On Developing An AI Powered Social Media Content Recommendation Tool

Tanmay Ratnaparkhe

•

Published: March 16th, 2021

save

Tanmay Ratnaparkhe

Founder, Predis

$1K

revenue/mo

Founders

Employees

Predis

predis.ai

from Pune, Maharashtra, India

started February 2021

$1,000

revenue/mo

Founders

Employees

Discover what tools recommends to grow your business!

platform

trello

SendGrid

social media

Twitter

Instagram

productivity tool-icon

Trello

payments

Chargebee

Stripe

Razorpay

analytics

Google Analytics

design

Figma

crm

Zoho

other

HotJar

Discover what books Tanmay recommends to grow your business!

Want more updates on Predis? Check out these stories:

1. Our AI Tool Generated $200K In 6 Months (2022) ➜

Hello! Who are you and what business did you start?

We are a team of three. This is my second venture. The first one was a tech company in the F&B space and got acquired by the largest digital products company in India. The other 2 co-founders are alumni of the prestigious Indian Institute Of Technology-Roorkee and have rich data science research experience and have participated and won various AI/ML competitions on global platforms including Kaggle and Analytics Vidhya.

Our flagship product is PredisAI. It is an AI-based tool that helps you write better social media posts and generate more engagement for them. This is done via our AI predicting how well will your social media post perform and also giving suggestions to improve it.

Our customer is anyone looking to improve their social media game and drive more engagement. As of now we only support Instagram and are targeting Social Media Agencies, Companies inhouse social media cells, and Influencers.

We are a pre-revenue startup and currently have 50 daily active users.

The unique thing about us is that All current Social Media tools are post-facto. They either help in scheduling or show like/impression-based analytics after a post is published. We are trying to predict how a post will perform (engagement) and suggest improvements to improve that before a post is published. This is radically different than all other Social Media Tools.

on-developing-an-ai-powered-social-media-content-recommendation-tool

What's your backstory and how did you come up with the idea?

Tanmay Ratnaparkhe is an engineer and started a software company in the F&B/restaurant space. In 6.6 years we grew it to 2500 clients in 72 cities in India and 5 cities globally with a team of 45 members. I sold the company to the largest digital products company in India and worked there for 2 years to aid in the expansion and transition. Predis.ai is the second venture for me.

Aakash and Akshay passed out of the Indian Institute of Technology Roorkee (in the top 1% of Indian engineering college pass-outs) and worked in a couple of high growth startups in India and Singapore where we solved research-based problems in Data Science. After 2 years of working there, we branched out and started predis.ai. Predis.ai was not the first idea we worked upon. The first idea was to make AI for AI :) It was complicated in every aspect - building, selling, deploying. This is why we thought of pivoting and looking at applications of the concept in different industries.

The thought process was this:

AI needs a lot of data to work with. We wanted a domain where we can get a lot of public data available.
Since the last startup was on the cost side, we wanted to be on the revenue side as the potential is theoretically infinite.
The Product needs to be 10X better than the nearest competition.
The Product needs to be SAAS (that’s what we understood) and DIY as it scales the best. It also needs to be as practically simple as one payment page only so that people could just come and pay us.
Software has anyways been eating the world. Social Media has already changed the way companies communicate with their customers. However, social media is a black box for most companies/ agencies. How to craft a good social media strategy is still an unanswered question.
Based on this and our core strengths (Founder with a successful exit and Strong Data science experience, and after talking to a lot of people we thought of attempting a seemingly impossible problem - predicting how your social media posts will behave and hence recommending what can be improved and hence helping you generate more engagement. The end goal is to become an AI-based Social Media assistant which can craft posts for you.
Idea validation has happened in the following way:
Talking to a lot of Social media agencies, influencers, business owners on the problem and the potential solution.
Getting an MVP out in the market and getting beta users to play with it.
Getting 50 active daily beta users using the product daily. The best part is that this has happened organically.

Tanmay Ratnaparkhe, Aakash Kerawat and Akshay Karangale

Take us through the process of designing, prototyping, and manufacturing your first product.

We have always followed an MVP - Minimum Viable Product approach. The first thing was to zero in on the outcome which we could provide to users in phase one.

Never try and cheat. Be genuine even if it means you need to turn away the client. Trust is the biggest asset you can build and it is very easy to lose it.

Next was to make wireframe-based designs on how the flow will look like.

We then decided on the tech stack - based on the current team’s comfort levels, we decided on the first set of technical tools/frameworks which will enable us to push the MVP out in the market. We also had a look at the APIs offered by each of the platforms - Twitter, Facebook, Linkedin, etc and see what all types of offerings would be possible.

Our first product cut was over Twitter as the APIs were straightforward.

