Suresh Kondamudi is the Co-Founder and CTO of CleverTap.
This article is adapted from a talk he gave at Product-Led Summit, London 2019.
"We started out in 2013. Today, we operate out of six regions. We have about 200 people now working in various locations."
The Clevertap homepage defines the company as:
A mobile marketing platform that helps high-growth mobile apps retain users. We look to improve retention rates with unified customer data and we are the only platform that uses AI and machine learning to personalise the customer experience.
We are not deterred by scale. Our ability to process billions of events in milliseconds allows businesses to respond to changes in customer behavior immediately – no matter how large the user base.
"Our customers typically include growth teams and product teams within consumer businesses.
Growth teams sometimes roll into marketing teams. Sometimes they roll into the product teams. Realistically, a significant amount of your KPI’s will, in some way, be related to user retention."
The problem with user retention
"When you talk about growth teams, they work on two different important problems. One is user acquisition. The other is user retention.
User acquisition is expensive. It costs anywhere between five to ten dollars per install.
Most importantly, the money goes down the drain if you do not work on the subsequent retention.
User retention is cheaper, it typically costs around one-tenth of the cost of acquisition. But, it's hard.
Here's a very depressing graph. Unfortunately, these are the actual numbers from a real case study.
This is called the leaky bucket problem and it's very frustrating for growth teams. You fill your bucket to the brim with lots of shiny new users. But, slowly and surely, 'drip' 'drip' 'drip'. Before you know it, you've lost 70% of newly acquired users within seven days.
Growth teams usually have a lot of ideas. But these ideas are iterative. It takes a long time to carry them out. We don't have this time. After waiting for these ideas to launch, you might have already lost too many of your initial users.
These guys also struggle with a lack of necessary infrastructure. They lack the tools to solve these problems speedily."
How to solve the leaky bucket
"There are two prongs to the solution
1) Effective engagement
2) Providing a delightful customer experience
Thematically, these are the two sub-areas but sometimes it feels like one of the same since you must use effective engagement to provide this delightful consumer experience with a subtle difference."
What is effective engagement?
"It boils down to creating engagement hooks to deliver contextually relevant experiences and appropriate moments with preferred touchpoints.
The notions of suitable context and communication paths couldn't be more important.
Take this for example:
Imagine a SmartFridge. You run out of milk. The IoT consensus device processes this and then sends this as a message to you on your mobile phone. Great. How handy! You were just about to walk past the shops on your way back from work.
In this context, the customer is delighted with the product. How could you not be?
But, then you get another message. It's 2 am. Your phone lights up. *MESSAGE* 'Your water jug is less than 3 quarters full.' You roll back over. Irritated. You see the phone light flash up again. *MESSAGE* 'In 48hrs, you will run out of strawberries'.
These messages have now just become spam. Understanding the context by which they are sent and received is paramount.
Here's an example of a service that understands the importance of context.
You’re a real jet setter. You're flying all over the world for your work, regularly checking into hotels, checking out the next day, flying to your next meeting or conference. You have a very tight schedule and you’re constantly rushed off your feet.
On one occasion your flight's delayed. It happens. You land at 11 am. Your important meeting is at 12 pm.
Now enter the app.
And as soon as you land, a message pops up, it tells you that there's no way to make it to the meeting if you go to your hotel first. It suggests you take two cabs. One for you, to the meeting room and one to take your luggage to the hotel. It proceeds to order them for you.
This app will have been triangulating this information that the flight is delayed and mapping the traffic conditions.
Whilst you go straight to the meeting room, the app checks into the hotel for you, providing some sort of a display key. Following your meeting, you walk straight to your hotel room. No need to check-in. The room ambiance is tuned as per your taste, based on your previous visits.
Now, this is a very good experience.
Notice that this is not just the work of one app. It's not just one company. It's multiple companies, collaborating to provide the seamless end to end experience
If it's a good experience, people tend to remember it.
And what people tend to remember is the experience rather than just the service."
"Let's explore the models that will help us to orchestrate these perfect experiences."
Hierarchy of engagement
"This is a really simple tool for any given business.
It boils down to breaking down your larger goal into smaller milestones so that you can solve the problems at hand."
Acknowledgement is at the bottom of the hierarchy. Interest is somewhere in the middle and at the top is the Conversion.
