PolicyBazaar.com is India’s largest and leading insurance aggregator website. It specialises in making comparative analysis of the insurance products of various insurance policies based on price, quality and key benefits. With this information, its customers can make informed decisions about the insurance products they need to purchase. Having bagged 90 percent of the aggregator market for all types of insurance policies in India, the Gurgaon-based company is much ahead of traditional insurance providers in the online or digital space. The company claims to be 15 – 20 times larger than its nearest competitor. PolicyBazaar.com sells auto, commercial, life, health, term, motor, two-wheeler, and travel insurance policies, as well as ULIP products. It does 300,000 transactions a month. It has been in operation since 2008 and is growing at a CAGR of 100% over the past three years. The latest round of funding from SoftBank (nearly $200 Mn or INR 1,300 Cr), will soon make PolicyBazaar.com a unicorn.
Another plus point for the company is its excellent conversion rates – its average visitor to conversion rate is 4% – 5%. Compare that to the e-commerce average which is just 0.5% – 1.5%. It boasts of a healthy business backed by a robust call centre of advisors that help customers choose the right insurance products.
“We started as a lead aggregator wherein we used to get customers, show them some information and pass on the lead to our partners,” said Ashish Gupta, CTO & Co-founder, PolicyBazaar.com. “But that did not work well for us because the partners did not have the infrastructure to sell online. So we migrated as an insurance aggregator, wherein we sell the policies — now we are taking the next step in also ensuring that the claims etc. are managed by us. We have become like an end-to-end insurance provider — and a front-end for the brands.”
As PolicyBazaar.com scales its business, it faced a challenge with its workforce. The biggest challenge was scaling up without increasing the workforce too much. After all, it is just not feasible to double the workforce every year.
“This is a very people-intensive industry. Our mandate is to grow 100% year-on-year for the next 2 – 3 years. But how do you do this by increasing manpower 30 – 40%? That’s where automation, intelligence, and analytics helps,” said Gupta.
How technology helped
Gupta gave us a few examples to explain his point.
Last year PolicyBazaar.com successfully executed a project called PBee Connect — an assisted chatbot. More than half of its car insurance is done through the bot. That’s a whopping million dollars of premium collected every month!
The bot has also improved agent productivity dramatically. Earlier, an agent used to do 1.5 – 2 sales transactions a day. Assisted by the bot, the same agent can now do 7 – 8 sales transactions a day, informed Gupta.
“The bot is doing this on its own 80% of the time. The agent just steps in when a complex question is asked, 15% of the time,” said Gupta.
This has increased agent productivity by a factor of 4X and that has propelled the growth for the aggregator’s motor insurance business.
PolicyBazaar.com will soon introduce progressive dialling or automated dialling. At present, it is the human agent who decides when to call and who to call. But that decision will soon be made by an analytics system and the call will be made automatically.
“We have already done this in one business unit. That gave us a phenomenal 50% increase in dial-out rates,” informed Gupta. “This kind of efficiency and work productivity also helps the advisor to earn more. Our efficiencies have gone up from 40% to 85% — in the call efficiency rate. That means we are calling back customers within 30 – 45 minutes. So it is a win-win situation for everyone.”
Harnessing data for customer insights
The next step for PolicyBazaar.com is unlocking its analytics through voice. It clocks 35 million minutes of talk time every month, and that is a gold mine of data that tells it a lot about its customers. The company wants to harness this data to learn more about its customers, their investment patterns and also about their families. This will help it sell more customised products, such as short-term insurance products.
It is now converting all the recorded calls into text and running an NLP (Natural Language Processing) engine. This exercise is already yielding some interesting insights. For instance, when it analysed its health calls it found that the 7th or the 8th most common request from the customer was cashless.
“We never did think that cashless was such a big deal in health. But the moment we saw that, we built it into our product. Now we have cashless hospital network on our site,” said Gupta.
According to Gupta, analytics and automation have helped the company to answer more queries and make the product more robust; that makes it a self-serve product, reducing its reliance on human agents to assist the customer.
Analytics is also helping the company get accurate information on claims settlement ratio, which is critical in the insurance industry. This is the number of claims made to a particular insurance agency, clocked against the number of times it honoured the claim.
“Our engine will detect that you have asked for the claims settlement ratio and it will send you an SMS with all the data. And the advisor will just need to tell the customer about the SMS instead of rattling out the numbers,” said Gupta.
His vision is to use technology like analytics to provide this critical information with less human effort. And it also makes for a great customer experience.
PolicyBazaar.com is also combining customer data with claims data and then creating new, personalised products such as short-term insurance, which traditional insurers would not be able to do.
“We have collected all this consumer data, in terms of what they came in and what they wanted to buy; what did they eventually buy. What does their family look like? What does their income look like? We are marrying this with claims data. We get claims data from our insurers, we compile all of it and look for correlations. Then we create new products around that, like new short-term insurance products, that last only for the duration of an event, like say travel. This is not possible in the offline world,” said Gupta.
While this has worked for its auto insurance business, it next wants to do this for term, health and home insurance.
PolicyBazaar.com is also using cloud technology to save costs. It moved its infrastructure to the AWS platform and that helped it reduce costs.
“We are now completely hosted on AWS and that brought in cost savings. Our costs are now 20 – 25% lower than what it was on-premise,” informed Gupta. “We see fluctuations at the end of the month or fiscal year. For on-prem, we always had to cater to the maximum. Now we don’t as we have a load balancing server – we use four servers during the day and only one or two at night since we are an India centric business.”
AWS offers PolicyBazaar.com a programmatic process that lets them automatically shut down unused servers after 9.00 pm. The company pays only for the resources it consumes. Of course, employees can still deploy servers after 9:00 pm should the need arise.
Servers can also be provisioned automatically as the business scales.
Gupta recounts an incident where automatic provisioning of servers was so seamless, the company did not realise that their infrastructure had scaled up until the end of the day. The day was 31st March, the last day of the financial year in India. Its business saw a huge demand for two-wheeler insurance on that day because it is fashionable to do things on the last day before a deadline in India. For its two-wheeler journey, PolicyBazaar.com normally deploys 2 – 3 servers, but on that day, it scaled to 16 servers on the AWS cloud platform.
“We did not realize it until the end of the day when looking at the report. There was no downtime and no manual intervention. The scaling happened automatically, and no extra effort was put in from our side,” said Gupta.
PolicyBazaar.com is also benefitting from other services on the AWS platform, such as Amazon Polly, a service that turns text to human-like speech (with local accents), using deep learning.
For instance, the company creates text scripts to inform customers about missing documents and reminders for medical check-ups. This is very personalised or subjective and the message differs for each customer. The Amazon Polly service converts the text scripts into voice and this is then transmitted to the customers.
Apart from this, AWS is helping PolicyBazaar.com on voice to text, and then converting that into analytics. PolicyBazaar.com had to upgrade its infrastructure for this, so it invested in GPUs for audio processing. They also use TensorFlow and cloud – and Amazon SageMaker, to build, train, and deploy machine learning models at scale.
“We don’t have to worry about our infra as we leave it to AWS. We just focus on what we are good at. We think AWS is the most mature product. The ability to control servers programmatically is so superior in AWS. The processes are also very efficient. I do not have to approve every server that’s deployed. I just get a report at the end of the month. So our processes are improving and people are getting more comfortable,” concludes Gupta.