From 0-1-10-100: Lessons from building consumer products for India - Part 1 || Bhavik Kaul (ex-CPO at LazyPay)
We speak to Bhavik Kaul about building Hike, Dunzo & LazyPay
Welcome to the Product Roadmap #7!
This edition is long, so long that we had to split it into 2 (or maybe 3) sections.
And that makes sense because Bhavik has spent more than a decade building consumer products for the Indian user at Microsoft, Hike, Dunzo, and then LazyPay.
Our conversation spanned well over two hours, and this post (Part 1) is about Bhavik’s insights building consumer products - we talk about the insight which powered the growth of Hike, increasing the frequency of usage at Dunzo, the hack launching credit on UPI at LazyPay, and a framework to build retention products. The next post will cover the softer aspects - how to figure out if you’re suited for a PM role, evaluating PM candidates, building influence as a PM and much more.
Warning: if you’re the kind who wants quick summaries, this article is not going to be easy for you. But then again, you can only get so much insight from summaries :). If you spend the time, there’s lots of juicy stuff in there (like Hike’s retention at its peak.)
Here we go!
Here are the key sections we covered
Overview of Bhavik's career path from engineering to product management roles.
Product Experiences at Hike - The growth strategies like free talktime, hidden chats, stickers that helped Hike grow rapidly.
Key metrics tracked at Hike - The important metrics like DAU, DAU/WAU, retention that were tracked to measure growth and engagement.
Metrics tracked at Dunzo and LazyPay - orders, activations, and retention that were tracked at Dunzo and LazyPay to measure business performance.
Product development at LazyPay to increase retention and monetization
Discussion on the credit-on-UPI product built at LazyPay to simultaneously increase retention, growth and monetization.Framework for increasing frequency and retention - understanding organic demand, adjacent needs and factors influencing frequency change.
How Bhavik drunk walked into Product Management - the importance of serendipity
Mithun Madhusudan: Bhavik thanks a ton for taking the time to speak to us. To start off, maybe you could tell us a bit about how your career has progressed?
Bhavik Kaul: So I have sort of drunk walked into this career. This was not planned. I've always liked building stuff. I started in engineering, joined a company to work on java. For the first couple of years, I was just learning since my background was not in computer engineering, but after that I got the knack of building prototypes. I transitioned into a pre-sales engineer role. I worked on stacks such as Adobe Flash & Adobe Air & Microsoft Silverlight which helped me create amazing user experiences.
The first critical juncture in my life was when I met a bunch of my friends, to try and figure out what to do next. I had a close friend who was the marketing lead at NDTV and he said we should do something in organic farming on which I had no clue, but somehow I went down that path. Then another friend was working for a consulting company in the US and was looking at the accounts of Coca-Cola and Iron Mountain. He told me that he had found some major cost benefits - and if he could build a product for them he could bill them double. I found some commonality with my need to tinker and build something. So I just started working there along with my day job, but within three or four months I realized that the day job was not working, so I quit.
I did that for about two years - the experience was phenomenal. I had not expected to make more money than my day job, but I eventually ended up making like thrice the amount of money.
But I think what hit me was that I was not able to scale these things to become automated - like you can make a product or a process and then it becomes a snowball, going down the mountain and picking up momentum. It was not working that way and I couldn't figure out why.
In those days the startup ecosystem was also not great. There weren't that many people whom one could walk up to and ask for advice. So I did the next best thing in those days - I went for an MBA which was a bad decision, largely wasted a year, but I think it gave me the right break. I needed to reset. Post that I also made another bad decision to get into consulting. That was I think the second critical point where I realized that I definitely didn’t like it. But the good part is that I met a couple of friends who were from good colleges and through them I got a break into my first proper product management gig and that was at Microsoft.
There was some interesting serendipity there as well. All my interview rounds had gone well, but I think I got my job because the VP had used a product that I had built in my organic farming days and he must have been one of 50 people who had used it, and our conversation was only about “how did you build it?”
The two three years at Microsoft were phenomenal because I had not had proper training at thinking like a PM. And that's what I got there.
