The 60x Year — Product Led Growth at Atlys
I was having a conversation with Evan Sharp (Co-founder of Pinterest & Atlys Investor), when he said “There’s no one right way to build a company.”
Not all companies need to build product as their moat. Brian Chesky really cared about the Airbnb brand, Steve Jobs was obsessed with design. For Mohak [founder of Atlys] and I, it was product. In fact, this was one of the reasons I joined Atlys.
Over the past year, we grew 60x in visas at Atlys; product led growth as you call it. But what is product led growth? I used to hear this time term a lot but now, I have a better idea of what it means.
Product led growth shows up in different ways — some subtle and some obvious. It’s when everyone in the company talks about product, when lunch conversions are surrounding product. It’s when growth is driven by product features and launches, when the culture is shaped in a way that everyone looks to product when growth is slow.
In an obvious way, it’s when product is used as a lever to solve broad problems. Let me give you an example. Trust has been a huge problem statement for us at Atlys for many reasons:
- India is a low trust society.
- Visas have high stakes: travel plans, which include high value purchases such as flight tickets & hotel bookings, are dependent on visas coming on time.
- The visa process involves sharing sensitive information such as passport information, bank statements and ITRs.
There are several ways to solve for this problem statement. The branding/content mind would say “Let’s get Alia Bhatt or another celebrity as the face of the brand”. A distribution mind would say “Let’s partner with Etihad.” I mostly thought of this problem through a product lens. How can product solve for trust?
The best consumer products are not products that are used daily [airbnb is an infrequent product]. They are also not products that go viral [clubhouse went viral]. In my view, the best products are those that people use over sustained periods and refer their friends and family.
This is the Atlys story so far: our product philosophy, the wins and the mistakes we made. If you looking for ways to scale from zero to ten through product, I hope this sheds some light on what to do and what to avoid.
It’s mind-boggling to think that just over a year ago, we were working out of a coffee shop in Delhi, speaking to our first customers on the phone. One thing in all of us was an innate desire to solve for the customer, inspired by all the products we loved and did not make.
Content
(I) Our Path to Product-Market Fit
True product market fit occurs when you choose the right value proposition for your target customer. But how do you figure out the right acquisitive value proposition? We had the following framework in mind to figure out our value proposition:
- Singular
One value prop reduces cognitive load and gives users clarity on why they should use your product or service. Singular value propositions also help in word of mouth and virality because existing users find it easier to communicate the product to their friends and family all users communicating the same thing helps in magnifying the communication of the value proposition.
Take Zepto for example, their value prop was not “low prices, online groceries, huge assortment, and fast delivery.” They focused on one thing: 10-minute grocery delivery. Users used Zepto for that proposition and only spoke about the 10-minute delivery when they spoke about Zepto.
2. Focus on the pain for which people will change their behaviour
Two opposing forces work together in product adoption: Distaste to change your behaviour [switching cost] & pain in current behaviour. Product adoption seems to happen when using your product, the (reduction in pain) > the (distaste in changing behaviour).
Before PMF we iterated on several value props: ease of use, speed in completing the visa process, and price transparency. We realised since we were an infrequent process, people didn’t care about these things, as they were okay with spending a couple more hours, doing an inefficient process once in 3 months. The “pain” wasn’t strong enough in the context of the frequency of the problem for users to change their behaviour.
As we spent more time speaking to our customers, we discovered the strongest pain point in the visa process: anxiety. People were anxious about when their visas would arrive, and what was happening with their visas, since they had travel plans contingent on it. What’s on the complete opposite spectrum of anxiety: extreme certainty. We had our winning value prop — our on-time guarantee. We promised a specific timeframe by the minute and offered a money-back guarantee if we didn’t deliver on time. And boom — it worked.
3. Avoid Noticed Value Props
Some value propositions are not acquisitive and rather noticed. They are hard to communicate and customers don’t trust them. An example of noticed value props that we tried was “hassle-free”. But what does that really mean? Now, once users used our product they realised that it was hassle-free. But as a value prop to acquire users, it never worked. It’s the same as Lyft having a value prop of “Better Drivers”. Sure, Lyft might have more hospitable and friendly drivers, but that’s something users will notice when they use the product.
Figure out your value prop before continuing to build more and more features. As founders, we are all optimists and love to build product. It is easy to fall into the next feature fallacy, where you feel the next feature would make the product good enough to get users. Not working, “adding search will make users use it”. Still not working “Let’s add a referral program”.
This rarely works. For us at Atlys, when we iterated on the right value prop, it just worked. We had no search, bare minimum home page, no reviews, no docs required, no in-house OCRs for passport scanning, and no group application feature. And yet, end of first week after launch, we were doing 11 visas a day.
