Leveraging Data: An Ace Up OLX’s Sleeve
Akanksha Dhamija, Director of Customer Centricity and Analytics at OLX gives the low-down on sussing customers out with data.
Data Is ‘Everything’
Until 3 years ago, OLX never asked any information from its user. You could tour the entire website without providing a shred of personal information. No name, no number, no email address, preferences… anything — zilch, nothing, nada.
“Until three years ago, OLX never asked any information from its user.”
Just because you don’t feed any information doesn’t necessarily mean you can’t get data. Take OLX for example. If you take a customer — across all OLX Group companies — per month, the numbers (data) generated with respect to this user looks something like this.
These figures scream one thing — customer footprint.
Yes, customers leave footprints behind. And the absence of any — personal — information asked from them doesn’t deter data analytics from unearthing customer behaviour.
“Whether or not you ask customers explicit questions, they leave behind a footprint for you to look at and analyze.”
And the potential to make the most of this data is limitless. Uber, for instance, uses consumer data — and AI — to reflect surge pricing. In fact, it’s even aware of your phone’s battery levels and is able to determine if you’re likely to pay the surged price.
“This is the power of data. It’s not just what you’re asking the user or what the user is doing on the platform, but there is more power to it. It just depends on how much you want to leverage.”
The What, When & How of Customer Data
As is the case with any business, some customers stay while others recede. OLX has provided a platform for many success stories. But for some, it just doesn’t work out.
Why?
According to a case study showcased by Akanksha, sellers usually retract because the buyer:
- Thinks that the products are fake or cheap
- Underquotes
- Doesn’t show up
The conundrum here is the discovery of one buyer persona after much research. With millions of users and biased sample sizes, there’s very little research that can be done. Moreover, research boils down to the design and the outcome.
“The qualitative insights might be exceptional for such samples but you don’t really know how the business is behaving. So it becomes important for us to figure this out with the data on our platform.”
At this stage, asking 3 questions does the trick — this is where data analytics kicks in.
‘What’ are the leading indicators?
Basically, it’s looking out for red flags about the business (or customer data). In the case of OLX, detecting change in user activity, such as buyer enquiries, number of listings, and renewal of paid packages were the leading indicators.
Some other parameters include: app (and social media) reviews, earmarking of (any) repetitive feedback coming in through support staff, and feedback from account managers if they figure out there has been a drop in sales.
‘Why’ is it happening?
This is the stage where all the collected data is exhaustively analyzed to find out ‘why’ red flags are being raised. Some questions that can be raised (in the order below) in this step are:
- Is the posting funnel broken? As in, is the user able to successfully complete the listing process?
- Is the user not getting enough leads?
- Why is the user not getting enough leads?
- Is the user getting irrelevant leads?
There can only be two answers to these: YES or NO. A broken funnel would require the team to get to the bottom of what went wrong.
If ‘NO’, you’d explore leads/enquiries. At this stage, a broken funnel must be revisited among other things like a defective — or a lag in the — chat process etc. It’s just a matter of listing out all possibilities and striking each out that isn’t pulling the strings.
If the first 3 questions return A-OKs, (relevant) leads must be dissected. And how can that be done?
“On OLX, we have something called a “Make an Offer” that buyers use to put forth their offer. It’s rather simple to analyze (the kind of quotations that are coming in), but if you don’t have something like that then you essentially do some level of text binding … any offer (from a buyer) less than 60 to 70% of the sellers expected price will tell you that it’s an offer that will not be considered (by the seller). ”
‘How’ can it be fixed?
There are many ways you can go about doing this.
1. Change the product feature set
Product features are its distinguishing characteristics. This could mean the manner in which content is showcased, how much visibility is given to the ads (of a user) on the platform, and so on.
In OLX’s case, the content showcase was a feature that underwent a change.
And this ‘change’ was crucial as a user’s ads might not turn up on the first page despite the category — under which the user’s items are listed — being allocated a share of the listings.
