独家优惠奖金 100% 高达 1 BTC + 180 免费旋转

Why Most Conversion Rates Are Wrong and How You Can Do Better

It is sad to watch competent people making ill-informed decisions because they think conversion rates are simple and straightforward.

Conversion rates are some of the most important metrics. They help predict future productivity, identify bottlenecks, and calculate ROI. You have to get them right.

“What is our conversion rate this quarter?” is a deceptively simple question. You just take number of things that “converted” and divide it by a total number of things, right?

A typical conversation goes like this:
- Executive: What is our conversion rate this quarter?
- Marketing Manager: 2.35%
- Executive: How do you calculate that? (A probing question to verify that the respondent knows what they are doing)
- Marketing Manager: We had 35213 leads this quarter and 827 of them have converted. (Exact numbers indicate that the respondent is competent and prepared)
- Executive: Thanks. Great job!

If you are this executive, take an extra step and ask for the lists of 827 and 35213 things and a short explanation for how something ends up on these lists. Chances are you will be surprised. You may discover that your “actionable analytics” looks like this:

Calculating conversion rates correctly is much harder than seems on the surface. This is why many people only do the surface. They make calculations that look like real conversion rates and result in a believable number, but are plain wrong.

The remainder of this post goes over the most common mistakes (or omissions) people make when calculating conversion rates. Eliminate them to make your conversion rates great again.

Often things convert into other things. Leads convert to Opportunities, and Website Visitors convert to Form Submissions. A common mistake people make is calculating conversion rates as <number of opportunities> / <number of leads>. This leads to subtle mistakes that significantly change the result. The correct formula is <number of leads converted to opportunities> / <number of leads>.

You cannot compare apples to oranges and you cannot divide apples by oranges. You can only divide a subset of apples by apples.

But aren’t “leads that convert to opportunities” the same as “opportunities”? Sometimes they are, but you cannot rely on this assumption. Even if it works today, it may stop working anytime someone changes the process slightly.

Make sure that you always divide apples by apples to avoid errors you would not even think of.

When calculating <converted things> / <things>, make sure that <converted things> are from the same bucket as <things>. Specifically, each converted thing should be present in the list of things.

For example, if you calculate <leads converted to opportunities from Oct 13 to Oct 20> / <leads created from Oct 13 to Oct 20>, you will get something that looks like a conversion rate, but is not. Here, leads that converted are not a subset of leads created. You used different buckets for numerator and denominator. The correct formula would is <leads created from Oct 13 to Oct 20 and converted to opportunities> / <leads created from Oct 13 to Oct 20>.

A good way to avoid this mistake is to list all items in the numerator and denominator and check that all items in the numerator list are also present in denominator list. If not — you are doing something wrong.

Sometimes things convert more than once. Same website visitor can submit multiple forms multiple times.

Due to complexity of tracking such cases, people often take into account only one conversion, usually the latest. This leads to a non-intuitive problem: fake conversion growth.

Imagine 10 out of 100 leads converted to Opportunity in October. Then, 10 out of 100 leads converted to Opportunity in November. Out of 10 November leads, 2 previously converted in October as well. If you take into account only the latest conversion, you will see 8% conversions in October and 10% in November, which make it look like November was a better month (a wrong conclusion).

If your items can convert multiple times, you have a choice of two options:

a) Make sure you record every conversion event.

b) Don’t compare conversion between different time intervals.

But even option (a) has a problem. Returning to our October-November example above, if you add up conversions between 2 months, you’ll have 20 leads converted. But in fact only 18 leads did (two of them converted twice, remember?). There is no right answer — you will have to choose between a few bad ones. Because conversion rates are hard.

If you think previous problems make calculating conversion rates hard, check out this one. It was just a warm-up so far.

In a simplistic model of the world, conversion is something that either happens or does not happen. In reality, there are two more potential outcomes:

If your conversion can take a long time — say, a year to close an enterprise deal — you won’t know the actual conversion rate for a year. This is often unacceptable — you can’t wait a year to make business decisions. You have to make some assumptions, each having its pros and cons.

Here are your options for dealing with delayed conversions:

Option 1: Calculate what converted so far.

The formula is <things that converted so far> / <total things>.

This is the simplest option, most common, and the most incorrect. Don’t ever use it. Choosing this option will make your conversion rates change (grow) over time, and will make different conversion rates incomparable (previous quarter will have an unfair advantage against the current quarter).

Option 2: Select Conversion Horizont(s)

Add time interval to the definition of conversion rate. For example, with the conversion horizon of 3 months, the formula is:

<number of things converted within 3 month of their creation> / <number of things created more than 3 months ago>.

This will make your conversion rates smaller than they actually are, and introduce the delay equal to the duration of the horizon, but will allow you to compare different time intervals and cohorts.

It is often a good idea to use multiple conversion horizons — shorter ones providing faster feedback, while longer ones providing more accuracy.

Option 3: Project Future Conversion

Since the root cause of the problem are things that did not convert yet, you can fix the problem by predicting whether each remaining thing is going to convert or not. You can use approximation of previous conversion rates based on thing’s parameters (e.g. age) or even apply a machine learning model. This could give you more accurate conversion rate, but only if your predictions are good.

Conversion rates are harder and less straightforward than most people think. They are also extremely important for making business decisions. It is worth your time and effort to ensure that you calculate conversion rates correctly.

The simplest way to validate your conversion rates is to get a list of things in the numerator, list of things in denominator, and compare them using your best judgement.

As soon as you calculate your conversion rates, your next question is “are my conversion rates good?”. There is a high demand for this answer across all industries and business functions.

After reading this article, you realize that calculating conversion rates is hard, there are many ways to do it, and if you make even a small mistake, your calculations become useless. Every company calculates conversion rates differently, uses different units, and have different definitions of conversion.

The reality is that you cannot use numbers from other companies to assess your conversion rates. It’s a very harsh reality to accept, so people don’t accept it. When there is a demand, there is a supply, so there is plenty of “industry benchmarks” for conversion rates of all kinds.

They are bullshit.

There is no way to take conversion rates from one company and compare them with conversion rates of another company in a meaningful way, unless companies do exactly the same thing with exactly the same inputs and exactly the same outputs, which almost never happens. The only reason “industry benchmarks” exist is that people desperately want them.

Accept the reality. Don’t be one of them.

What can you do instead? It’s better to spend your time and energy to answer a different question: how can I improve my conversion rates? To answer it:

Add a comment

What is your view?

Send

Related posts:

January 2021 Community AMA Recap

There have been many questions following the announcement of the proposed Dune Network/Free TON community merger. In order to address concerns held by the Dune community, and clear any uncertainty…

Rhetoric as Love

What happens when two actants, two bodies, two different souls, two similar anatomical structures interact? Sparks flying in my head, butterflies swarming in my tummy, love created at first sight…