Have you watched the movie, Minority report, where Tom Cruise saves lives by detecting crimes beforehand?
In this post, you will know how to proactively save your customers from churning by detecting it beforehand.
This is part of the Solving Churn blog series. We have spoken about how Churn is hard to solve with current methods in our previous post.
Value of Predictions
The Return on Investment
With more and more money pouring into internet companies, companies have been collectively spending billions of dollars in growing the user base on their platform and fight off the cut throat competition to survive in their respective domain or vertical.
But the sad reality is that companies on an average lose 80 per cent of their customers in the first 30 days of acquisition.
So, for example, if you had a marketing budget of $1000
Money Spent= $1000
Money wasted = $800
Future LTV wasted= 💸💸💸
Not many companies can survive at scale if the Churn problem is not mitigated as soon as possible.
2. Behaviour Predictions can save the day 🔮
I bike to work every day in Bangalore which is known for it's chaotic traffic. The likelihood of one reaching from point A to point B is determined by how good one is at anticipating movement of vehicles and pedestrians so that one can avert a potential crash.
This sort of skill comes from experience riding on these roads for a long period of time.
Let's break down how one is able to do this,
Input Data: Video of how the environment around me is like, how the vehicles and pedestrians are behaving etc.
Prediction Engine: Brain(Experience from previous rides)
Output: Preventive Proactive measures to reach my destination safely
This is exactly the same principle we can apply to our problem and proactively save users from Churning. The 'experience' comes from having gone through these roads safely, having a couple of crashes/falls in the past, the weather and million other signals which are subconscious to us.
If we can somehow experience or 'learn' how users have behaved in the past(those who churned and those who stayed) we can predict how your current users will behave in the future.
Growing your company with Predictions 🚀
Predicting Churn behaviour in users can be equally life saving for companies.
Input Data: The Data Universe
The Data Universe of a company
Almost all companies broadly collect the following data,
Product Data: Data around the products being offered by companies to the Customers. For example, Movie Genre, Actors etc. for Netflix, Food data for food delivery companies etc.
Marketing Data: All interactions between company and the customer. This can be marketing campaigns through Push notifications, email or SMS
Feedback data: All support related queries, ratings etc.
App events data: All the events which a user performs on the platform to achieve desired outcome.
Transactions data: Order Value, number of orders, Transaction history etc.
This helps you build an almost complete picture of the user for your Prediction engine to learn from.
Proprietary External Signals(like locality scores etc.) relevant to your company can drastically help increase the accuracy of predictions
Prediction Engine: Making a machine learn your business 🤖
Humans suck at analysing large amount of signals. With the recent spurt in behaviour data, manually building rules for 100s of possible signals of Churn is neither feasible nor accurate.
Comparison of performance of Analysing data between Humans and machines with increase in data points
Just like how you learn to avert danger with every different obstacle you face as a commuter, you can make a machine learn how different paths, activities and 100 other signals from the user's Data Universe led them to either Churn or stay on the platform.
Since Churn is measured in time windows(For example: Users who don't purchase in 7 Days), we give examples to our Prediction Engine to learn how behaviour over time windows has led to a resultant behaviour (Churned or Not Churned) in the specific window. In the image below, each dot represents a window and the red dot being the resultant behaviour.
Each row represents an example. The blue dots represent users' past behaviour over time. The red dot represents if they have churned or stayed Source
Now once the model has experienced how hundreds of thousands of users behave(Churned or active) based on all the signals from the Data Universe, you can Predict how your current and future users will behave in the next window.
With Predictions, your life becomes very simple,
Step 1: Focus
Step 1: Focus
Find the Churn Risk %age of users with the help of the Prediction Engine and segment out the high risk users.
Step 2: Understand
Understand the potential reasons for their high Churn risk to proactively improve your product and marketing efforts.
Step 3: Take Action
The best part is that once you start taking actions, the Prediction Engine will start recommending actions for different segments of users based on the performance of previous actions.
Putting it all together
3. The Return on Investment 💵
Focus on users who are actually at risk and hence save up on retention costs
Save up on acquisition cost to replace the Churning users
Protect the Future Revenue from Churning users
So your war chest for Marketing just got much bigger with Predictions and the saved money can be invested back into acquiring more users and proactively retaining Churning users to compound your growth week on week.
Lot of legacy tech is being disrupted with the advancement of Artificial Intelligence and it is high time we had a Zero to One solution in reducing Churn.
So now the question to you is, what would you rather do?
Spend more and save less users
Spend less and save more users ?
We are offering Free early access to our AI powered Churn Management Platform for a limited period. Sign Up now to transform how you retain your users.
Also, stay tuned for the next part of the series where we will talk more in detail about the existing prediction methods to forecast churn.
Subscribe to Marax AI Blog
Get the latest posts delivered right to your inbox