F2S: Edge of tomorrow
Backstory
ClearTax is one of the most ubiquitous financial platforms in the country. Over 3 million people file their taxes through our free tax filing tool every year. However, tax filing is not our only consumer offering. ClearTax invest is our mutual fund offering, which offers expert-curated packages of mutual funds tailored to a user’s needs. ClearTax invest offers a wide array of mutual funds from 30+ mutual fund houses at one place.
The opportunity
In the tax filing tool, after a return has been completely compiled, it needs to be submitted to the income tax server of the government of India. This process of submission to the server, filing the tax return and fetching back a status message from the Income-tax department server takes roughly anywhere from 3–6 minutes toward the end of the season, where the servers are particularly loaded.
During this time, we chose to experiment cross-selling curated mutual fund packages to users based on their specific needs. This upsell of mutual fund packages is referred to as F2S.
F2S 2018
During the tax filing season of 2018, the F2S flow primarily had two variants:
- For people who did not max out their section 80C investments, we recommended ELSS or tax saving mutual fund packages.
2. For people who maxed out their section 80C investments, we recommended fund packages aimed at wealth growth in general.
Both the stories showed up for relevant people while they were waiting for their tax return to be filed. This design worked on some basic principles of simple imagery, hyperinflation of the timeline to draw big numbers in the order or crores, simple bite-sized messaging on each screen. The final stage conversion, i.e, number of investments through the flow stood at 1%. Although this seems like a measly number, in the context of the number of people who file their taxes through our service and the critical juncture of time at which the up-sell is done, it is a decent number.
F2S 2019
The aim for 2019 was to better the final stage conversion, which stood at 1%. However, due to forces outside our control, we were already on the back step before we even started.
Due to a major market correction, the numbers on the returns of almost all mutual fund packages plummeted. For perspective, the fund package that sold the most in 2018, had a CAGR (Compound annual growth rate) of 23% in July — August 2018. Most of the funds were hovering around the 20% mark. However, the same package fell to 9.2% in July-Aug 2019. The average for most packages stood around 8%. While in reality, as seasoned investors will tell you, is a perfect time to invest, it does not make an attractive case to the newbie investor, which happens to be 80% of the market.
Furthermore, we observed that one’s peer network and the performance of their funds are often the most influential factors in driving a prospective investor over the line. And this, for many existing investors plunged into the red, which tends to scare away the new investors.
The process
We had a limited timeline of one month for this project, which mean that we had to have a complete release cycle almost every 3 days. For such a fast-paced project with a limited period of effect, we relied heavily on three things:
1. Metric feedback
An end to end funnel was created to monitor the click-through rate of each step in the process. In addition, we relied on heatmaps for each step of the funnel to analyse where users were looking at. We had a constant influx of these metrics thanks to metabase and hotjar. Occasionally, we also set up polls through hotjar. These proved to be invaluable resources to quickly identify issues, monitor efficacy and take calculated calls on the designs we rolled out.
2. Quick qualitative feedback
As a part of the tax filing NPS monitoring plan, we rolled out an additional opt-in for a small 15-minute user research session pertaining to the tax filing flow and its constituents. A simple reward system of coupons was set up to users. This ensured we made accurate qualitative interpretation of the data we are gathering. This double verification of problems through data and user stories ensured we kept working on the right problem and not the wrong one.
3. Constant iterations, A/B
Constant iterations were at the core of this process; at all levels. At a design and conceptual level, we ensured a constant flow of iterations, tested them out with users before rolling them out into production. On the tech side, we enabled infrastructure to support iterations and A/B tests on multiple levels. At any time, we could test up to three concepts of stories for the F2S module. In addition to this, we personalised the stories to the user segment by understanding their finances and tailoring recommendations accordingly. As we discovered from parallel research we conducted, people had different affinities to investment and these were mostly a factor of their income range and their age group. We segmented users accordingly and personalised stories for them.
