The right way to fail?
Kindigo went from zero to a fully funded 10 member startup to zero again, all in 6 months
The idea for Kindigo was born in December 2017 after a few weeks of research and brainstorming. We went from no idea to an idea to a funded startup with 10 member team to launching a product and finally to a failed startup in 6 months. This piece is about the failed startup with a hope that you can find some lessons on how to fail early.
The journey starts with me joining as an Entrepreneur in Residence (EIR) at Human Ventures, a startup studio in Manhattan.
We started with blank slate looking at the canvas of various stages and needs of human life at HV’s office.
After 2 weeks of shortlisting and then further shortlisting and then further shortlisting, we arrived at the idea of Kindigo, which was not even on the original board.
The reason I was drawn to the idea was my own encounter with bullying in my school days. The initial research showed clearly that cyberbullying was the top things parents were concerned about when it came to their child’s health. This was even above child obesity, drug abuse, and mental health.
We spend another 2 weeks doing more research including market size, competitors landscape etc. We surveyed 400 parents on cyberbullying and the results confirmed that the problem was real and being faced by almost every parent. What was astonishing was that compared to the size of the problem (1 in 2 children get cyberbullied, 1 in 2 victims have lower self-esteem) very little is being done to solve the problem. We used pollfish for our survey.
Armed with the confidence we started brainstorming on what the solution should look like.
With my fairly technical background, our understanding of the problem was also technical. The hypothesis went something like this:
We then came up with a solution which was technical. The gist of it was that we ingest all the interactions children are doing online (facebook, apps, websites etc) and then pass it through a machine learning based trained classifier which can flag cases of cyberbullying and then finally show the results on an app to parents.
Next, we did some mockups using a free tool called InVision.
Based on this idea Human Ventures gave me a starting capital to work with for the next 2 months. Our goal for this period was to come up with a proof of concept and a digital presence and a strong pitch to raise a pre-seed fund.
We started with working on our identity which mainly meant the name, logo, brand and a website.
We did some brainstorming on what to name the company and came up with “Trudy”, which stood for a trusted buddy. Once we had the name we used 99designs to crowdsource our logo.
For a brand, we hired a freelance designer to help us with a brand guide. What we learned here was that there is a full range of price available from pretty expensive to pretty cheap but using the right tools you can get 80% quality work (which is good enough if you want to launch soon) at a fairly cheap price. Another advantage of using the crowdsourcing tools is they are a lot more reliable in terms of time versus a design agency or a freelancer.
With an identity in hand, we wanted to launch a website so what we start waitlisting users. We used instapage to build and launch a website quickly. We also made our basic social accounts like FB, twitter, and LinkedIn. We used a service called teardwn to make sure the copywriting on the page was good.
On the product front, we started with writing a product requirement document (PRD) and making sure the product was possible by doing a tech feasibility analysis.
We hired a freelancer to build a proof of concept and I started building our classifier to flag cyberbullying. For building the classifier I got access to any public dataset available on cyberbullying. I found a few with the major one being the Wikipedia Talk Page labeled dataset which consisted of 260K annotated comments. Using this a few others I was able to build a bunch of classifiers and the one we finally picked did fairly good on a test dataset.
With the help of the part-time designer and our freelance developer, we were able to build the end to end working prototype in these 2 months.
The other thing we did at this time was to continue our User Research. Our last survey had given us fairly quantitative insights and to get more qualitative insights we decided to telephonic interviews with parents. We used a service called UserInterviews which connects you with targeted users for an interview for a fee. What we heard here was pretty much consistent with our understanding and helped us refine our target audience.
With a proof of concept, a digital identity, market and user research we were ready to raise money.
Being an EIR at Human Ventures, raising pre-seed for me meant convincing the parent company that the idea was good enough to put the VC money into it. So as a final thing we created our pitch deck and asked for dollars. We were funded.
After raising the funds, we spent the first week on planning and budgeting. Using the funds, over the next 1 month I was able to put a small team of part-time and full-time people to take care of various functions.
Our team on slack looked something like this
Everyone on the team except our Growth Hacker and me was remotely working either from India, Romania or USA
Our most immediate goal was to release the product so what we start our build, measure, learn cycle.