But in that duration, I also got lucky because this was the part of Microsoft which is not known as the Microsoft which is large, laggy, makes decisions in a year. I had just serendipitously fallen in the division which was created by this guy who had been acqui-hired by Satya. He was a startup founder and now he had become a CVP (Corporate Vice President), but after the acquisition he operated like a startup founder.
“I met companies which run big trucking businesses - they were doing like 250 crores of revenue a year, and everything was managed on five WhatsApp groups.”
The product that I had worked on was basically for SMBs in India. The idea was that Microsoft products are used by large companies, but there are 60-65 million SMBs, which have no use for things like Excel or PowerPoint. What do we do for them? So in those days we started talking to a lot of these SMBs.
I met companies which run big trucking businesses - they were doing like 250 crores of revenue a year, and everything was managed on five WhatsApp groups. I was blown away. How were they even keeping track of all this information?
So then me and three other PMs, we came up with this suite of applications - the first was a marketing app for small retailers. The second was a CRM embedded in a dialer app, and the third was a project management solution which looked like WhatsApp.
During that time I got really interested in chat and social products. And I got lucky because a friend of mine had interviewed at Hike messenger and told me about the app. I instantly felt i should talk to them.
Hike in those days had many things working for it.
a) Right amount of funding from the right people
b) They had also put together an insanely talented team and
c) they had found a bit of a product Market fit.
So I joined them and that was an insane journey. In fact, I think I attribute all of my product learnings to that one company.
I'll talk about my favorite features or products at Hike a bit later. I think I had the most fun and the most learning in just one company in my entire life, but these are functional learnings. These are not like people or process learning. But for a variety of reasons Hike didn’t really get to where it could have and I started to look out.
That's the time where hyperlocal had this weird S-curve happening. Everybody was talking about it. In Bangalore Dunzo-ing had become like a show off thing. During this time my wife was also running a startup of her own. She used Dunzo very extensively on WhatsApp. So she told me about it and said that you should go and talk to them if you're interested in knowing more. And that’s how Dunzo happened.
Then there was the entire product journey of going from WhatsApp to where Dunzo ended up later, which was a marketplace and dark store. And finally when I quit Dunzo it was because I had kept a personal goal for myself here - I’ll take the company from X to Y. I was evaluating whether beyond Y do I have anything to do here? And I realized that maybe my skill sets or what I wanted to do is or my aspirations are not exactly a match and so I looked out.
That's where I met the folks at Lazypay. The vision was to build a credit first Neo bank. I felt this was fascinating because the company was already at PMF and I could deploy my skills to help go from 10 to 100. With a matured investor like Prosus on the back it felt like a definite win.
Today LazyPay is doing a billion dollars in lending with about 115 million dollars of revenue and this year it will break even. So a great 4 years spent. In fact, I attribute this company to all the ‘scale’ learnings for me. I learnt how to build several businesses in parallel, how to hire and construct a large team, how to align over 500 people to a single vision and drive outcomes. Doing all of this in 4yrs was a crash course!
“I remember one time we spent about five hours designing a button.”
Mithun Madhusudan:
I think this is a great introduction. I've added 2-3 other questions that we will cover later, but first would love to hear about what you liked about working at Hike?
Bhavik Kaul:
Yeah, so interestingly I played several roles at Hike. I started off with the most unlikely role. I was hired in the game gaming team - I had zero gaming background. I was not even a gamer! But something clicked and I got that opportunity - probably the best thing that's happened in my product career because in the process right I got to learn the finer pieces of experience building.
I remember one time we spent about five hours designing a button. My boss, me and two designers were sitting in a room and I was writing out emotions - and how that emotion should look like. I’d say okay if I press this button the feedback I get should be similar to what I feel if I had pressed the Pillsbury Atta logo's stomach. And then designers were building that button in the room.
That was basically the first one year or so at Hike. I got induced into thinking about what design does for building emotion. From there I got into building growth products. And finally I was building nonverbal chatting or non text chatting.
Hike’s history, ICP, and initial PMF
Bhavik Kaul:
Hike has a very interesting history. I'll just quickly walk you through it. Hike was a bunch of companies that Bharti Softbank had funded.
Their thesis was - there is this ICP which is 12 - 25 year olds - and the user wants to show off.