Edge of tomorrow
Does anyone remember the Tom Cruise movie Edge of tomorrow? The one where he wakes up every day in the same scenario and makes slight progress on his mission? The whole process of this project felt exactly like that. We wake up, crunch the numbers, talk to users, make sense of them, iterate continuously, test concepts with users, roll them out for production and repeat the process; and push the needle by a small bit. Since the benchmark for us was to beat last year’s number, we set up the initial A/B against the F2S 2018 design.
User research and Insights
Through a user research session that was conducted before the whole process began, we identified a set of hypotheses that could work as potential storylines for this whole exercise.
Before going into every new version of F2S, we conducted user research session as mentioned earlier. Since the newbie investors/people who have not invested yet constituted 93% of our audience, the main aim was to optimise these flows and stories to these people in particular. The aim of the research was to find out both; what worked in a variant and what did not. So, we reached out to 5 users who had converted and 5 users who did not and asked them to re-run us through the F2S experience with think-aloud protocol. After starting from this exercise, we noted down peculiar behaviours and noted down the parts of messaging that worked, that failed and that did not come across in a lucid manner. This gave us an actionable list of stories we could iterate on for the period of one month.
Furthermore, it doesn’t take long to notice that the 2018 version of F2S had way too many clicks. This increases the number of pages or layers in the conversion funnel, which directly correlates to higher drop-offs from these different levels. The complete funnel could be broken down into two parts: The story funnel and the transaction funnel. For reference, look at the videos above.
Problem 1: Long Story funnel
The story funnel was mainly having the sales pitch to the user to invest. Currently, It had two layers in itself and 4 levels. 3 screens were being used to drive the point home and redirect the user to the fourth page: the fund page on the ClearTax website. There can be one reduction done on this front.
Problem 2: Long transaction funnel
The transaction funnel too has 4 different screens. In the limited amount of time the user has to go through the flow, this is too many screens for anyone’s liking. This, coupled with the anxiety of filing status invariably lead to a lot of drop off.
The form fields being filled in the transaction funnel are all data points that are already being collected in the filing flow. There is no reason these cannot be removed completely by simply pre-filling the same from the tax filing flow.
Problem 3: The fund page
Another major drop-off point for us was the fund page. Sure, everything before the transaction is exploratory and doesn’t require the user to mentally make a decision. But the fund page itself showed on the user study that it confused the user instead of guiding him/her. But this required extensive work that was beyond the scope of the timeline afforded to the project. So, we decided to do the following two things
1. Try and remove the fund page from the F2S flow
This was possible but required a week’s work in development. So, we started working on the possibility of removing the fund page from the flow.
2. Re-Design the Fund page to suit the needs of the user
This became a separate project altogether that I took up post the F2S season. More on that will be updated here soon.
Problem 4: Anxiety of filing
There was one constant insight into the user research from those who dropped off early in the funnel: Users did not like moving away from the screen when their filing was being done. This was a known fact to us, but its effect was far more than was initially predicted.
This meant that we needed to ensure the complete process had to be completed on the same screen and the design needed to facilitate this.
Practical Limitations and timeline tweaking
These were the effective high-level tasks for us to achieve during the F2S period:
- Remove the fund page from the flow
- Shorten the story funnel
- Shorten the transaction funnel by pre-filling user details in the checkout flow
- Enable checkout on F2S screen.
However, action items 1,3 and 4 from the list above required us to re-engineer our backend quite a bit, which would roughly take 2–2.5 weeks’ time. Till those couple of weeks ensued, the first couple of iterations of the F2S could just be on the story and not much else.
F2S 2019 V1: The minimalist approach
The first was that for the amateur investor, which accounted for around 90% of the market share, the flow did not make sense the story had gaps in between. For instance, the users were unable to figure out how they could save more taxes by investing in tax saving mutual funds. These were people for whom the basics of how section 80C worked was not clear. Added to that, there was no thread which transparently showed them the entire calculation by which they could have saved/made a certain amount of money.
