Where Mobile Analytics Ends - And Mobile Marketing Begins

I finally caved a couple of weeks ago and got a FitBit. All of my friends are doing it, and the added inspiration and amusement from David Sedaris’ essay in the New Yorker pushed me over the edge.

Despite an awareness of the sedentary nature of my lifestyle, I was still shocked by the first week’s data.  My dashboard told me that there were a mere 15 or 18 minutes of activity in my day. I had to take action. Simple adjustments such as getting out of the Muni at the next stop and walking back, or taking the stairs (just kidding, what I mean is I walked up the escalator) made a huge impact. Today I’m up to my 10,000 steps a day with moderate effort, and already planning how to improve on that.

That’s the thing about data: however interesting the story that data tells us, without the power to take action it can be disheartening.

It's the same when you consider your marketing efforts. The key metric to consider is uplift. This helps keep goals achievable, and facilitates a more concrete understanding of the cause and effect relationship between the message you send and your users’ behaviour.

Setting A Baseline

For example, if you’re running a campaign to encourage users to opt in for push notifications, first take stock of how many users have accepted your request in the past. It’s all very well to say that you want all, or 90% of your users to be accessible via push, but if only 35% of them have opted in, historically speaking, you might want to set a goal of 50% (which, incidentally is a solid target).  That would represent a 43% uplift - which is pretty much how I’ve done with my FitBit efforts.

Naturally, you should A/B test your messaging and your timing for these messages. Once again, I’d recommend comparing the uplift of each variant against the control, and each variant against each other as well. The golden rule in all data analysis, after all, is "compared to what?"

Let’s take another example - conversion rates - conversion meaning a first purchase in this case. Your data probably show that users tend to convert early or not at all.  Once you have your baseline, you can start to take action to improve that.  

You might decide to show a special offer to users new to the app, be that a reduced price on an item, or a month’s free subscription. At first, you’ll have to guess how many times is optimal to show this message. Let’s say we show it five times to each user who has not yet made a purchase.

This campaign will give you more information to work from.  Look at the conversion rates by impression count.  Maybe 200 people make a purchase after seeing it once and 120 people after seeing it twice, but only 35 make a purchase on the third impression with an even steeper drop off after that. This tells you that 3 times is enough (in this example - obviously your user base will respond in their own unique way).

So we’ll take that learning and continue to optimize.  Now we know 3 times is best, let’s test what price point works best. Do your users respond better to a one month or a two week free trial?  Does the $2.99 price point for your virtual currency work better than $0.99.  

In short, continue to gather data, continue to take action.

Meanwhile, I’m working on getting to 11,000 steps by incremental changes.  Wish me luck!