Conversion Optimization
What is conversion optimization
Conversion optimization is the idea of optimizing individual conversion points against how they are expected to perform. This better allows teams to identify potential opportunities and weaknesses their business units have.
The why and how
To accomplish this, we develop models to compare the expected performance to actual performance. We will explain how to understand different likelihoods of actual results compared to expected results, specifically in an example where a sales team wants to increase net conversions and how “low-quality leads” can help hit that goal.
Understanding the situation
Commonly, teams will look at the performance of different areas of their business unit and choose to improve them arbitrarily. The two most common arguments are “this metric is low, which means there is a lot of potential here” or “this metric is high, and we should try to push it higher.” These ideas are not inherently incorrect, but when they are followed without supporting data, they can lead to a waste of time and effort. To solve this, we want to model out our various metrics. Instead, what we want to do is review what each group’s conversion rate should be to then compare that to our actual performance.
A quick refresher on clustering
When examining any database of entities, it can be useful to group them in different ways. One common way is to group them based on the likelihood of converting and then learn which attributes tend to affect conversion rates. In machine learning, these clusters can be used for more advanced purposes. What we are doing is one the simplest ways to use clusters, but by getting into that mindset, you can begin setting up your business to be ready for more AI tools in the future. In this example, a cluster is similar to how most organizations would “rate leads.”
Let’s start with a simple example using dice and coins.
How likely will X or more conversions occur in a given group? To better understand this, think of dice. If you were to roll a single die, there is a one in six chance of getting a 6; you are unlikely to get it. However, if you roll 20 die, the odds are pretty good you will get at least one 6, about 97%. Now, if you want to get at least two 6s, the odds drop a little to 87%, and so on; as you increase the desired number of conversions, the chances of hitting that “goal” drop.
Now, think of a coin. The odds of getting heads on a coin are higher than that of getting a 6 on a die, and if you flip it 20 times, you are all but guaranteed to get at least one heads. As with the die, the more heads you want to get out of those 20 flips, the lower the odds of achieving the “goal.” If you were looking for five heads and five 6s, the odds would always be greater to get five heads; however, that does not mean we want to ignore the die and focus all our energy on the coin. What if you wanted 10 “successes” after rolling and flipping 20 times? Using some basic stats, you can find the optimal mix of both groups to reach ten successes; in this case, it would be two 6s and eight heads. This means it would be harder to get the ten successes by only flipping coins.
This is all to say that when you are setting goals and later trying to achieve them, it is important to know how your different clusters are expected to behave compared to how they are behaving. Just because one group converts more does not mean it converts better.
For our example, we will look at a theoretical business that expects to get 60 leads for its sales team in a given time period. These leads are then filtered into one of three clusters based on the historical likelihood of converting. We have three groups of 20 leads each. Leads in Group 1, red, have an average of 25% converting; in Group 2, yellow, an average of 50%; and in Group 3, green, 75%.
First, let’s look at two groups in isolation—the green and red groups, which have a 75% and 25% chance of converting any random lead, respectively. Although it may seem like the best strategy to convert all “green” leads before moving to another group, that is not the case. As you can see, although you are more likely to convert 11 leads from the green group than even a single lead from the red group, once you get to that point, each additional lead you try to convert becomes harder to close. We want to develop a matrix for optimizing the number of leads from each group to maximize the TOTAL number of leads converting.
This is also what I was referencing earlier with converting more versus better. If a sales team closes ten leads from the green group and five from the red group, they do a better job closing from the red group and should focus more on the green.
Note that additional math is required to merge what we will assume to be independent events. This article does not cover the math used to solve for these values.
Setting and then achieving conversion goals
Now that we, hopefully, have a better conceptualization of what it means to optimize conversions for different groupings, we can better examine how to set and achieve goals based on data. As a reminder, our example uses a sales team whose leads are broken into three groups of 20 leads with a 25%, 50%, and 75% chance to convert.
The first question is: What are our chances of converting X leads from the total 60 leads we have? The shifts in likelihood change very fast. You go from all but guaranteed to convert at least 22 leads to it being nearly impossible to convert more than 38. Where an organization wants to set its goal is up to them, but by understanding what the odds of achieving different goals are, they can better set those goals. In this case, we see that there is a 76% chance to convert at least 28 leads and a 34% chance to convert 32 or more leads. These could be good goals/stretch goals to give a team.
The next question is: How should we convert that number of leads? Again, we can turn to the math. Every combination of conversions from our three groups has a different likelihood of occurring. But, by identifying those most likely to happen, we can better support our sales team in achieving their goals. A visualization of the most optimal combinations can be viewed here; however, the matrix quickly becomes incredibly large. We will use the table below to better look at the combinations that matter to us.
Based on the above table, we can see which combinations of conversions are most likely to achieve the desired number of total conversions. For example, if the goal is to convert 30 leads, you would want to convert 4-6 leads from group 1, 9-11 from group 2, and 15-16 from group 3.
It should be noted that this information is not designed to say a sales team must convert within exactly these ranges; it is designed to help sales team members find opportunities they may be missing. If a team member converts eight leads from group 1 (red), that is incredible; however, it shows that their next targets should statistically be in one of the other groups.
Related Information
From the above, a sales team can better set and achieve their goals by utilizing simple modeling techniques. However, this process has several other benefits.
Encourages honesty in the lead generation team — Many organizations have goals for their lead gen team that too heavily focus on getting “high-quality” leads. Although this should still be a priority, it does create an inherent pressure to push more leads into that rank that may not fit there. By placing leads where they belong, we can create a more accurate mix and develop strategies to target the leads in those groups.
Reduces waste — Every potential lead is potential revenue. By predicting the expected conversion rates from each group, we can ensure our leads are not left on the table.
Set organizations up for more optimized testing — By breaking down leads into different sub-sections; an organization can begin to develop tests and analyze how different lead types react to different strategies.
Improve company predictions — By knowing how the different leads are expected to convert, an organization can more accurately model its projected closing numbers and revenue.