(II) Product Starts with User Insights
User insights is a popular term in product. But what is a user insight? It’s simply a deep understanding of users. The better you understand users in the context of your business, the better product you can build since you’re able to better address their needs. We built an insight engine to discover insights in the following ways:
- Analysing user behaviour data
- Start with a Hypothesis, Then Look at the Data
- Customer calls
Analyse user behaviour [1]
As you build and start looking at data to analyse user behaviour, you will start to see that users will naturally solve some problems for themselves.
We saw a large percentage of our users doing several visas in the same month. As we looked deeper, we observed it wasn’t because they travelled to multiple countries in the same month, but rather were travelling with a group & applied for the entire group themselves. We knew that Indians travel internationally in groups [either with friends or family], but what was interesting is that often one traveller applies for all.
This user behaviour was so strong that users engaged in behaviour the platform doesn’t natively support. This insight helped us build our group application flow, making it easier for users to do so.
Even Twitter, in its early days, saw users numbering their tweets (1,2,3,4..) as a workaround for sharing longer thoughts or stories that exceeded the character limit. This user-driven behaviour eventually led to the creation of the now-ubiquitous Twitter threads feature.
Start with a Hypothesis, Then Look at the Data
Before diving into your product funnel data, formulate a hypothesis about what you expect to see. Look for discrepancies between your relative assumptions and the actual user behaviour. These gaps often lead to valuable insights that can significantly improve your product. Customers in our funnel have to sign up, add their travel dates, upload passport and selfie photos. We hypothesised that the passport step would have the highest drop off.
Data, however, told a completely different story. Travel dates was the highest drop-off point — a huge discrepancy in our hypothesis. When I spoke with customers, I realised that the anxiety around visas was so high that users wanted to get the visa and only then plan their travel. Based on this insight, we launched product features that exponentially increased our conversion rates.
(III) First Principles Thinking on User Insights
Once you have any insight, think from first principles on why that insight exists. We had an insight that users travel in groups and usually, there is a main traveller who applies for the entire group. The first thing we asked was — why?
When it comes to purchasing things online, humans tend to do it for themselves. This is partly because humans have different preferences and needs that occur at different points in time. With regards to purchasing things in groups, humans purchase experiences in groups. This is because humans are inherently social beings, and like to have experiences with their friends and loved ones.
- Movies (experience): People purchase movie tickets in groups.
- Sports (experience): People book tennis courts, and cricket grounds.
- Travel (experience): Atlys
- Concerts (experience): Friends or family members buy tickets together to attend a live concert or music festival.
In all these cases, there is usually one person in the group who is the leader — and books the experience for everyone. This is because:
- Humans prefer organising themselves this way
- Convenience because the information remains the same for everyone: Date/timing in the case of movie, flight details in the case of travel etc.
- Better Coordination among group members
First principles thinking is extremely hard to do and you can think “from first principles on why that is”. It’s because it’s unnatural. We learn by emulating other people — [everything] from learning how to walk, learning how to talk, comes from copying others and modelling others.
Thinking of the group experience from first principles allowed us to create product journeys catering to natural human behaviour, helped us figure out from where we take inspiration [bookmyshow, StubHub] and better address user needs.
(IV) The Best Way to Improve a Funnel Step - Remove It
Humans today have low attention spans & have an extreme distaste for cognitive load. Cognitive load is the amount of mental effort to complete a task. Aim to minimise the cognitive load at each funnel step.
- ask a user to click on something, you lose 10% of them.
- ask a user to pick between two options, you lose 20% of them.
- ask a user to pick to input something, you lose 40% of them.
The highest conversion rate of a funnel step is removing that funnel step. I always asked myself one question: can we do without it?
- Reduce the Number of Steps [2]
We had a screen that showed progress after each step in the visa application process. So 4 steps means 4 different screens for the user. What did we do? We pushed the progress bar to the top, eliminating the need for 4 different screens and reducing the number of steps in the funnel.
2. When you can’t reduce steps, minimise cognitive load at each step
There are several ways to reduce cognitive load — use smart defaults, guide users through multiple options through social proof, minimise clicks and scroll with optimal placement of features.
We became so obsessed with this that when users put their number on our login page, we auto-clicked continue so they didn’t have to [98%+ users put in the correct number].
Think about this: If you want to ask someone out on a date, what’s better?
- Are you free to eat dinner at Big Chill after 8 pm tomorrow?
- Are you free to grab a meal sometime?
It’s the first one because it significantly reduces cognitive load for the other person to say yes. Try it.
The 80–20 [Pareto principle] rule is my biggest weapon when it comes to reducing cognitive load. We had a multi-country feature where users could apply for multiple countries in the same application. The insight here was that for certain countries, people planned trips to multiple countries together.