(If you remember they had also come with a bunch of these ads saying life mein thoda hike kar dia. That's how the name Hike had also come up.)
The thesis was that - what you need to show off is knowledge and being smart. So we said chat will remain chat, but we will have jokes, news, facts, specific offers for you - so effectively the app became ‘access to information’ through a chat interface. There was a feed which only showed you ‘Aaj Ka Joke’ and then you posted that joke in that chat to show off.
That was the entire thesis and interestingly this found a lot of takers in college towns or college states. 90% of our user base was from Maharashtra, Tamil, Nadu, Kerala, Karnataka and within that 25% was just Mumbai and Pune.
Driving Organic Acquisition - Hike’s Referral Hack
Here’s what we found out about our users from interviews. Remember, this was pre Jio. So the network was running on 2G and the internet was expensive, so how do I bring my network online? Talktime was expensive. So here’s the first massive growth/retention/frequency hack - we gave away free talk time as an incentive for you to get your network on Hike. Hike is today renowned for doing this growth hack called ‘Talktime rewards’.
If you install the app you will get talktime and SMS data worth Rs 50, and if you refer someone, that person will get Rs 25. If you install the app and send a message to your friend and if that person is not on the internet or not on Hike, he gets the message as an SMS. So you send it from the internet, and he gets it on SMS. (this was state of the art back then).
In fact, after Hike a large number of startups copied this growth hack to get people online. This was the first major uptick.
Now you had like this bunch of people coming to show off and their group started forming, and we started looking to build more value props. In interviews we started talking to these kids and asking what’s happening in their life. Everybody kept saying this one thing - ‘I have a group of close friends, who I message from my Dad’s phone, and then sit and delete all those messages at night’.
The genesis of Hike’s hidden chat and privacy features
The real a-ha moment came when we heard this story - A user told us that my mom is very smart, so when I go to take a shower, she opens my phone looking for what I have chatted about. I installed this Locker app (to lock access to Hike). But she's so smart that she's figured out that it's a Locker app. Now she's telling me to unlock it and if you don’t unlock it I’ll tell your father.
The insight inside came out of him telling us ‘what if I unlock the app and my mother doesn’t find out’. So essentially a lock pattern, which when you unlock successfully or unsuccessfully, looks exactly the same.
And that eventually became the second biggest growth hack for Hike which was called hidden chats. So you could set up a pattern to lock the Hike app. Now the beauty of that was that you set up a pattern and you unlock it, irrespective of whether the pattern is correct or incorrect, the user doesn’t get any feedback. The user sees a response saying - app is unlocked. But the app gets really unlocked (for you to see what is in it) in only one case - if the pattern is correct. So only in the right case you would get to see the hidden chat. But you got a positive response in both cases.
Eventually we plateaued out on our DAU, we reached about five million DAU and we had started noticing that there were two concentric network circles forming - one, people who had come for the show off reasons. So they were in concentric circles of 10+ people.
Then by volume actually there was a larger group of people whose Active Social Network (people who they talk to and share with) was three or four. So they had smaller social groups but by volume they were maybe more than 50% of the entire base.
When we started going down deeper, we realized that people had found an organic product market fit - Me, my closest friends, and my girlfriend or boyfriend. And there wasn't a single product in the market solving for it. Hidden chat had solved for it and the feed had solved for it.
I had actually seen that for some people their feed had become their love story. From the time that they started talking to each other onward and it was like an album (the feed). It was crazy how this had evolved to that point.
Hike Stickers - Solving for frequency of usage
This was also the time where I was trying to solve for frequency specifically for DAU/WAU. So if your organic frequency of usage is maybe 1 message a day & 2 mins per day. How do I increase the number of sessions & messages to 15x that number? It’s not by typing - it's by clicking. Now the modality is clear. But could it also massively emphasize the emotion because typing mein emotion kya hai? In an “I love you” text there is very low emotion. So that's where I started. I started working on stickers as the first differentiation and that became the differentiation for Hike against the entire market.
In the beginning we knew our users were lovey dovey couples so we made some (sticker) packs of ‘love’ words, cuddly, bunny stickers. It really took off - if you looked at a graph of core/casual/ power users, more casual users started to become core users, and more core users started to become power users - their frequency went up by like 50 to 100 percent. All they were doing is tera kiss mera kiss, main blush kar raha hoon tu bhi blush kar.
