Skip to content

Variation in Your Sales and Marketing Results

James Roloff
James Roloff
6 min read
Variation in Your Sales and Marketing Results

You’re stressed and anxious. Your numbers are NOT looking good for the month. You know that the upcoming sales meeting is going to be rough. It’s a terrible feeling, especially as you don’t believe you’ve done anything differently than last month.

Last month… when your team was rewarded with a pizza party to celebrate the great sales numbers. The company president even called to congratulate you on the great results.

Same inputs. Different outputs.  Did you do something wrong? Your management is sure acting like it.

But the answer is simple - variation.

Every system has inputs, processes, and outputs. Every system also has variations.

Anybody working in quality control is well aware of this. Their job is to closely monitor the output of systems to identify variations that cause quality control issues.

The job of a quality control manager is to try and reduce the variation down to a reasonable level. In the case of manufacturing chicken tenders, some variation in size and shape is okay. In the case of pharmaceutical manufacturing, almost no variation is allowed.

Quality control teams know that to control variance, you have to look at the system as a whole. What are the components that lead to variation? What factors influence output?

These same principles can also be applied to business development processes, including your sales numbers, lead generation, and account activity.

Example of Variations in a System

Let’s first look at an example of a system in everyday life - driving to the grocery store.

The system itself is pretty simple:

  • Inputs - the driver and the car
  • Processes - the route, the traffic lights, and the functioning of the car.
  • Output - the time it takes to drive to the store.

On average, driving to the grocery store takes me about 12 minutes. This is the expectation that I would have in terms of the “performance” of the system.

Every time I drive to the store, I have the same inputs and the same processes. So the output should always be the same, too, right?

Not exactly. Over the last eighteen trips, here are how many minutes it took me: 9, 10, 14, 12, 12, 12, 11, 10, 12, 12, 11, 10, 19, 13, 11, 13, 13, 11.

Here is a graph showing the number of occurrences at each duration:

As you can see, while it takes an average of 12 minutes to get to the store, there is variation in the system's outputs. There are factors in the system that cause variation in the time it takes to drive to the store.

In this case, it wouldn’t be statistically significant if it took me 10 minutes, nor would it be strange if it took 13 minutes. Those trips would be “normal” for a system with variations.  Maybe I hit all the green lights? Or maybe I was stuck behind a slow driver?

However, there was that one trip that took 19 minutes. Was that the result of normal variation in the system, or did something unique happen there?
By understanding the variation in the system, we can do two things:

  1. Benchmark system performance and analyze individual data points.
  2. Implement process changes to improve results and address variation.

Taken together, this helps demystify our performance and sets us up for success in executing our strategy.

Measuring Variation

Variation is typically measured in the disbursement of a set of data points. You’ve seen this before in reviewing a “normal” distribution, as seen in a bell curve like the example below.

Source: Wikipedia

Many parts of our lives follow this pattern, including the height of people or IQ scores. But it can also be applied to system outputs, like my grocery store drive example above. Or, more relevant to our business strategy, systems like sales performance or lead generation.

The distribution of data points tells us to what extent variation can be expected in a system. The further a data point is away from the average, the less statistically likely it was caused by “normal” factors. It becomes more likely that something special happened that caused it.  On the other hand, if a data point exists within two standard deviations of the average, normal variation was likely at play.

One way to review this data is with Control Charts, also known as ​​Shewhart or statistical process control charts. These plot data points in a way that makes it easy to visualize whether or not your system is in “control” or if there are anomalies with any data points.

Here is the control chart for my grocery store trips:

You can see each trip plotted on the graph. The red lines are the control limits. These are approximately 2.5 standard deviations away from the average. Meaning, that it’s statistically unlikely that any data point from this same system would fall out of this range.

From the chart, we can draw a couple of conclusions.

  1. My grocery store trips from about 8 to 15 minutes are normal. I wouldn’t have to do anything differently to arrive in 9 minutes instead of 14 minutes. This type of variation is expected based on system factors (traffic lights, traffic around me, etc).
  2. The 19-minute trip was statistically significant. This trip is likely a special cause and should be analyzed to see what happened. In this driving example, maybe I encountered an accident that blocked an intersection. Or I had to change my route to fill up my gas tank quickly.

By measuring and understanding your system’s variation, you can identify what is normal and what results from a special cause. We call these common and special cause variations:

  • Common cause variation results from natural or expected variation in the process.
  • Special cause variation results from unexpected variations or factors in the process.

Once the thresholds of your system’s variation are established, you can analyze your data to determine if outputs are common or special cause.

Improving Your Sales and Marketing Results

So, let’s return to your sales results. It’s been a bad month; you’re nervous. You know your manager will be “having a talk” with you about your results.

But was it really a bad month? Or were your sales in line with natural variability in your sales system?  

Below is an example Control Chart of leads generated by an actual client by month.

On average, this business development person generated 12.5 qualified leads a month. However, their system variation had an upper limit of 23 leads per month and a lower limit of 1.8 leads per month.  

What does this mean? Based on their current sales and marketing systems, anywhere between 2 and 23 leads per month are to be excepted.

Getting 7 leads in February could result from the same activity, processes, and methods that generated 18 leads in April! (I hope you still get a special lunch with the boss for April’s results.)

To make this point clear - common cause variation in your sales and marketing results doesn’t necessarily mean anything unique happened (good or bad).

Armed with this knowledge of variation, let’s talk about how you can use it to improve your sales and marketing process. Here are 5 tips to get started:

#1 Capture Good Data

Understanding your system results and variation begins with good data. Your entire sales and marketing team needs to be accurately tracking data. Ensure that your sales CRM and marketing software is set up correctly and collecting the correct information. You’ll also need to work with your team to create standard definitions for tracked values.

#2 Calculate Your System Variation

Once you have a data set of outputs (sales numbers, leads, conversions, etc.), you can run your variation analysis. You’ll want to find your process averages and the statistical distribution of the data (i.e., the standard deviations). Creating graphs in Excel and using built-in statistics tools are great starting points. You can use software packages like QIMacros and Minitab for more advanced analysis.

#3 Don’t React to Common Cause Variation

It’s time to STOP having pizza parties for good months and tough conversations for bad months if the data points result from common cause variation. Reacting to single data points that are range bound of the control limits is not only unnecessary, but it can also be harmful if you over-react.  

#4 Separate Signal vs. Noise

You can better separate the signal from the noise with your variation understood. Where are the outliers? What are the special cause variation events? Digging deeper into these data points can help you better understand unique circumstances that affected your system.

#5 Change the system

And finally, if you are unhappy with the distribution of your data points, there is only one thing you can do - change the system. If you want to increase your average or reduce variation, you need to find the factors impacting your results and improve upon them. This might mean increasing training, adjusting ad spend, changing sales methodology, or more. You can't force the results to change without improving the system that creates them.  

Comments