The initial screen had high cognitive load: a mental effort to go through multiple options, search function, multiple CTAs. The Pareto rule helped in reducing cognitive load. 90% of users selected between 3 countries, around 60% of users selected the same country, which was closest to the destination.
We decreased cognitive load by a) reducing options to 3 b) aiding decision making by the “closest to X” tag and c) removing unnecessary CTAs and copy
Just by reducing cognitive load, we increased % of people adding another country from 3% to 6.8%.
(V) User Journey — The Need of the Moment
Coming back to the core of building great products — insights. To have insights, you need to know what’s going on in the users mind; and users have different things on their minds at different points in time in their journey.
Let’s take the example of going on a date again. When you meet your potential partner for the first time, you’re thinking “I hope this isn’t awkward”. As the date proceeds, you’re thinking different things — it could be “This person is not my vibe” or “This person is really interesting”. If things are going well, you become more comfortable with the person.
Similarly, users had different things on their minds across the entire product journey. To understand this, the first step is drawing out [mood-board] the entire user journey; and the user journey doesn’t start when users land your product, it starts before that.
Thinking about what’s going on in the user’s mind at each step helped us figure out some of our key insights, that helped us in our zero-to-one journey. Having context of the user journey is also fundamental to figuring out the placement of features. New users arriving on your product page/onboarding page will have very different needs, emotions, and questions compared to users who have already transacted with you or completed a core action in your funnel.
VI) Metrics — Your Best Friend Who Can Be Your Worst Enemy
- Core Action
Figure out a core action for your product that takes precedence over every other metric. As Sarah Tavel describes core action, “The action that is the very foundation and essence of your product. Pinterest would not exist without pinning. Twitter would not exist without tweeting.”
Why is it so important?
Startup = Growth; and the unit of growth is time. In essence, startups are a race against time. As Paul Graham says, “Speed defines startups. Focus enables speed.”
Now, choosing the wrong metrics makes you lose time by making you focus on the wrong things in the long term. We made this mistake too. For a brief period, we were chasing the “sign up” metric blindly, which resulted in us launching a feature of a fixed bottom sign up modal on a certain scroll depth. It did increase sign ups, but ended up being a poorer user experience, without contributing to our core action of submitting a visa application.
We made this mistake again. To improve our passport conversion step, we added an option to skip/do it later. This obviously improved the conversion of the passport step, but again did not drive our core action of submitting a visa application.
This is not to say don’t focus on any other metric. Several supporting actions are needed to happen to complete the core action. These metrics, should always however, impact your core action metric.
2. Defining Success Metrics Not Related to Core Action
We had two insights based on customer feedback.
a) Users did not trust our platform & b) Users were highly anxious between the time they applied and the time their visa arrived.
Thus we had the following goals: Increase Trust and Reduce Anxiety. But how do we define these?
To figure out the success metric of these, we figured out how the user behaviour [metric] would change if we achieved these goals of increasing trust and reducing anxiety.
Increase Trust
What metric changes when trust increases?
To figure this out, I looked at the cohort of users that had high trust. Users coming from word of mouth have higher trust compared to the users that came from ads; and what we saw was fascinating: both the users had very similar conversion rates throughout the onboarding/document uploading funnel — the major difference was in the % of people who entered our funnel from our country landing pages. We had our metric: conversion rates from country landing pages to entering the funnel. If we were able to build trust, that conversion metric would increase.
Reducing Anxiety
What metric changes when anxiety increases?
This time, I looked at the cohort of users with high anxiety — the ones whose visas got delayed. What behaviour changed when we delayed visas? We saw it was the page-views of the visa tracking screen. When we delayed visas, app tracking screen page-views shot up instantly. When you’re anxious, you tend to open the app more often; We got our metric.
Reduced Anxiety means users opened the app less frequently and over time we aimed to reduce this — with extreme transparency, and notifications.
In the end, I feel what worked for us was that we were opinionated. It’s more important to have your own framework and opinions than to follow the framework we followed. We could also build an equally good product by gamifying the visa experience, similar to CRED. But we had an opinion and stuck to it: People strongly disliked the visa experience. The best visa experience is no visa experience, and our job was to get as close to that as possible.
Notes
[1] Look at user retention cohorts to discover insights: Since we are an infrequent product, retention wasn’t our focus at the start. If you are building a frequent product, figuring out what cohorts of users retain helps you figure out what they value the most. Zepto realised that users who were receiving their orders within 10 minutes (just because of proximity to the grocery store) had a much higher retention compared to those who had higher delivery times. They realised what users valued by looking at what retains them.
[2] At Duolingo, product team wanted to remove the onboarding step of “where did you hear about us” to reduce the funnel steps, but kept it because the marketing team insisted on it.