We also saw that because our user base was in Maharashtra, Tamil Nadu, Karnataka etc - there are these cultural nuances. So we started stickers in regional languages ‘you can say the same thing but in different languages and it means even more’. From there on I think we did several crazy things. I think the peak was when we started realizing that maybe we can animate the flat sticker, make it emit sound.
From a thesis perspective we were like if there is a kiss, then there should be a kissing sound, there should be a dhak dhak sound for the heart, and it should become bigger.
We also had to solve for the size of the sticker. That was the technology challenge. I think one sticker’s size used to be like 20-30 KB and the animated version of that was maybe around 1.5MB. And that size was a big deal for 2G users.
So we spend a lot of time creating the network layer, to chunk out data in such a way that it downloads in the right fashion without hampering the experience.
We did some really crazy stuff. We did full page animations. For example, a girlfriend is pissed off with her boyfriend. How do you show that she's pissed off and how does that anger reach the boyfriend? So we made full page interactive animations. If you click on that notification, the entire animation takes over your entire page, and you can’t see anything else.
Then it shows the glass crack - the animation was beautiful. The glass cracks and there is a sound - and people used to think oh no is my phone cracked.
Mithun Madhusudan:
I think there are a lot of parallels on this stickers piece. Because this is the piece that basically drove monetization in audio chat rooms. People are buying this stuff so that they can send it out to their friends and people in the chat rooms. And we also had to eventually do a bunch of stuff on the tech side. Obviously now, it's a little bit more routine compared to maybe five to ten years ago when you were doing it.
Bhavik Kaul:
I think if you look at the entire Hike journey, I eventually started orienting on these things. Full page animations, then we started doing sticker recommendations, then we brought games in chat. So in fact, I created one demo where there is teen patti on top, with people in a group, and below it there is chat. Then we realized there is no fun in the chat unless you are using cuss words. So we created this version where you write text and it got converted into a sticker - this led to the launch of the feature called "text to stickers" which became quite the rage. Now with genAI tech Apple in its 2024 annual event announced "text to emoji". The world has come a full circle for me.
Then we did this to solve for frequency and retention, if you sync your contact book and instead of saying ‘Hi’ you send him an interesting sticker. And that had way more conversion in getting people to talk to each other because it broke the barrier. You are now shocked and when you open that you could see some fun, quirky responses. Automatically your barrier to talk to each other went down and that's what increased the network graph and our DAU/WAU numbers. I think that was definitely the top thing that I've loved doing.
Why Hike didn’t work - pivoting from a closed network to an open network
Mithun Madhusudan:
Right. I think a lot of this seems a little bit before it's time. Everybody says reduce friction, become more visual, become more animated. So, maybe there's an interesting point - obviously Hike didn't work and they're doing a bunch of other stuff today. But do you think that is the reason that Hike was a bit before its time? What went wrong?
Bhavik Kaul:
Every product’s genesis, network and the use case of that network are already defined. For example Instagram from day one was an open network - I did something and I want to show off. But our product from day one was a closed network. What happened was that we wanted to sort of bridge the gap. If you want a close Network, here you go. You have a close network. If you want an open network. Here you go. You have an open network. We didn't perceive that this is not how people see products. As soon as there was this concept of an open network (on Hike) everybody who felt that they were here for privacy freaked out.
There was a way to technically create a barrier. Like on Instagram you can be private or public. That's the same thing that we had made but it failed. People churned out by the truckload. So that was the single biggest mistake at Hike. If we hadn’t done that, Hike could still have been large, albeit not like the size of Instagram, but I would have easily thought that it would be at like 20-30 million DAU - that’s the number of people who wanted privacy. It was this need to become Instagram which really broke the camel’s back. The core Network started churning as soon as they felt that the privacy was taken away from them.
The Key Metrics Measured At Hike
Mithun Madhusudan:
One last question before we close up this point. I think an interesting set of numbers from you would be what a good activity engagement and retention numbers for an app like this. Maybe if you can remember Hike’s numbers at its peak - D7 D30 retention, DAU/MAU. And why those metrics are important according to you.
