Ex-Amazonian’s TargetingMantra finds a way online into the minds of customers
How often have you scoured online for a good buy, only to give up in disgust, swearing never to return to that shopping portal? Don’t you find it annoying when you have finished reading a book, and your store has no sensible recommendations on what you should read next? When you buy a shirt, wouldn’t it be nice if someone could recommend the tie and cufflinks that go with it?
These are common irritants, if you are a regular online shopper in India. That’s because most consumer Internet enterprises are only now waking up to personalization. Amazon did, years ago. They got a team of super smart machine learning experts to mine machine data and build a recommendation system. An Indian, Saurabh Nangia, helmed this project.
Saurabh built the similarities engine from scratch to replace their existing system responsible for generating recommendations. He noticed the difference it made to the whole user experience, hugely increasing the probability of a buy. Web companies in emerging markets like that of India, however, lacked the bandwidth for a dedicated team in-house to take care of this. Saurabh noticed this, and decided to move to India. It didn’t take him long to convince his friend Rahul Singh and a couple of his teammates from Amazon to join in. Together, in early 2013, they kickstarted TargetingMantra in Gurgaon.
What do they do?
TargetingMantra learns user models for predicting preferences of customers. “We use real-time data that comprises how customers are interacting on the website, their past data, their personal preferences and the concerned company’s catalog data to dynamically train our user models. We believe our solution primarily requires expertise across three fields of computer science – big data (computation of large amount of data), machine learning (algorithms for making computers learn from data) and data mining (making sense from raw data),” explains Saurabh .
They are currently providing more than 15 solutions for personalization such as – ‘Similar items’, ‘People who bought also bought’, ‘Frequently bought together’, ‘New Arrivals’, ‘Best Sellers’, ‘Recommendations for customers based on their interest’, ‘Recommendation Emails’, ‘Customized Banner Ads’, etc. Companies can also track their performance and make strategic decisions through their analytics dashboard.
Rahul believes TargetingMantra has some advantages over others in the market currently. In fact, it is one of very few companies globally that provide such services:
- One stop solution: “We offer personalization, big data analytics and targeting solutions with a single integration.”
- Instantly available: “Any new team will take close to two years to develop a good algorithm. We offer it in two days, thus helping the client realize at least 10% of annual revenue which would have otherwise been lost.”
- Customised algorithms: “We believe that a one-size-fits-all approach cannot work when it comes to recommendation systems. We use a variety of different algorithms based on the kind of traffic companies get.”
- Best-in-class algorithm: “Our algorithms have provided substantial improvement to various companies against their internal solution and other external products. Our clients are welcome to AB test us against any other solution for a free trial period of 30 days.”
Having said that, TargetingMantra does have two huge challenges ahead of them. “Few companies in this region are yet mature enough in the industry life cycle to appreciate the importance of our services,” Saurabh says.
They are also finding it tough to hire good engineers who understand machine learning and big data analytics. Most of the experts are expensive, and prefer to work in Europe or the US. Currently they are a lean and mean team of five. Saurabh, of course, focuses on product development. Besides five years’ experience working with recommendation systems and machine learning, he holds a Masters in Computer Science, specializing in Machine Learning from the University of Illinois Urbana-Champaign (USA), and a BTech in the same from IIT Guwahati (India). Rahul holds the business development end of the company as his expertise is in market assessment, entry and growth strategy. He has an MBA from SP Jain (Dubai and Singapore), B.E. in Electrical & Electronics, and MSc. in Physics from BITS, Pilani.
Their target market includes all online businesses, from e-commerce companies to media portals, job outlets, and marriage websites, as well as blogs of all kinds. Currently, they are focusing on India but plan to expand into South East Asia soon. “So far, we are limiting to onboarding only 3-4 new clients every month. In a few months, we plan to raise money to increase our team size and help us expand our services in Asia and Africa,” Saurabh says.
“We have already validated our product in the market. TargetingMantra has breached a major milestone of processing half a billion data points across their customers since inception. We will continue to build and offer more features that can help to improve the conversion rate for site visitors,” Rahul adds.
YourStory asked them for their secret sauce. And here’s what they told us:
- Lean Startup Model – Seek continuous validation of your idea in the market before expansion. Iterate often and take customer feedback at every point. There is no point developing a product that no one wants.
- Patience and Perseverance – Startups usually don’t go viral right away. And B2B startups are especially slightly slower as the sales cycle is much longer. Therefore, it is important to be patient and listen to customer feedback and continue working on it.