“When we give ourselves permission to fail, we, at the same time, give ourselves permission to excel.” – Eloise Ristad
For three years, Akshaya Aron and Dipankar Sarkar worked on natural language processing (NLP), Machine Learning (ML), neural networks and Artificial Intelligence (AI). This helped them in making the first fully automated bot on WhatsApp where one can add a number, and start chatting with it. They ran it for four months – got 40k users without spending any money, got five million messages exchanged on the platform – and finally shut it down, for not being a viable business.
But, this is not the end of the story.
Earlier in April this year, the duo gained pace again and came up with Octo.ai – initially as a platform for fully automating a company’s internet marketing needs (push notifications, personalised feeds, etc.) – and later in July as an analytics platform built for Machine Learning.
They also recently raised Seed funding of $200K at a valuation of $1.5 Mn. The investment was led by a host of angel investors, which included Rohan and Arjun Malhotra (Investopad); Rahul Khanna (Trifecta Capital); Rakesh Agrawal (US based- seed in Lyft, Cruise, Shyp, Poshmark and Lendup), Sidharth Rao (WebChutney), Outbox Ventures, Rajan Navani; Gautam Gandhi (Ex-Google, Angel Investor); Jaspreet Bindra; and Gagan Duggal.
The team of seven is currently based in New Delhi.
Journey From A WhatsApp Bot To An Open Source Analytics Platform
The duo initially decided to use personalisation algorithms and intent capture algorithms they had made for the WhatsApp bot, and launch Octo as an automated marketing solution. “We were offering intelligent push notifications and personalised feeds along with the usual dashboard, segmentation, etc. However, our initial sales cycle, made it obvious that some of the assumptions we made were flawed,” said Akshaya.
The one critical assumption was that the data would be available to run algorithms. As Akshaya explains it further, high-end user personalisation problems require data to be structured correctly. But when a company starts, the heavy dependence on free/ freemium third-party analytic services – Google analytics, Mixpanel, and others – results in heavy data loss as most analytic providers lock it into their systems.
In this scenario, they would need to be in the company’s stack for 3-6 months. They figured while selling that every client ( enterprise clients especially ) don’t want to share their data with startups and the threshold for entry is high.
“At the start of this month, we open-sourced our product code, it is a combined version of MixPanel and Segment.io. Any startup/enterprise can deploy it on their own servers and customise it as well. This has been done to give control of data back to the companies and enable everyone to use ML & AI on their data sets. Till now they have all been locked in on paid third party services” said Akshaya.
Here is how it works: (before using Octo)
(After using Octo)
Growth Hacking Companies’ Marketing Strategies
According to Dipankar, the framework of Octo collects all the analytics data and allows real data science/ML going forward. Also, as one size doesn’t fit all, companies need freedom to modify data structures to accommodate algorithms better suited for their business case.
“We are open-source, hence infinitely customisable, as per the need of the client. This allows companies to extend the system for their own internal Machine learning algorithms and workflows. We estimate over 70% reduction in their external SaaS costs when they start scaling,” he added.
Majorly, companies can make the best use of Octo in four possible ways:
Use it as their in-house analytics solution: Instead of using Mixpanel, GA or any other third-party tools they can just install Octo, put the one line code on their website and apps and have their own version running! The biggest advantage is owning the data and also being able to customise the reports as per business need.
Autonomous push marketing/marketing automation: Companies can install Octo, and send personalised push messages based on triggers to each user to drive more active users and engagement. Like all marketing automation platforms, Octo comes built in with an advanced rule engine and an advanced routing mechanism to handle all marketing communication. eg. If Octo’s analytics show a user is engaging less, an automatic notification can go to them to get them back on the system.
In-app and on-site personalisation: Companies can use the recommendations and segment-of-one features to personalise their app and website for each user, just like Facebook or Netflix. Octo increases time on site and revenue by providing content from the system which will engage the user.
Marketplace seller analytics: Marketplaces can use it to power their seller analytics via a multi-tenant and highly scalable system. The system can also provide sellers with recommendations for products and key insights on how their products are doing.
Monetisation, Competition And Challenges Ahead
The revenue model of Octo is similar to other open-source companies, like MongoDB, and Red Hat. They will charge companies for customisations (traditionally enterprise clients who need a custom solution built for themselves + technical support), hosted solutions of their open-source offering, consulting and training on Octo.
The team is currently competing with Adobe Omniture, Google Analytics, Mixpanel, Segment.io, Kahuna and various other companies.
When asked about the challenges, Akshaya replied, “The biggest challenge with any open-source technology is adoption. As we are a new company, big companies who would benefit a lot from such technology are a little hesitant in adopting us. Other than that the long decision cycles in bigger companies is always something which hurts small startups.”
Pushing Internet Marketing To Beyond Basic Software Stack
The team currently has an internal target to reach 100 clients by the end of 2016.
The Octo founders aim to push Internet marketing to the next level much beyond the basic software stack, where machines are autonomously taking actions based on user behaviour insights. Right now all these tools are used by marketing teams to make inferences and perform actions. The duo further believes that an AI system should perform these actions as it will be more efficient and obviously more personalised.
“In India, we are one of those few team startups who are actually going to do fundamental changes. We want Octo to be used as Mangodb, or SQL, so that whenever any developer builds an app, they usually add an entire code library to it, and open source really helps in it,” said Dipankar.
According to a June 2016 report by Nasscom, the Indian analytics industry is expected to reach $16 Bn by 2025, from $2 Bn currently. India is presently among the top 10 destinations for analytics, with AI and deep learning algorithms creeping gradually in all aspects of even the best of the crowd.
The coup de grace in it is segment-of-one marketing and analytics-driven marketing automation. Industry veterans believe this market to be headed towards use of Machine Learning and neural networks on data sets. This will help companies to serve users better and to improve their experience in a mobile application or on a website – with startups like Octo around to facilitate the process.