Another interesting aspect of investment we discovered during the user research was that the most important factor in making the decision to invest was that of social proof from one’s immediate peer network. In the absence of the immediate peer network, we noticed that experts in the financial field were the next best at dragging users over the line of their first investment. Both these powerful aspects needed to be productised and built into the flow.
To work around these problems, we designed a single screen story to the user with transparent and complete calculations. Added to that, we added some expert validation in the screen while the rest of the transaction flow remained similar; for the time being.
To have a reference point, we pitched this version against the last year’s version and waited for numbers.
As an emergency measure, we rolled out the 2018 variant to 90% of the traffic and the 2019 version to 10%.
Insights
This was a crude reality check for us. Our hypothesis proved to be way off the mark. We immediately set out to find out why. A quick study pointed out three things:
- Users thinking that they would hamper their filing flow if they clicked on the CTR.
- The “Invest now” button scared the users to know more about the fund. They had an assumption that they would directly checkout once they clicked on it.
- Not enough data to generate trust.
- The screens being loaded with text and hence being ignored by the user.
Since it was still early days, we were getting quite low traffic on the income tax filing flow. So, we decided to take up the first three observations and iterate with them. Regarding the final insight, we decided to pursue that observation one more time. The assumption was that less traffic might also impact this data point.
F2S 2019 V2: The two-screen approach
By this time, the second week of July, we had some aspects of being able to check out in place. So, apart from the aforementioned three inputs carried over from the last research dump, we created a design that allowed us to have a CTA that would take us directly to the transactional flow, cutting out the fund page in the process.
V2, Wealth growth variant
Please click on images to enlarge.
V2, Tax saving variant
Please click on images to enlarge
Insights
Another shocker. But this confirmed that the data-heavy approach does not work. However, the positives were that the softer CTAs worked better than a direct, harder CTA. With this, we decided to immediately ditch the data-heavy approach and decided to go back to light storytelling with one-liner messaging and a single image. Another insight was that the social proof messaging was simply not working.
F2S V3: Back to basics
With the F2S 2018 miles ahead of both the variants we had made, we decided to adopt the first screen of the same and proceed ahead with changes in messaging after that. We also went ahead and inserted a complete screen just for social proof.
Please click on the images to enlarge
Since the first screen was same for both variants, we monitored the final screen CTRs. For the first time, we started the beat the last year’s version. But we barely had two weeks’ time left to do whatever we had to do. The social proofing screen was working and adding to the conversion of the page.
Post this observation, we changed the ratios to 50% each.
F2S 2019 V4: The checkout experience
By this time, we redesigned the F2S page to eliminate all possible distractions and optimised the layout to suit the flow for better conversion. We added much more information about a fund to generate trust in the added real estate we conjured due to the revamped layout of the page.
We now had the technical framework we wanted in place and we were ready to play as we wanted. We further floated a plan for users to start with a minimal investment of 500 rupees. In addition to this, on the final checkout screen, we removed other distractions of referrals etc and introduced some graph elements to see if they impacted user trust.
The new layout gave us a fat bump in the conversion numbers. This was further proof that we were heading in the right direction. The simplified layout and not allowing the users to jump out of this flow and into a new tab was working.
F2S V5: Goal-based approach
With all the other aspects in place, now it was purely down to the kind of storytelling that was done on these pages. We needed to iterate on stories and needed hooks for drawing users in at the first screen.
In one of the research sessions, we found out that a lot of users had no innate desire to invest in mutual fund packages. We also found out that working backwards from life and financial goals worked in drawing in users, as this kind of financial planning was non-existent and appealed to a majority of the users.
Here is the goal-based flow:
The goal-based approach gave us a considerable amount of increase in the first screen conversion. This was at a time we were doing more than 50,000 filers in a single day. The innocuous entry point, combined with the interactive element, that gave a sense of personalisation of financial plan to the users was working.