Bhavik Kaul:
So we were largely tracking four metrics - ones that we always talked about. The first one was the daily active user base. This was mostly important because this was a vanity metric which showed how large you have become and how much market share the team managed. This obviously does not take into consideration churn. You can be adding a lot of people on top (through user acquisition). The second metric of choice was daily active divided by weekly active, and the idea here was to measure engagement, like how often people come in and how often they log in. I think I remember at that time daily active was 22-23 million users and the DAU by WAU was close to I think 57%.
Mithun Madhusudan:
57% is a very large number.
Bhavik Kaul:
Yeah, but you can also understand why because it's a close network.
Mithun Madhusudan:
Yeah, and it's a chat product, with daily use case.
Bhavik Kaul:
Then the third metric that you used to track was your retention numbers. Most people track D15 D30, but we started tracking hourly, like Hour3, Hour5, Hour7. We realized that maybe this person needs to get some validation quickly, discovery of a friend, discovery of a feature, discovery of a value prop. Our D1 retention was almost 95%. I think D30 had reached 70-75% back then. And the 25% churn existed because there was a bunch of people who didn't have friends who were coming here for the informational features. We had news, we had jokes, and eventually, we had games, and then we had two classes of games. We had deep experience games, which we had created (Oh, by the way, Hike has a patent for it).
So all of that used to happen. Wherever the network had for whatever reason not sit together, there was churn, and wherever you found at least two of your friends, the retention was through the roof.
We also used to track time spent per day, because in the social apps world time share comes from taking time from other apps.
Those are the four top metrics. The fifth one was very interesting. We had come up with this metric and specifically to understand what kind of network graph we have, is it very dense or slightly sparse? We used to call it ‘Active Social Network’ or ASN - within a week, how many unique people do I have any double-sided interaction with? If I send you a chat and you like my picture, that's also considered an interaction. You don't have to chat back. And if I do it 15 times with you also that's considered one ASN. So how many unique people do I interact with? This was a worry, okay, and that's what got us to think that we should become an open network, and at that time, the ASN was 3.1 or something which showed that this was not a very deep network - I'm not actively connected with a very large group of people, but this is also the value prop - my friend, my close friends, and my dating partner. This was not supposed to be all the people in the world or my classmates or my school group.
Mithun Madhusudan:
I think it's almost like crossing the chasm, right? You have a deep core use case, but by nature of the product or company, you want a larger use case, and you're not able to successfully navigate. Your core user base also drops, and the new ones also don't stick around.
Bhavik Kaul:
Yeah, and so the risk that we were always running was that if one person churned in that group, then you led to a very high chance that the rest also will churn because the core group is broken.
So that was actually one of the deciding reasons why we had also thought that we should do an open network - maybe the group will also get bigger. But yeah, I think that that didn't really work.
Comparison of Hike Metrics with what was measured at Dunzo
Vibhav V:
Yeah, since we are on the topic of metrics. What about at Dunzo and PayU, what did you track there?
Bhavik Kaul:
So that's a great question. Let’s do a like to like comparison. Like DAU at Hike, we had at Dunzo orders per day, orders per week, orders per month. At LazyPay - we tracked originations, not number of users. It’s a very dhandha type business. So originations grow and that makes revenue grow.
At LazyPay, the problem that I was set out to solve for very very early was activation rate. So that was a very big problem. And after that, the D7 retention rate became the big problem.
Activation rate at Dunzo when I joined was very low, I think it was in the single digits. 100 people came in and 9-10% transacted. And over a period of 2 years that number had gone up to as high as 40, between 40 to 50%. The retention rates were another worry, because most people had found single factor use cases - I use Dunzo for pick up and drop. Dunzo was insanely popular for pickup and drop.
When we started adding several different categories (for added use cases) over those two years, we realized that people had preconceived notions - if I see food as a category here, is this Swiggy. And then I had to find different design solutions to make people realize that no, this is actually Swiggy plus pick and drop, meat delivery, grocery from your closest store, printouts all in one. The additional use cases were important because we were always chasing frequency per week as a metric. By the way, we were always on top even back then - swiggy and zomato used to have 3 point something as a frequency per week. We were already at over four and a half. I was trying to push it to eight or nine.
