Finding the correct Conversion Rate
How do you know which conversion rate to use to measure success? Lance Nelson of Mezzo Labs shares his ideas on which metrics to use, and when.
When you’re talking about a process such as an online purchase funnel, it depends whether you want to measure the overall effectiveness of your form, or the intermediate steps in that process, such as moving from order summary to order confirmation. The latter methodology is particularly useful if you are examining the form’s design because you think customers are dropping out at a certain point.
So there’s overall conversion and step conversion, and it’s very important to use the correct one to fit the purpose.
Overall conversion vs. step conversion
One of our financial services clients recently got their conversion rates slightly confused in an insight report.
The hypothesis was that their application form was too long, and that a high level of drop-out was occurring before the prospect clicked Submit to process their application. Web analytics data would be used to validate this theory.
Their report presented findings with the aid of a funnel diagram, in which the step conversion from arriving at the form to the first field, Name, was 48%. No cause for alarm you may think – many prospects click through to forms just to see how long they are, what’s needed to complete it etc.
The next step conversion, for those who completed the form by clicking Submit, was then presented in the report as 67%. Again, no cause for alarm – this proved that if prospects started the form, two-thirds were likely to complete it. However, this was also presented as the overall conversion rate. 67%! This would be considered a success by most marketers, and not cause for any alarm.
Using the correct conversion rate
The client had made the error of dividing the final number of applicants by the number who had completed the first form field, which resulted in the vastly inflated conversion rate of 67%.
The correct calculation for the overall conversion rate would require dividing the final number of applicants by those who had arrived at the form. Using this calculation gave the correct overall conversion rate of 32%, which would have backed up the original hypothesis.
Thankfully, the client took my advice to change the conversion rate calculation and the report was very well received by the key stakeholders. I was also thanked for pointing out the error and for enhancing the report, effectively making the client look good!
Basically, there are two lessons here:
- Use the right calculation method to answer the right business question
- Show your calculation method clearly in your notes
In this case, step conversion would be used to report on form usability, and overall conversion to report on performance of the entire funnel process. The difference between them is fundamental to understanding your online conversion rate.
There may be other considerations to bear in mind:
- Calculation. Do all parts of your organisation agree to use the same method to calculate conversion? Does the UX team say conversion is 5% and the media team say it is 20% because they calculate it differently?
- Filters. Are you filtering out visitors who have absolutely zero chance of converting? Internal traffic, spiders/bots, casual browsers who are not really in the market for your product, e.g. curious visitors who just want to browse your site with no intention of booking.
- Metrics. Are you capturing interactions at session level, not page-view level? In other words, are you counting visits to each step of the funnel only once, irrespective of whether the visitor has toggled back and forth?
- Technology. Is the page built in a way that might affect the way tags are fired? For example, dynamic forms built in AJAX may mean you need to customise the tags to capture interactions with all sections of the form, not just the whole page.
Once you’ve taken care of these, you should be in a good place not only to accurately measure conversion, but also to optimise your rates through audience segmentation, multivariate testing and other techniques. And that’s for another blog post!