removes casual visitors leaves a higher proportion of committed ones, and the rate rises while the business gets less.

It can also fall for a good reason: a successful campaign brings a large number of people earlier in their decision, and the rate falls while enquiries rise.

Managing to the rate rather than to the enquiries produces decisions that are exactly backwards in both cases.

The cases where it misleads

• Traffic fell, so the rate rose.
• A campaign brought early-stage visitors.
• A page started ranking for something broad.
• Bot traffic inflated the denominator.
• A seasonal peak brought browsers.

The third is common and confusing. A page that starts ranking for a general term brings a lot of people who were never going to enquire, which depresses the rate while being unambiguously good.

Define what counts first

Since the numerator is frequently as unclear as the denominator.

Does a phone call count, and can you even see it. Does a form submission that turns out to be spam. Does somebody who enquired twice count once or twice.

Two people in the same business will produce different figures because they are counting different things, which makes the number useless for comparison.

Write down what counts, once, and keep it, even if the definition is imperfect.

An imperfect definition applied consistently is more useful than a better one applied differently each time.

And note that most small businesses cannot count calls at all, which means their conversion rate omits the channel that produces most of their work.

Watch the absolute number alongside

The simple correction, and it is enough for most businesses.

Look at enquiries per week as a count, and at the rate, together.

When both rise, something is working. When the count rises and the rate falls, you are reaching more people less precisely, which is usually fine.

When the count falls and the rate rises, that is the dangerous case that looks like success in a report.

The count is the thing the business actually experiences, so it should be the number in front, with the rate beside it as context.

A worked example

A business was told their conversion rate had improved from about two per cent to about three and a half over a quarter.

That was accurate, and it had happened because traffic had fallen by roughly half after a page stopped ranking.

Enquiries had fallen from around forty a month to around twenty-eight.

The rate, read alone, showed a substantial improvement, and the business was doing considerably less work.

Putting the count beside the rate made it obvious in a glance, and the conversation changed from what did we do well to what did we lose.

Segment before comparing

Since a site-wide rate averages things that have nothing to do with each other.

Traffic from a search for a specific service converts very differently from traffic to a blog post, and averaging them produces a number describing neither.

Look at the rate by landing page, and separately for paid and unpaid traffic, which is two reports and considerably more informative.

Mobile and desktop are also worth separating, since a large gap between them is usually a fixable problem rather than an audience difference.

The site-wide figure is mainly useful for noticing that something changed, after which the segments tell you what.

Ignore the benchmarks

Because they are the most common misuse of this number.

Published industry averages combine businesses with completely different traffic mixes, definitions, and measurement, so a comparison against them means very little.

A site whose traffic is mostly people searching for a specific service will show a high rate. One whose traffic is mostly people reading useful articles will show a low one, and may be doing better.

The only comparison worth making is against your own figure over time, with the definition unchanged.

Anybody quoting an industry benchmark at you is comparing your business against an average of businesses you know nothing about.

The number that actually matters

Worth saying, since conversion rate gets more attention than it deserves.

Enquiries that become work, and what that work is worth, is the figure a business runs on.

A site producing fewer, better enquiries can be outperforming one producing many poor ones at every rate.

That means tracking what happens after the enquiry, which happens outside the website and is where most small businesses have no record at all.

A simple list of enquiries with a note of which became jobs is more valuable than any refinement of the rate.

Check what the denominator contains

Since the visitor count itself is less solid than it looks.

Automated traffic, your own visits, and your developer's testing all appear as visitors unless something is filtering them out.

On a small site those can be a meaningful share, and they only ever push the rate down, which produces a business worrying about a number that partly describes itself.

Exclude your own address and any known automated sources, which most analytics tools support and few small sites have configured.

Do it before comparing anything, since changing the filtering later makes the periods either side incomparable.

The counter-case

The rate is still useful.

For comparing two versions of the same page with similar traffic, it is exactly the right measure, since the denominator is comparable.

It is also the correct number when evaluating paid traffic, where you control the denominator and are paying for it.

And it is a reasonable early warning: a sharp move in the rate means something changed, even if it does not say what.

Define what counts, watch the count beside the rate, segment before comparing, and ignore the benchmarks.

What to do

1. Write down what counts as a conversion.
2. Put the enquiry count in front of the rate.
3. Check whether traffic moved before celebrating.
4. Segment by landing page and by source.
5. Separate paid from unpaid.
6. Compare only against yourself.
7. Track which enquiries became work.

Step three catches the case that looks best in a report and is worst for the business, which is a rate that rose because the traffic fell.

Choosing what to track at all is covered in deciding what not to measure.

Frequently asked questions

West Coast Media Solutions Inc. provides web design, web development, hosting, digital marketing, and business consulting to organisations across Canada, drawing on more than twenty-five years in the field.