Exploration vs Refinement: Which approach to follow when optimizing your mobile app? (appiterate.com)

During my high school, I learned a very interesting concept in mathematics once. It was the concept of Global Maxima and Global Minima. Before I tell you whatever it has to do with your app and its optimization, let me briefly explain the concept to you. It goes something like this. If you plot any graph, there are bound to be highs and lows in it. Any point where the graph rises and then falls, thus forming a peak is called a ‘Maxima’. A graph can have several such maxima.

Now, I don’t like to take math too seriously, so I figure it won’t be much trouble if I draw myself on this graph and try climbing one of these ‘peaks’. As you can clearly see, I’m setting out to reach the top and finally when I do, I seem quite happy with the achievement. But is it really such a cause to be happy about? You see, I did end up reaching the highest point of the peak but not the highest point of the highest peak.
 
So where does this lesson from my math class apply while optimizing our apps?
 
It teaches us the importance of exploration over refinement.
 
 

Explore, not refine

 
Let’s take a scenario where you have an e-commerce app. As in any e-commerce app, you want to present the list of items you are wishing to sell on a page. Suppose in the present version, there are six items visible at a time and the user has to scroll down to view more. Now you want to decide on the magic number of items that must be shown at a time in the top down scroll that would be most attractive to the users. Here what you are really trying to do is ‘refine’ your app. Your app has a certain style and you don’t want to toy with it.

Going by this approach, you might end up finding out that your hits go highest when that number is four. Anything less than four reduces hits and going over four also shows a decline in the number of hits. Just like in my math graph, you have reached a maxima. But the question to ask ourselves here is whether this maxima is a global maxima or a local one? You can be rest assured that almost in all the cases it would be a local maxima. So what would exactly be a Global Maxima? To find that out you will have a take a different approach, most likely even climb down from that peak, in simpler words you’d have to ‘explore’.

In your e-commerce app’s case, it’s possible, just possible that the answer to the maximum hits lies not in the number of items in the top down scroll, but maybe using a different kind of scroll altogether. Going by my math problem’s analogy I’d say you’d have to go to an entirely different segment of the graph. Using this approach, you might land on a variation which uses side scroll to swipe between the item list. Now it’s just a question of deciding how many items to display on the side scroll, which, lets say turns out to be four. Hence you’ve reached the global maxima and I’ve climbed the highest point of the highest peak.

Of course there is always a risk. Trying out such radical changes might also lead to global minima, i.e. a variation which the user utterly hates. And it might very well be that you go bankrupt before being able to try out all the variations. But all my math teachers used to say “There is always a solution”. And true to their word, here too there is one.

It’s A/B Testing.

If you happen to have ten thousand users, surely you can’t present all the crazy variations to every one of them, else you risk losing your entire user base. So you A/B test, i.e. you show a variation to not all but just a small segment of your users, say fifty. Now even if it’s a worst case scenario and the variation is completely absurd, you only risk losing 0.5% of users and on the plus side you now know that you need to work on a variation in the opposite direction. And the best case scenario being you hit the right chord and those five users love the variation, you can right away roll out the winning variation