Building multiple use cases for Dunzo
Mithun Madhusudan:
And this four four and a half was basically because you had multiple use cases for the pickup.
Bhavik Kaul:
Because he had multiple use cases. And eventually, our frequency had gone up to seven to eight per week.
It was largely because we had to teach the user, there are different categories and you can use all the categories over here. Because people who got in and thought this is like Swiggy and that was what that's the perception that I had tried to break. This is not like Swiggy. This is different. Eventually, one of the big decisions that we had made was to rebrand, and so for exactly the same reason.
So if you look, look across the delivery network, everybody back then, used to be shades of yellow, orange, red. And if you remember, Dunzo used to be this bubblegum yellow, and the brand wasn’t differentiated. So we went extreme right field and we said, we'll go full grunge: black plus green neon.
We did a lot of brand studies - there is all this whole brand archetype chart. So that tells you that for companies who want to be like you, what kind of brand archetype do they identify with. We were coming up to be the friendly neighborhood delivery person, approachable and nice. That was what the persona was, and the brand that we were trying to build was the bang opposite. We wanted to be extremely cool, above the masses, more sassy. Also the type of copies that we were using was very sarcastic if you noticed back then.
So by the end of it all at Dunzo, the frequency had gone up, and the D30 retention also significantly went up.
LazyPay Use cases and Metrics
On LazyPay, originations was a major metric of choice, but retention was a massive issue on LazyPay.
Mithun Madhusudan:
How do you define retention?
Bhavik Kaul:
Then the next transaction or loan.
Bhavik Kaul:
The biggest problem at LazyPay was the use case that the consumer had organically understood. These are people who had very low penetration of credit cards. For people who had credit cards, had a very high revolve rate on those credit cards. So credit availability and enough credit was like a problem with them. There were two major use cases that had come out - lifestyle aspiration and emergency.
So lifestyle aspiration, I'm a bachelor and I earn. Might as well go out, order food, go to movies, get gifts for my girlfriend. And emergency, was the end of month emergency. I get my salary on the first, and I lose track and 20-22nd of the month all my money is over. And now what do I do when I have to order groceries or I have to recharge my phone?
The value prop was LazyPay is very fast, very transparent and requires zero onboarding compared to other loans. Other loans required a lengthy onboarding. We were like you authenticated your mobile number and here is the money. This is magic right? This worked well, but the downside was that because it had a limited use case, and was available on limited merchants the frequency got limited.
The second thing was because we were underwriting a lot of people who had low availability of credit, their credit lines also used to get over. And there was no real way to keep increasing their credit lines because they were fundamentally hard to underwrite. So the retention rate at that point on the platform was about, D30 was about 20 odd percent. Today this has gone up to 40 odd percent.
By the way, this is at a portfolio level. When you do this cross section, you will notice that there will be segments that show D30 retention as high as 80% because everybody gets different credit lines. So I think the right credit line with the right use case works. Also, while this is a different topic - on the retention side, that is exactly why the first literal first product that I built on LazyPay was Credit on UPI.
Credit on UPI LazyPay - increasing both frequency and retention
Because I had to do two things simultaneously. When I joined LazyPay, Covid had hit and the business was getting hit as well. While the business is growing, monetization was a big issue. So I had to get two or three things done. One I had to get rapid rapid growth shown in that year. Second - the only way I could grow sustainably was if the retention number increases, Third I have to find a way that I could also monetize. All this in one year.
So I created this credit on UPI product - which increased use cases because credit becomes ubiquitous. You can use it at a 1000 different places. In fact, the way that users were using credit on UPI when I had built it versus how it's getting used today is fundamentally different.
So there are three modalities of UPI. There is scan and pay, there is an intent flow from other apps and then there is a collect flow. Collect flows are where you enter your VPA and then you have to actually pay. Interestingly, what happened was that for me the first one and the third one had worked really well.
So people had found product market fit in paying at grocery thelas to their house help to petrol pumps. This had become a massive area of our PMF.