Most significantly in the conversion we have the Core Action.
If you're in E-commerce, the core action is the participant. If it's something like Spotify, the core action would be listening or adding to the playlist.
Here we are breaking the customer engagement down.
The next step would be mapping your users onto this and then navigating them through these milestones. You need infrastructure to do that."
Customer lifecycle management
"This is more granular."
"We can see the milestones of the user journey have been mapped. From the brand new users, to your onboarded ones. Then, to your completed first transaction. The second purchase is another milestone.
Once the customer gets to the next stage, you lock them in with some sort of loyalty program. Airlines do this well with their use of 'air-miles' and the like.
Sometimes, people churn. It happens. But, you must react with a program for that.
These steps are not easy, nor are they linear. People can jump from stage to stage.
If used effectively, retention metrics can be improved depending on the business. But, realistically, how do you do it?
These charts look good on paper, but what are the tools that will enable you to succeed?"
The building blocks
"Both of these models focus on effective engagement. On delivering that right message at the right time that works on providing great experiences for the customer.
Putting these models into practice isn't necessarily the easiest task."
Step 1) Think about the 'who' 'what' and 'when' of this experience
- Who do you engage with?
- What content do you engage with?
- When do you choose to engage?
"Also, what are the preferred touchpoints? Which channels of communication are you going to use?"
Step 2) Profile the customers
"We call it customer 360.
You have to profile in terms of demographics, geography, technographics, behavioral data, interests.
Do this until you know everything relevant about the customer as possible."
Step 3) Build a strong segmentation suite
"You need to be able to segment people on a real-time basis.
Why real-time? Because you need to preserve the context.
Remember that SmartFridge? When taken slightly out of context, useful communication can become spam.
To build the segmentation list for this, there's a lot of context and a lot of real time consciousness required.
Your move into this so-called automation suite consists of navigating the customer through the user milestones.
That's not going to happen manually. The ideal onboarding of a customer will need to be automated by you."
Achieving the ideal onboarding workflow
Think...what is the ideal?
"You might think you already know it because of the domain knowledge that you have. You might have been in the industry for a while. Even in that case, there are so many micro-patterns that emerge which you should be leveraging."
How do you move into ideal onboarding automation?
"When moving from onboarding to the first transaction and then to the second transaction, the automation workflow can look very different at each stage.
We can use a very well known technique that is used to segment users to map to these stages. It's called RFM analysis. It's a classic.
You use this to segment users into 10 different categories and then map into one of these lifecycle stages. You can use tools and necessary infrastructure to navigate customers from one stage to the next.
To do that, you need to have this automation suite and you need to optimize this 'who' 'what ‘ and when' aspect while figuring out automation flow.
The only way is to test. You experiment. You measure. You keep experimenting until you find a satisfactory result. 'Can I improve that?' Of course, you can.
Data scientists play a significant role in optimizing this aspect. Based on the past data, they can layout templates for ideal onboarding. If you have this, you can improve this loop.
Remember, this is iterative. There’s a reason why it's hard. It's more of a science than an art.
Remember to collaborate. Multiple apps can collaborate to provide a seamless experience.
This is what it takes to implement effective engagement and in the process, delightful consumer experience."
"When it comes to solving these problems at scale, it can become very complicated.
One of the major problems comes with talking about customer 360 and triangulating this information across data silos.
We sometimes assume that AWS has an answer for everything. But in reality, the Cloud is just someone else's computer. You need to solve this problem more deeply.
Solving these retention problems at scale is challenging but it is possible."
This is an Indian streaming service.
It operates at an immense scale, over Southeast Asia in areas like Singapore, Indonesia, Malaysia. Because of its location within these populous areas, the scale of the company is immense.
These guys have improved their user retention for a record-breaking 10.3M concurrent users.
- RFM Based Segmentation
- User Retention Cohorts
This is a fantasy gaming gap.
People make a lot of money playing with virtual teams. Again, it comes from India, which has a large population.
They retained 5 times more customers with advanced user insights.
70 percent of their inactive and churned users were re-engaged with personalized campaigns.
- User Activation & Retention
- Brand Recall
- Marketing Automation
These are real examples that deployed models we have discussed. So, what are you waiting for?"