Manhattan has some of the most expensive real estate anywhere in the world. Prices in the Big Apple average over $1,700 per square foot. Closer home, a house in HSR Layout in Bangalore gets a little more affordable at around $80 per square foot.
Real estate is a wild game. But if you’re glued in with all the land grab happening in virtual real estate (read: metaverse), you’ll be all too familiar with app real estate. Teams fight for a better share of real estate on the main app.
And when 5% of Indian households shop with you, Meesho’s ‘app’ real estate is as hot as they come 😉
Ask any Product Designer at Meesho. Virtual real estate is a fight to the hilt 😜
But how important is real estate on our app? 😉
This is a story where an inch-long feature increased our orders per visitor by 3.5%!
You can’t put a price on that 😝
Two months. Three People. Millions earned.
Nearly 70% of our orders come from Tier 3/4 markets. We’re the only Indian e-commerce company with such deep penetration in a core market.
Our User Research led us to an interesting insight. As plain sight as this sounds, 'Visual cues' are far better for our unique demographic than text-based ones. Our goal was to reduce text-based search dependencies for our users.
'Sort', for example, another ‘search’ filter remains underutilised on the app.
Other apps also have ‘Filter’ as an option — but these text-based solutions are not as natural for our audience.
Rather than text, a visual element is used to filter out search results. Such filters are called High Visibility Filters (HVF).
No saunters, only sprints 🏃
It took around 4 sprints — each one being 15 days — for our engineering team to create and implement the feature onto our app.
The main problem was narrowing down the applicable filters, and then automating their appearance on the screen.
We started with three broad filters:
- Category (kurtis, shoes, cricket equipment)
- Price (under 99, under 499)
In addition to these, we created hierarchical filters to aid our users:
- Category-invariant filters (new arrival, next day dispatch)
- Taxonomy filters (cotton/silk, for 5 year-olds, plus-sized)
10 such filters are auto-populated as and when a new feed is created. Ten slots and 150+ applicable filters on any feed? 🤯
How did we arrive at the top 10 filters?
We built a relevance model that shortlists filters based on their relevance score, computed based on the user's previous browsing and order behaviour.
The shortlisted filters are shuffled at the user level, where every user sees a unique set of filters. The filters are then ranked and fixed depending on their performance on the feed.
The first two buttons are category-invariant — New Arrivals and High Quality, followed by a shuffled mix of all the other filter types.
Button up, folks!
One of the main tasks while executing this project was planning the content of the buttons: how would they look on the app?
While planning for the filters, we had to plan the content of the buttons. There was a lot of tinkering, but here's how we reached the final placement:
The first decision was to keep four and a half buttons on any screen. Notice the last button split into two?
Wondering why the half? 🤔
If we’d attached a whole number, say four or five buttons, a user might not have realised there are more options. The 'half' button is a way to scroll further, thus further discovery.
Finally, we had to decide the images that would go with every filter.
Keeping in line with the user research, we wanted to enable the usage of the HVF even if the user doesn’t read the text underneath. We selected appropriate images for all categories such as cotton and kurtis, and were ready to go.
One inch, many millions
Testing period: 2 weeks ⏳
We immediately observed a surge of 3.5% in orders per visitor!
We were ready to rumble 🥊
We rolled out the feature to 100% of our users in December — and it was a rousing success!
Fun fact: After the introduction of HVFs, we’ve noticed a surge in adoption of filters by 80%!
But how does the adoption of filters help us?
Applying a filter makes the search result more relevant for the user. It makes it more likely to convert — that is, add to cart, and ultimately purchase the product.
This seems like a simple, small feature. And it is. But its bottom line contributions to Meesho are immense. We run constant experiments to get better at our work. We love doing fun things that help define and change e-commerce behaviours.
Want a playground to dabble with?
Well, we’re hiring! Join us — check out meesho.io for our open positions! ☺️
Btw... Here’s the list of people who made HVFs a reality:
Engineering: Aayush Singla, Siddharth Bulia, Sandeep Saini, Ujwal M
Product & Design: Richa Patel, Gagan Mahajan