The second thing that happened - a bunch of people who didn't have access to credit also could not use credit for buying shoes and other stuff on Amazon, and so they started putting our (LazyPay) VPA there and then paying with credit so they effectively got a credit card. Even though they didn't have a credit card.
Mithun Madhusudan:
So it was a LazyPay based credit line and you’re calling it credit on UPI because he's using his UPI handle.
Bhavik Kaul:
Correct. But what's happened now is that this entire product has become like an add-on to a credit card. So you get a RuPay credit card and you can use it through this modality. The use case now has increased. The acceptance of UPI credit has reduced drastically because now there is an MDR associated. In our case there wasn't any. So a lot of merchants have started shutting down or limiting access to where you can use credit on UPI, and I'm not sure whether you've seen data around it. But that's one of the biggest challenges that products in this case are having, like, a lot of merchants don't enable it. It's not a UPI transaction anymore. It's a credit card transaction where coding is through UPI.
Mithun Madhusudan:
Okay. So do you think since inception all of these guys like guys like Kiwi and whoever is doing it have that MDR.
Bhavik Kaul:
They yeah, they get MDR. I think the bank gets about 2%, and they split whatever. There is one caveat - there is a threshold. So, there’s no MDR below a certain amount. So, people have also because of that reason aggressively started changing the nature of transactions. The majority of my transactions offline were under 2,000 because of these use cases and now they are trying to push a restaurant, they're trying to push organized retail, they're trying to push modern retail grocery. So, if you go to some of them and check where is upgraded getting used now, it will have changed a bit
That UPI credit happened to be one of the punts which actually worked to increase both frequency and retention on the platform. I remember that we were doing a non-UPI credit user had an average credit line of like 3k to 4k, and UPI credit user 25K. Utilization on this side was like about 30, 35%, on this side 60% (UPI on Credit). We also added a feature where you could choose to pay in parts which was like an alternate proxy revolve product, and we saw almost 40, 45% of the outstanding amount on the UPI credit being revolved.
And the, and the loss rates over those were very minimal, to the extent that the third thing that I had to solve, profitability, got solved. So the BNPL business by definition does not make a lot of money in India because the MDR that you get from merchants is just enough to pay for your CAC and losses. So it just breaks even in the best case, but now with UPI credit it became like a massively profitable business.
Mithun Madhusudan:
But I think you were saying here that you had to shut it down or something, if I remember.
Bhavik Kaul:
Yeah, this is because, eventually, NPCI came back and said that, you can only issue VPAs if they are interoperable.
And we had issued VPAs which were only operable for LazyPay. And there was no actual way to do it. Although we sat with NPCI. In that time, if you remember, Jupiter had also launched Bullet, and then, they were massively marketing it, and this was like much after our LazyPay product had reached our scale. What happened is that the RBI came back and said that you cannot load credit lines on a PPI (pre paid instrument) product.
The circular said you can load a PPI with another PPI, with a credit card, with a bank account, just not a credit line.
So, it was crazy. I think everyone was doing extremely well. I think the credit cards that was also like the time when I had also launched our card, Slice card, Uni card. I think everyone was doing card originations, and that was becoming larger than what some banks were doing. I remember that I compared my, then, quarter, numbers with IDFCs numbers, and I was doing almost 75% of what IDFCt was doing. Slice had beaten IDFC, easily, and a couple of more banks. It was a crazy time, and all of our business, by the way, we were not built on poor mathematics. They were all fairly profitable. Although the break even was like about two to three years. So, it was rightly constructed.
I think that was the metric stuff that you were talking about. At both Dunzo and PayU, there were topline growth metrics, and, retention metrics, which I was tracking. At Dunzo, the only time that I spent a lot of time was activation.
Framework to increase Retention: Organic Frequency and Need Adjacency
Mithun Madhusudan:
I think one theme that is coming out at both DunZo and LazyPay - one horizontal way of thinking about adding frequency and retention is adding more use cases. And in both cases it's categories, right? For Dunzo, it is like different things you can do, and pick-up and drop.
Bhavik Kaul:
So, basically, my framework has been this: number one, you understand the organic frequency of the need of the product or service that you're trying to solve. Because in 99% of cases, you will not be able to change the organic frequency.
