Predis.ai – 2 Months Overview (Apr/May 2021)

A lot can happen in a startup in 2 months. What’s been happening at Predis?

We launched Predis.ai on 8th April on Product Hunt🥳🥳🥳. It was the result of months of work – largely done in a sleep-deprived and over-caffeinated, albeit hopeful state ☕ (and while managing an infant and a 2-year-old in the case of 1 of the founders). Like any startup, we had an amazing number of unforeseen twists and turns, including a pivot from Twitter to Instagram.

For those who have no idea what Predis.ai is – We are an AI product startup that has the following revolutionary benefits:

1. Instagram creators (Social Media Agencies/Influencers) use our AI to ideate and craft better posts in half the time. This is because our AI is trained to give suggestions on how to improve Instagram posts before they are published 🕔.
2. Instagram creators (Social Media Agencies/Influencers) use our AI to visually understand:
   • How their own content has fared, what has generated max engagement 📈.
   • How their competitors are doing, what content works for them, and hence what needs to be picked up for the next content calendar😈
3. Both these features are not available in any other products as of now (to the best of our knowledge).

It has been just over 2 months since we launched and we thought of taking a step back and reflecting on where we are.

The reflection also helps in understanding what has gone right and what has gone wrong for us. We thought about publishing our learnings as this might be insightful for other founders as well.

We have divided the blog into these sections:

1. Why we pivoted from Twitter to Instagram.
2. How our Product Hunt launch fared.
3. Why we decided to offer LTD on Pitchground.
4. Key learnings from user metrics. 

Why pivot to Instagram?

Our product needs platform data to train our AI which then derives insights shown to the users. Our API Analysis showed that Twitter APIs are most open and hence we thought of starting with the Twitter platform. However, as we started talking to more potential users, it came out that Instagram made more sense as there was much more marketing happening there. Twitter is absolutely not off the cards though; we will get back to it once we add a new host of features we have planned out for Instagram 💪.

Launch our startup on Product Hunt

We decided to launch on Product Hunt as it’s the most logical way of putting your product out for the world to see. The PH launch was a mini-disaster 😭. We were genuinely hoping to make the top 3 apps of the day but ended up being the 14th or 15th. Upon analyzing, we figured that we did the following things wrong:

• Hunted it ourselves. This happened because we read the Product Hunt blog which recommends hunting for products yourselves. However, it does not work out in real life.

That u actually need a good hunter. I read the PH blog which recommended hunting your own products. As a PH newbie, Missed the community leverage. Later almost every one told me I was wrong in hunting myself.

— Tanmay Ratnaparkhe (@TanmayRatnapark) June 11, 2021

• We did not have a community and hence there were no people rooting for us. Product Hunt is a large community where people do help each other out. Since we were practically new to the PH/FB/Slack/Reddit groups ecosystem, it was impossible for us to have a fan army. 
• In a typical way of fixing things on a running train, we asked a lot of friends to create PH accounts to upvote us. This resulted in us getting some upvotes and staying in the competition but we were pushed out of the main page by the PH algorithm – Which we believe gives more weight to the upvotes from established PH members. Even today if you see, most of the comments have red balloons on top of them😂
• Once we were out of the main page, it was just a race to the bottom with the top products getting more eyeballs 👀 and Predis getting only a few. 
• However, all was not lost 😇 We still got a bump in our non-existent traffic

• This also got us 30 new users who gave us some nice feedback.
• Obviously, the traffic from Product Hunt always has such trajectory, and we had expected the drop. 
• The idea was to test the waters and despite the flaws in our launch, we were able to accomplish that 💪.

Post the launch, we wanted to do the following:

1. Reach out to more users. 
2. Get some revenue. We are bootstrapped and a little bit of revenue does not hurt anyone, right? This is also good as users who have invested some money in your product are easy to retain and also share more feedback!
3. Hence, we started looking at launchpads which will get us both. 
4. Life Time Deals (LTDs as they are called) have become very popular and there are a lot of platforms for it.
5. We decided to go ahead with Pitchground as they were ready to promote us. We could have landed ourselves in the AppSumo Marketplace, but did not have any marketing firepower to push ourselves 😅.

Launched on Pitchground.

What worked?

• Pitchground delivered some traffic to the website. We saw an uptick in the free signups too 📈.
• We got some paying users and more importantly got feedback. In 2 months, we have made around $4.5k in total revenue from 50 users 💰. After Pitchground’s part, we are left with $1.8k in revenue. 

