Funnels: Optimizing Funnels
Once you are tracking your funnels, and have some benchmarks in hand, it should become clear where you have opportunities to improve conversion rates and hence improve your overall business. For example, your funnel might look like the following:
There is a steep drop off between users visiting the site and searching for shoes! Clearly something must be amiss, so we can start to optimize this funnel by changing the user experience. The best way is to simply remove one of the funnel steps if at all possible! If there are fewer steps then you have fewer chances to lose the customer (hence, Amazon’s one-click purchase).
Removing steps is easier said than done, so a more likely solution is to update our website to make search more prominent and make sure everyone knows they should be searching. That should increase the number of people who search after visiting the website and hence our total purchases! After we do that (in this example), our funnel looks like this:
Many more people are searching but, wait, FEWER people are buying??!! We increased the number of customers engaged with step 2 in our funnel, so why did it hurt our overall performance in step 5?
Funnel optimization is a tricky business because the steps in your funnel might be related in ways you aren’t aware. This example is based off of a real world experience with a company who did not realize that their most loyal customers didn’t even use search. When they redesigned the site around search, those loyal users couldn’t shop the way they wanted and left the site!
To optimize our funnel and improve conversions, we need to be able to test improvements and see how they affect the overall funnel. Luckily, we can use our good friend A/B Testing! In the above example, the change to search would be our Group B and we could quickly compare the results against the current site, Group A, and realize that we’d made it worse instead of better.
A/B Testing can itself be tricky for funnels because it’s possible that the driving cause of conversion rates are characteristics of customers that you can’t see. For example, conversion rates vary wildly with customer acquisition sources. If you use an ad network, you might not know where users are coming from when they click your ads. Your performance might be great one day and bad the next simply because the ad network is showing your ad on different sites.
Even so, if you run your A/B Tests carefully and for long enough to avoid short term noise, you should get good results quickly and optimize your funnel performance. There will be cases when you cannot optimize a given funnel any further. If that happens you will just need to throw that funnel away and build a new one.
Quote of the Day: “Premature optimization is the root of all evil” – Donald Knuth