F2S V6: Advisory
By this time, we had completed the mammoth user research session that was running in parallel to the F2S ordeal. We learnt a great deal about the segmentation of users and what affected them the most. Personalisation, as shown in the previous example proved to be a powerful tool to motivate people to walk over that invisible line of making your first investment.
In parallel, we had a hypothesis of investment being easier when being accompanied by advisory from an expert. This was a completely new vertical that we were going to launch post the F2S season and the F2S module provided us with a perfect opportunity to test this product idea with the consumers at scale.
In V6, we offered a personalised plan based on dependents, age, income and major financial necessity. In addition, there was the pull of extra validation that came from this being an advisory flow.
The advisory based flow was a huge success by the numbers. It gave us a major bump in numbers. This was at a time when we were doing over 80,000 filers a day. In context, it was a huge win in multiple ways. It had also validated the concept of an advisory or a recommendation based approach to investing. This was in the last few days of the season and the season ended in 3 days after this variant.
Final Numbers
Overall, we more than quadrupled the conversion rate of F2S over the course of the season.
Learnings
1. Numbers do not lie
Even having worked with metrics multiple times before never prepped me for what I experienced here. This was sprints on steroids. The constant monitoring of numbers at each step of the funnel, on each day helped us race faster than we ever could.
The numbers were our holy grail when it came to validating stories, interaction patterns, pages and even copy. I have not covered all of the cases of iterations and A/B tests here as the list is quite huge. There were a bunch of cases of bugs and failures that we zeroed in on no time because of the numbers.
The faster you monitor, the faster you know. The faster you know, the faster you inquire. The faster you inquire, the faster you fix. The faster you fix, the faster you grow.
2. Slice and Dice
Slicing and dicing designs based on daily feedback of numbers and users was extremely new to me. Often times, the solution lies in between solutions and presents itself as an unexpected stitch together of multiple solutions.
Without the numbers, it would have taken ages and a lot of analysis paralysis to realise which solution works better at what step. But with the constant monitoring of data, we found small pieces that individually worked better at different stages and stitched them together in no time to get the most out of the system.
3. Have your tools ready
A big drawback of this exercise was the lack of an exclusive graphic designer for the process. And having joined a couple of days prior to the season kick-off, I had no hand in at a ready-made visual inventory. This was a major feedback throughout the season. We have now conjured up a complete visual inventory, modularised it and are locked and loaded for the next season.
We also struggled a bit with having the right tech and data architecture in place to be ready for the season. A lot of things broke in between, but we finally managed to build an all-round system that works for such a fast-paced environment
4. Storytelling is king
In the beginning, we started off with more data which we assumed was the way to go. The entire team was sure of being transparent and comprehensive was the way to go. But this assumption was shattered by the users. Bite-sized chunks that form a coherent narrative have given us the most juice. The stronger and crisper your story, better the results.
5. A/Bs are the A B Cs of product
During the course of this whole exercise, we pitched all sorts of things against each other. Stories, copy, imagery, layouts, designs, numbers; you name it. Consequently, we have now defined guidelines for each of this thanks to the numbers thrown into our face. Test as much as you can, learn as much as you can.
6. Observe and you shall know
We managed to find winning storylines in the end just because we listened to users. Not to their voices, but to their unsaid needs that usually lay beyond their voices. If it weren’t for the research we conducted to identify what works and what doesn’t, a lot of stories that we experimented on would never have been stumbled upon.
Epilogue
The F2S Module was by far the most intense sprint of work that happened. The cycles were insanely short, the data extremely crisp, the people extremely agile and the numbers, unabashedly truthful. I had a wonderful time working on this project and it was a lot of people that made this project truly special. Design-wise, it might not be the best project; but learnings wise, it sure is.
This project was done under the guidance of Akshay Bharadwaj and in conjunction with Aditya Raghunath, Medha Seth, Mayank Abhijeet, Vikas Verma and Aravind Bachu. Shout out to you all!