After building the first MVP, we reached out to prospective customers through friends of friends network. We also ran a few Facebook/Instagram ad campaigns to see how many people would be interested. The Facebook/Instagram ad campaigns were simple - just a small description of the problem/solution and a link where people could submit their email ids in case they were interested. The campaign was a success and we got a lot of users who were interested in our offerings and were ready to try out our product.

As we spoke to more and more customers while building this, we realized that people were more interested in Instagram instead of Twitter.

This made us change our focus and look solely at Instagram instead of Twitter.

As of now we are supporting only Instagram and looking to perfect our offerings for Instagram before launching for other platforms.

Describe the process of launching the business.

The first and the most important thing for us was our Website. This was also necessary as we envisage ourselves as an online-only business for some time.

We spent a good amount of time working through the content and the creatives for the website. Once that was done, we also did the design and the implementation of the website.

Side by Side, we also have been in the constant MVP/Feedback state for the last 2 months. This includes launching a feature, talking to prospective customers about it, and taking their feedback.

Financially, we have been bootstrapping the business till now and have been very frugal in our overall approach. Since Covid hit, all 3 of us had to move in back with our parents as it was not possible to stay in Bangalore - India was under a strict lockdown and once our cook went away, the writing on the wall was clear. This also became a blessing in disguise as our personal expenses reduced drastically.

Bootstrapping an operating frugally also means utilizing free plans from almost all service vendors we use, optimizing the cloud servers to lower the cloud service spend (which is our biggest expense), focusing more on guerilla marketing tactics instead of starting to spend money on Ads. All of the expenses right now are borne by us.

We are also starting to set-up our social presence - selling social media marketing tools means we need to definitely have a strong social media game.

We have also set-up a knowledge base on our website which helps customers learn about the product and resolves any queries.

Since launch, what has worked to attract and retain customers?

We are in the process of a full-fledged launch. However, the following has happened till now.

As our first version of the product went live, we put in Google Analytics and Hotjar to track user behavior. To our dismay, we saw a near 100% drop rate at the time of user onboarding. This basically meant that our onboarding flow was so complicated that users were finding it impossible to complete it and hence though a lot of users came to signup and use the product, practically no one was using this.

This pushed us back to the drawing board and we revamped the onboarding flow to make it super easy for users to sign up and start using Predis. This resulted in a big change in the onboarding funnel and now almost 95% of the users who sign up do try running at least one prediction on Predis.

The before and after flows can be seen here: Before - After.

We have also started reaching out to users via emails to get their feedback and nudge them back to the platform.

How are you doing today and what does the future look like?

Like all founders, we are fairly optimistic about the business. We are at a fairly early stage and have a lot to improve on all fronts. The short term and long term goals are as follows:

Short term:

Optimize the entire Predis experience for Instagram
Add all other platforms.

Long term:

Full-blown AI assistant which will create posts (with creatives) to be sent for your approval. Will be posted once approved.

What platform/tools do you use for your business?

We use a lot of tools as of today:

Zoho Apps/Google Docs - Take care of emails/excel sheets.
Trello - Planning/Bug tracking happens here.
SendGrid - Customer outreach via emails.
Chargebee + Stripe + Razorpay - Managing Subscriptions and payments.
Figma - Wireframing, Designing flows.
Amazon Web Services/Google Cloud - Cloud service to keep the product up and running.
Google Analytics - User Analytics
Hotjar - Heatmaps and Behaviour Analytics
Tawk.io - User Chat Support

What have been the most influential books, podcasts, or other resources?

We read a lot of articles over the internet. Following are some of the things we keep on reading from:

Hacker News/ a16z blogs
Medium articles
Books - Godfather - If you read it closely enough, it is not a mafia story but is a book on how to conduct business.

Advice for other entrepreneurs who want to get started or are just starting?

Try to start on the revenue increment side instead of the cost side.
Try and build something which is 10X instead of incremental changes over the existing products.
Don't hesitate in spending money on things which will give a good return. Fundamentally trying to do everything by yourself means you are investing your own time and hence money which could be better used somewhere else.
Focus only on the things which you do best. Outsource everything else.
Make the first 10 sales yourself.
Always listen to customers. There will be a lot of learnings from observing how customers are interacting with your product.
Always have metrics and tracking for every action you take. Gut feeling is great, validating that gut feeling via metrics and results is pure awesomeness.
Never try and cheat. Be genuine even if it means you need to turn away the client. Trust is the biggest asset you can build and it is very easy to lose it. This applies to everyone you interact with - Customers, Employees, Investors.
Try and build something which has value for people. While selling, sell the value only and not the features.
No one gets rich in a short time. The hard+smart work takes time to compound but explodes once you put enough of it.