An incident which happened during this time was our lawyer applying for a trademark on Trudy came-back with research saying it will be hard to get a trademark as someone already had a trademark on Trudy in the tech space. This resulted in us losing almost a week in finding a new name and moving all our digital properties to it. After using another crowdsourcing tool squadhelp. We received 1000+ name suggestions and we ended up using Kindigo which was a cool wordplay on Kin-to-go, Kind and Indigo. This service helped us do an audience survey on our shortlisted names and also do trademark research.
As we were working on hiring a team, finishing our prototype to something which can be released, we started running Facebook ads with 2 fold objective of getting uses waitlisted and also learning which demographic responded the best to our ads. We did probably some 100+ ads tinkering with the image, messaging and the audience to achieve the same.
The other activity we started doing was interviewing kids. We realized that we have not put enough effort into understanding the point of view of the most important stakeholder in this problem: the child. We reached out to parents who had signed up and we had interviewed earlier to allow us to talk to their kids.
The biggest learning out of this exercise for us was, which again in retrospect looks obvious was kids do not like escalating things to parents till shit hits the fan. Kids try to handle these situations themselves and only think of talking to their parents when they think things have really gone out of control.
Another notable thing which happened during this time was that the outreach we had done to experts in this space bore fruit. We got meetings with
Also what I realized was whatever technology we built, it will be really hard to identify Cyberbullying. Take this example, a child bullies another child in school by calling them a pig. Now once they go home, the same child posts a picture of a pig on the other child’s facebook. To an ordinary looker, an image of a pig can be an internal joke or kids being kids, but all the other classmates know that the child is being called a pig through this action. There is no artificial intelligence algorithm which can detect that.
The fact that technology might not solve this also resonated with my understanding of the problem which has been gradually increasing with time spent in this space.
What I realized was that we were building a tool which will give more control to parents on their kids and probably make them happy about taking action but would not solve the cyberbullying problem.
I shared my apprehension with our investors.
Another big blow happened when we were just 4 days away from releasing our app and as we’re submitting our app to Facebook for review the Cambridge Analytica scandal happened. You can read more about how it impacted us here but in short, the ability to not support Facebook as a platform meant our pilot would end up being fairly ineffective.
We started this month knowing that we missed our target of releasing the product and decided to release the product a couple of weeks late by releasing Instagram monitoring instead of Facebook in the MVP.
Another blow came shortly when we realized that Instagram decided to cut access to all its API.
We found some non-scalable hacks and released the product with Instagram (we scrapped by storing user passwords), Twitter and Youtube as supported services in our MVP.
The remaining month was a painful journey on learning how not having facebook supported and a non-reliable integration on Instagram made our product really ineffective.
This started with us sending emails to about 100 people out of the 400 people who had waitlisted on our website. We got 13 signups from these mailers. We were using decent analytics in form of mixpanel, fullstory to track our users and learn from them.
Based on our funnel and feedback we tried to fix our onboarding funnel by doing simple changes like removing the requirement of phone number at signup, improving our help section. We did see improvements in our conversions on users signing up and adding their child’s accounts. We also did a live session of new users using the product on a shared screen.
We had some minor wins like Kindigo blog listed as Top 10 Cyberbullying Blogs and Websites To Follow in 2018 by Feed post. We also release support for new services like Gmail, Reddit, Pinterest etc.
After a month of optimizing user onboarding, talking to users we could see that with facebook not supported and Instagram being a hack the product was not working for the user. We could have continued with optimizing the signups but the core of it was not working.
The above combined with the fact that we knew this product was not the solution to the problem we had head out to solve made little sense for us to continue the product.
We were left with the option to either pivot to a new product or altogether shut down the company.
There was no exciting pivot in sight. This coupled with the fact that we were unable to bring a co-founder onboard made doing the pivot a really hard choice.
We had only spent around 10% of the funding we had raised by now and with not enough firepower in me to run this alone, it felt to me that returning the money was a responsible thing to do.
I finally decided to shut the company and not use any of the remaining funds. During this whole time, my investors at Human Ventures were super understanding of my situation and never made it feel like this was a disaster. Instead what we probably learned was a lesson in failing early.
P.S: This post was written in June 2018 and some of the content like the website of the startup is no longer maintained/available.
The right way to fail?
Research & References of The right way to fail?|A&C Accounting And Tax Services