Funnel analysis: investigating why users drop

You probably know that funnel is the #1 tool to analyse users behaviour. Most of the times it is not the funnel that helps find answers though: the funnel helps to identify symptoms of the problems, while some other techniques help to find the root problem.

Couple of techniques are described below:


Users do not search for flights

About 30% of our users did nothing on “Add Flight” screen. We wanted to know why, so we came up with some hypotheses:

Add Flight — Screen 1
  1. Users don’t have a flight at the moment they open the app for the 1st time — that’s why we added “Sample flight”. It did help, but didn’t completely fix the problem.
  2. Users do not find the airline/airport they are searching for — so we decided to anonymously track what users enter into the search fields. And we started to see some exciting patterns: people enter some weird.. oh no, not weird… some strange.. nope.. some foreign letters! They were trying to search “Домо” for Domodedovo airport in Moscow and we had no assumption they do that.. I know it sounds obvious now, but believe me this was not obvious for us at the moment: we used English wordings all the time and had no idea someone would use Russian, or Spanish, or Arabic… Eventually we have parsed Wikipedia using dbpedia to extract airport and airline translation into native languages we support in the app… So, check your assumptions and habits and phone device settings ☺ And make sure you log user input for further investigation. But do it anonymously
Add Flight — Screen 2

Eventually we were able to decrease the “Not found” rate from 30% to 5%. Any ideas why those 5% are still there? You might wanna check the app to find out the answer ;-)


Users do not search for ground transportation

Another problem was with our ground transportation screen: only 64% of users searched for taxi. What was the problem with 36%?

Taxi Search — Screen 1

We had no idea, but our experience (see previous section) helped us to find the problem: we started to track what people enter into the address field. And it came out there were 2 main problems with it:

  1. Users copy/pasted addresses from their hotel confirmation emails and our “autocomplete” feature did not account for this
  2. People usually entered just hotel names, but we filtered (don’t ask me why) “estabilshments” when fetching results from Google Places API.
Taxi Search — Screen 2

We were able to improve the conversion from 64% to 81%, but still have another 19% of users not choosing address.. Any ideas why?


The last thing I wanted to notice is that sometimes the best way to investigate users drop is to just ask users (!). 15 minutes of Skype talk or a 30minute meeting with your user may yield much more insights that hours of brainstorming in office.

What about your insights on reasons users drop in funnel? Let me know.