As a result, the former — category-first — page on the OLX app was replaced with pure content. It was crafted in a way that displayed least- visited categories over the really popular ones (like cars and mobiles which hog 60–70% of user activity on OLX) that are in demand.
“ Populating the cold-shouldered categories first increased the inflow of users (in these categories).”
The “content first” method was also made hyperlocal. That is to say, the algorithm was tweaked to populate and display listings based on the distance from the location entered by the user.
And this show of listings was further explored and made to display as per the status of the location (Tier 1, Tier 2 and so on), thereby solving the problem of fresh content always taking precedence over older listings.
“The algorithm was based on the data we had… we ensured that we leveraged whatever information that was available.”
As seen from the graph above, this retouch solved a core business problem distinguished by other categories (apart from cars and mobiles)on the platform not getting enough user interaction.
“This was a great insight actually. We used to think that people come on OLX with specific intent . . . No! You serve them content; they will browse; they will act on it.”
2. Offer an incentive
This part is self-explanatory. In order to retain the user, certain incentives in the form of renewal bonuses and discounts on paid packages are offered.
3. Personalize business packages
This means a restructuring of pricing plans to suit the user. Coming from different locations, catering to specific categories, creating ads for varying demographics means the number of leads for each user will be different.
So, in a way, keeping the same pricing model for all users doesn’t cut it. Moreover, a certain percentage of leads will always hit a dead end. This might put off a user, or worse, result in an exodus as the ROI for the paid plan doesn’t match the user’s expectations.
OLX leveraged data — such as GDP per capita of cities and states, number of leads, the size of sellers wallet — to reprice their plans. As is evident from the chart, this gamble paid off and caused revenue and number of users to surge.
What’s more, even users who wouldn’t pay for any plans in the past came aboard the bandwagon after plans were repriced.
Even with all the data, sometimes it’s just hard to ascertain user persona
With more than 40 million users dropping by OLX every month, there is just too much to work with. And coming up with user personas — distilled to perfection — becomes an obstacle.
“If you actually get to that (40+ million users per month) level, there are many things to look at. You can’t possibly make a decision if you look at things — for so many use cases — at a granular level.”
So, how do you come up with user personas?
The solution is simple — segregate users into different groups based on the behaviours they exhibit. These behaviours are derived from data, and the user personas created by studying the averages of the same data.
How Much Data Is Too Much? And How Do You Know When To Stop?
“Data is never too much. The more you ask, the more you track, the better it is. There are things which you might feel you don’t need; track them anyway.”
Akanksha is spot on with her answer — data is never too much. You never know what you might need, so track whatever you can.
She also had an incredible anecdote to share which harped on the endless possibilities of data. It’s about how users’ mobile carriers were being recorded to know their point of origin. Adding to it, she mentioned how they were able to figure out why users weren’t staying on OLX after a certain point.
“…if you’re on Airtel and you know the price points for data plans, you can make sense of how long the user will stick around on OLX (until the data pack is exhausted). ”
The Takeaway
“A lot of teams and businesses are stuck at reporting (what happened) and don’t go beyond. Why? Because we do a lot of fancy dashboards. Tech teams build these dashboards at the request of the BI and Analytics team for many reasons: segments, drill-downs… but dashboards can’t help you beyond a point.”
It actually takes a human being to scrutinize the data to understand why something happened and what can be done to fix it.
Turning the data from business insights can be the secret ingredient to a business’ long-term success. Stop guessing; start leveraging data to have the biggest impact.
All About TechBytes
This chapter on Leveraging Data by Akanksha Dhamija is a part of the Techbyte series.
The idea behind this series is to invite external stakeholders to present and discuss interesting topics from their domains. This gives exposure and perspective to the entire Aasaanjobs team on the exciting stuff going on in the outside world and covers anything from new tech, scaling technology, new trends in the tech space, and so on.
Highlights of The Session
Why work with OLX People?
We don’t give you a job; we build your career and offer an opportunity to work with the brightest minds.
Join our team. Forward your CV to ta@olxpeople.com
