Growth prediction: How to predict the increase in number of transactions after a feature is added?
Sudhanshu Gautam
3 replies
Hey Makers!
I am just exploring about product design and management. I stumbled upon this problem of predicting the growth after a feature is added to an existing product.
I know it's a bit vague of a question but I am not looking for specific solutions too. Let's understand the problem (generic) first.
There is an app ABC which has 100 million monthly active users. The product owners are planning to introduce a feature Y which solves a very basic problem of the users.
Let's assume it like a shop app. Users end up wasting their time to find out the item they want whereas there is data enough to predict what the user might want. The app is still doing good and so are the competitors.
Even though it is not a big hurdle for the business, the users go through a tedious process which they don't address or think about but assume it is their fault at not being efficient.
What is at stake is the UX purely.
In order to solve this problem, the product people have come up with this idea which will be incorporated in form of feature Y.
Now the question arises to quantify the benefits of building this feature at a *high* cost (both technically and operationally). How do you come up with that number?
And going further, how do you justify it?
What is being proposed is basically a hypothesis of a feature that must result in a good UX, but how can that be quantified?
Please provide your views on the process. If there are any resources you refer to, please share them too.
Thank you very much for your time :)
Replies
Ryan Hoover@rrhoover
This is a great exercise to go through for just about every feature you build, @sudhanshu_gautam2. Estimating impact can be difficult, but generally easier when building on top of an existing product.
First, I would try to create a funnel. It could be a single step (e.g. upvoting a common on PH) or multiple steps (e.g. registering a new account on PH). Start at the top of the funnel (i.e. step 1) and try to estimate the number of people you'll reach. E.g. At PH we already know how many people visit the site every day so if we're optimizing the registration flow, we have a clear estimate for step 1. Then make some smart assumptions – based on historical data or industry averages – of how many people might complete step 1. Then do the same estimation for step 2. And so on...
At the end this should give you an estimated impact for the feature. If you find the results too small, rethink the design and try something bolder.
Now if you're building something brand new, this process is much harder but you can try to run tests or analyze similar competitors to estimate these numbers.
Share
@rrhoover thanks for the reply. You explained real good.
But what about the case when you are building something that doesn't exist as a feature.
For example, when facebook rolled out the reactions on posts. How did they justify it?
How did they justify that there would be x amount of growth when this hypothesis is implemented?
Or did they even justify it with numbers?
According to me, if engineering of a new feature is concerned, the growth must satisfy the requirements.
What do you think about that? In the second step, what would that data matrix look like to decide further when the market hasn't seen the feature.
@sudhanshu_gautam2 this could be a real long answer but I'm afraid I have to run shortly.
In the beginning you have to use your intuition ideally informed by data or evidence. E.g. Is there a unique behavior among a certain population that you can build for?