Interview by Anil Chopra, VP-Research and Consulting, CMR
Khatabook, a name that was unheard of a few years ago has become a leading name today as a provider of digital tools that help Indian MSMEs support their business. Its utility solutions became so popular that the company registered over 4 Crore merchants in less than one year. Data science and AI have played a key role in the success of Khatabook among other factors.
We spoke to Arpit Agarwal, VP- Analytics & Data Science at Khatabook to understand how the company has been leveraging data science and AI to solve the problems faced by the MSMEs.
What are some of the day-to-day business challenges faced by Indian MSMEs that Khatabook is trying to solve? What are the metrics that determine the successful resolution of that problem?
We’ve tried to pick up small use cases one by one depending on their criticality. While most people were building solutions for Sec A class of people, nobody had really explored these areas for investments or tried to find new solutions to these problems.
One problem we’re trying to solve is building digital solutions related to bookkeeping and credit recovery and any challenges MSMEs face in their day-to-day businesses. A day in the life of a retailer and distributor involves ordering inventory, trying to bring back customers, ensuring staff attendance, and salaries. We’re building solutions around it.
We launched a simple app in which people could record any credit transactions. The app also helps them to set reminders. It made recovery very easy instead of the hassles of recording everything. MSMEs who started using the app have been able to increase the recovery by 3x using this app.
After taking our solutions to MSMEs in tier 3 and 4 cities, we started building a solution for MSMEs which can help them manage their staffs and salaries. An average small business generally has a skeleton staff of not more than 4-5 staff, and it becomes a problem even if just one of them is absent.
Some employees also take the advance from time to time. It becomes difficult for the owner to keep track of all the staff related tasks. We understood all these challenges and launched Pagar Khaata, which was able to manage the attendance and salaries of the small businesses. Within two months, the number of Pagar Khaata users sky-rocketed to 7.1 million.
This growth has happened because the use cases that our solutions are built around have been unexplored till now.
We have also launched an app called MyStore for small businesses to help them take their shops online. For a Kirana store selling within one kilometer radius, there’s no real online identity for the shop. Our solution helped in giving them an online identity, to which they can upload their products catalog, generates a sharable link, and can share it with their prospective customers for ordering.
We’ve just started and have a long way to go because it’s a deep market.
Which categories of the MSMEs have recorded the highest adoption of Khatabook so far?
The highest adoption has been recorded in the micro and small business segment, across just about every type of category such as Grocery, Kirana, Electronics, Mobiles, Accessories, etc. - especially the shop owners who have a lot of inventory.
Khatabook is very easy to use and meant for people who were not using any accounting software at all. Medium companies would be using accounting software like Tally and would have the workforce to maintain their accounts. This is where we hit our growth the fastest.
Now, we have equal penetration even in medium enterprises. People in this segment would have to go on their desktop, put dates and reminders, etc. or remain dependant on their accountants. Khatabook is a simple app that people even in this segment can use on their own without depending upon anyone. They simply record an entry and send reminders.
We recently acquired Biz Analyst, which sits on top of the accounting software and provides business intelligence. Tally, for example, records minute details about suppliers, but it doesn’t give the intelligence like who has the highest outstanding, etc. This software sits on top of Tally and is both web and mobile-based. It gives recommendations where you can slice and dice customers, know who’re the people to recover money from, who haven’t ordered from you in a long time, etc.
Where does Data Science fit into the scheme of things?
The use of Data Science can be categorized into two buckets. The first bucket is to understand customers really well, and the second is to enhance their product and user experience, e.g. This customer hasn’t paid up - it's time to send them a reminder or This customer prefers to buy often from- Can I recommend them something?
When I say user experience, the most critical thing in it is that their money should be in safe hands. We’ve built exhaustive fraud detection models that work in real-time. Transactions are possible on Khatabook. It will send you a reminder and a payment link to make the payment. If someone is doing frauds, it safeguards the merchants from it.
We’ve worked extensively on fraud detection and currently, the percentage is very less, but whatever has happened, we’re able to capture the 95% in real-time. Fraud is also likely to happen when you’re collecting money through the Khatabook app. The second is in the QR area. We have khatabook QR, which can be scanned to pay.
Another feature is smart suggestions. If there’s a customer who purchases often, then the app should be intelligent enough to keep the repeated entries marked on top. Moreover, Khatabook is available in 13 languages, which also takes care of the language barrier faced by the small businesses. We also have voice recognition is in the pipeline, but making it multi-lingual can be a challenge.
How is AI being leveraged to help MSMEs?
We’re using it to understand users very well. So, for instance, we check where they’re coming from, location, cities, business size, credit behaviour, credit recovery behaviour, etc.
We then think about the solutions that will fit them the best. e.g. people who’re not being able to recover money fast, can we create recommendations on how they can recover the money? Can we, for instance, flag the transactions as risky?
Another one is predicting whether the merchant is likely to use the app in the next week/month/etc. We’re using this model to predict that to understand their pain points and take corrective actions. All this goes back as feedback to the product team. It helps us customize our campaigns, and engage with those who’re interested.
We also use AI in lead gen modelling, as Khatabook has a base of 40-50+ million customers. How to use this base to cross-sell. Would it, for instance, have a use case for Pagar Khata, or My Store, and how they’re adapting to other apps, etc.
We use it to optimize our marketing strengths. Right now, we can proudly say that we have one of the lowest cost of acquiring customers. Even apps like Pagar Khata we’ve done at 1/10th of the cost of our nearest competitors. The app has a premium version that’s paid and provides premium features that will provide significant incremental value to the users.
Does data science help individual customers such as Kirana store owners?
Yes. For instance, My Store has suggestive products features. Imagine there are millions of SKUs to choose from, but it’s very difficult for someone to go through all of them. Data Science helps determine what will be the most probable products that the person would like to upload in the catalog.
It helps people upload their catalogs very easily. There was a significant uplift in the use of products. It also keeps track of uploading patterns and gives them intelligence of what’s selling best. E.g. if they have 1,000 things to upload, it tells them the 5 things that will increase the propensity to buy.
We’re using another branch of AI in khatabook called NLP. As we’re using 13 languages, Khatabook can decode them and convert unstructured data to structured data, by taking feedback from multi-lingual customers.
The Power of Data
Khatabook has millions of people who’ve installed it and using it regularly. You can imagine the amount of data that we’ve recorded about Indian MSMEs. We’ve created an MSME Index to draw insights about the MSMEs pan India at a level of pin-code, street, etc. That’s the power we have. We’ve even released an MSME report from it, which can be used by the policymakers, the banks for underwriting, or can be used by MSMEs themselves to know the trend upfront.