• Did webinars and generated some traffic, and I’ve now been working on conducting better webinars. Since I haven’t done webinars before, I was pretty nervous before the first one. I think I got significantly better by 2 and 3. See for yourselves here 😇

• We got feedback on the UI from the Pitchground founders – How the UI sucks and improving UI will improve conversions. We had a similar thought process and was on our roadmap and hence we thought we’ll close on this before we start on the next set of AI-driven features.

What did not work?

• Predis did not get as much traction as expected 😰. We were expecting a total of 10k over 3 months which meant that we had to be at $6k by end of 2 months. However, after the 3rd webinar, we have covered some ground 💪. 

• Depended too much on getting ads run by PG to generate traction. To be fair to Pitchground, if we were in their place, we would also want to promote the top-selling deals heavily to maximize the revenue. Also, Predis.ai is targeting a specific niche of users(Instagram Marketers as of now) and hence our penetration is lower with respect to other products that target a large set of users.

Learnings

1. We need to have a better buzz machine across Facebook groups, Instagram, Reddit, MPSocial, BHW, and other places where our target users frequent. 
2. We understood how LTDs work and how they are perceived by users. LTDs are double-edged swords and one should use them judiciously. For us, we will want to do 1 run on AppSumo before we focus full-time on our monthly recurring revenue. 

Key Learnings along with user metrics 

Concurrently, we have been reaching out to potential users via various channels and the stats look like the following. We have had 318 signups in the last 2 months. The source-wise stats are the following.

As you see the drop-off rate is a lot for users from all sources. This is kind of expected as we are still figuring out the PM fit. For us, the next step was to analyze the drop-off rates.

Analysis of Drop off rate

We know that our drop off rate is a lot. However, this is also something that is expected and we intend to improve this over time. We start the analysis by checking the drop off rates according to the following metrics :

1. Source of users
2. Product usage vs Drop off rates
3. Device vs drop off rates. 

Source of users for our startup

1. As seen in the above table, the least drop-off rate comes from users from Pitchground and word of mouth (personal connections/outreach).
2. This shows that users from these sources have the best idea of what the product is about and have the right expectations from it. 
3. Takeaways – Focus more on Predis.ai values on other platforms so that users have a better idea before they come to signup.

Product usage vs Drop-off rates

Next, we wanted to see how the product usage impacted the user drop-off rate. The key takeaways from this exercise are:

• We have 2 major parts of the product – Improve Post and Explore Content. Improve post is a tactical module and helps users improve their posts on a day-to-day basis. It also helps them build on a base idea into a caption and finally the post. The other part – Explore content is more strategic in nature and is used to understand what kind of content is working for the handle and also for doing competitor analysis. 
• First, we checked the drop-off vs how many times users used Improve post feature. We divided our users into Heavy / Medium and low usage categories. The usage vs drop off for the Improve Post section looks like this:

• Obviously, the heavy users stick around and have a very low drop-off rate and Max drop-off is from Low usage users. 
• Key takeaways here –
  1. Need to improve the onboarding flow to make sure users run more predictions and hence convert into heavy users.
  2. Need a way to nudge the medium users into running more predictions and hence converting them into heavy users. 
• The drop-off rate for users trying out content analysis looks like this. Here also, If we somehow make more people run content analysis, the drop-off will come down by 50% for such type of users. Obviously, there may be some drop-off due to resistance. However, we think it is worth the risk.

• Another important metric for us to check is how do users who link their Instagram business account fare. If we get people to add handles, the drop off will come down by 40% – 81% to 55%

• Key Take-aways –
  1. Need to get more users to link their handles. This will also let us reduce the friction further where we can use the last post itself to show value. 
  2. This also let us show the handle’s own analysis without any further action. 
  3. PitchGround and Word Of Mouth are working the best with the least drop-off. Other channels have lower contribution and higher drop-off

What’s next for our startup

• Based on this data, we are revamping our UX and the onboarding flow with the following objectives.
  1.  Need to get users to run 10 predictions to become heavy users and hence reach a 9% drop-off rate.
  2. Need to make new users experience the explore content/competition analysis so as it decreases the drop off to 41%. 
  3. Getting users to link their handles at the start will make sure we are able to show a lot of value to them and hence getting the drop off down by 40%
• Looking to launch in-app/personalized notifications to entice users to try other modules and also nudge them to come back.
• This blog also marks our start of Build in Public. You can expect regular updates from us.
• User Outreach
  • Looking at driving more users to the top of the funnel so that the retention also becomes more.
  • Looking to start making more Reels on Instagram😢. Honestly, this is more painful than any other thing in our startup. However, this is most important as we can empathize more with creators and also helps us get more traction on Instagram!

If you liked reading this and have any feedback for us, please reach out to us on Twitter! Also, feel free to share if you think someone else might benefit from this😊