Second, you understand if the need is isolated. Or if there are adjoining needs, which may not be very clear at the onset, but, you know they exist. For example, when I was building the grocery dark store business at Dunzo, what I realized was that if I step out and I talk to several users, if users go to buy say atta, chini, dal, they will eventually also want to buy fresh groceries. And interestingly by sheer probability, they started, finding places where you can find both. And so, the option of doing only one is out.
And this, I think, BigBasket had realized very, very long ago. In fact, BB, from day one, was trying to crack the vertical supply chain on fresh grocery, for exactly this reason. Grofers had tried to do the opposite. They had tried to become only staples. No fresh grocery at all.For about two three years, Grofers was just this, they had become like the APMC market of Staples. And eventually, they also realize that people look for these baskets,, and then they obviously pivoted for several reasons. One of them was this guy - they also wanted to sell fresh groceries.
So the second point is understanding is the need isolated or there are adjacent several needs.
The third point is understanding what factors could influence the ICP to change their frequency either on the core need or on the adjacent need.
So, for example, take the insurance business, it's like a yearly business. You sell an insurance what will make the user come back again? But adjacent to the insurance business there are other businesses, for example, take motor insurance. You get your car repaired several times in the year. You buy car accessories. And maybe you require other stuff - challans, fast tag recharge. This is an interesting hack to get retention and frequency, and the day that you have to get your insurance done. Actually insurance is the product where you have the highest margin of like 30, 40% margin on the premiums. You'll be like I’ll buy insurance from this guy only.
Mithun Madhusudan:
Interesting that you talk about insurance, because this is what I think CRED is doing, I love their whole garage, or whatever thing is.
Bhavik Kaul:
All of them are doing it now. I don't know whether a lot of people may not have designed it as well as CRED. But Acko now has this offline set of garages. Do you know this?
Mithun Madhusudan:
No, no.
Bhavik Kaul:
They’re selling used cars online on their app. They have all the features that CRED also has, challan payment, a lot of other, tracking, tracking your insurance, and now they're created a bunch of actual Acko garages where you can send your car for repair. Because they see this is how they're going to both make money. But also, this is in the same ecosystem of needs.
Vibhav V:
Basically, don't fight the organic demand and then you have adjacencies that you put together which will help in retention and frequency.
Bhavik Kaul:
Yeah. So in the credit business there is a Golden Rule, which everybody has been pitching VCs is: acquire people from BNPL, uplevel them to a credit card, or a UPI credit business, and then sell them loans. Because, clearly, selling them loans is the hardest because everybody's trying to do that. To sell loans directly from BNPL is very hard because the need is very different, and you don't have enough data to underwrite. So, until you do this need adjacency in a nice staggered way, you don't like a large valuable profitable business.
Mithun Madhusudan:
But there are a lot of players who've gone into direct credit cards. All of these new age vertical credit cards. Scapia is probably a good example?
Bhavik Kaul:
There will always be these very specific niche use cases, which will sell anyways, because Scapia credit card is selling on one value proposition 0 forex, and their entire business model changes. But you also notice that - Why are you trying to become make my trip plus Credit Card, when your credit card business is large enough itself? It's because of these adjacencies. What will make people use this card repeatedly? Because the organic frequency of the businesses, once a quarter, or half a year - twice a year, or three times a year.
———-———-———-———-———-———-———-———-———-———-———-———-
That’s it for Part 1. The next post will cover softer aspects key PM skills, how to evaluate PM candidates, characteristics of great PMs, getting promoted and quitting a company, and building influence as a PM.
References:
(1) Types of networks: https://a16z.com/hidden-networks-network-effects-that-dont-look-like-network-effects/
(2) The most important consumer metrics: www.lennysnewsletter.com/p/the-most-important-consumer-metrics
(3) NPCI credit card regulations: https://razorpay.com/blog/npci-circular-on-upi-interoperability/
(4) Hike - The Indian Messenger
While I was reading this, my first thought was ‘hmm it’s like Lenny’s newsletter’ and there was truly nothing like that in Indian context. Interestingly the about section states the same thing :)
Took sometime to go through this. Really appreciate Bhavik’s insights and authenticity