Walid Daniel Dib is the co-founder and CEO of hala, a UAE-based insuretech startup

Spoiler: There was absolutely no grace involved.

If you’ve been in the Middle East and North Africa (Mena) startup ecosystem, you’ve definitely heard that “90 per cent of startups fail before they do X or Y”. A report conducted by CB Insights on US startups suggests that “70 per cent of upstart tech companies fail — usually around 20 months after first raising financing (with around $1.3 million in total funding closed)”.

CB Insights: Why Startups Fail

I am fascinated by the above graph. Based on a survey with 101 founders, the most likely reason for failure was no market need, at 42 per cent. The least likely reason (7 per cent) for a startup’s demise was due to a failure to pivot in time.

To clarify, not all pivots are great. The above graph also shows that 10 per cent of founders failed because of a pivot gone bad. Still, founders were four times as likely to fail due to a poor market fit.

Having been in the exact same position myself just a few months ago, I can imagine only two scenarios for this. Founders in my shoes who have closed up shop were either:

1. Too stubborn to admit that their hypothesis was poor, and decided they’d rather burn their cash to prove their point than switch tracks; or
2. Too worried about the repercussions of pivoting. Questions that have personally kept me up at night include: What will my existing and future investors think? What about my existing clients? What if we fail once we pivot? What will happen to my colleagues then?

Mistake #1: building a solution looking for a problem

The most eventful memory of 2018 for me was co-founding Addenda, the B2B precursor to hala insurance. My brother Karim and I formally quit our jobs in July of that year and started Addenda with the intention of disrupting back office workflows between insurance companies. 

the now defunct www.addenda.tech landing page

Addenda was the result of our obsession with blockchain and decentralisation. Armed with just a few years of experience in the insurance sector, I drafted a vague business plan showcasing how much insurers can save on operational costs by using blockchain.

The idea was this: in a region oversaturated with insurance companies that agree on so very little, we thought we could capitalise on their lack of trust between each other. What we didn’t know at the time was that blockchain was just the foot in the door to get insurance CEOs to sit down on the same table and cooperate.

Everyone won. Insurers got on front page headlines being “innovative” by transacting on-chain, and we got to learn the actual problems and inefficiencies insurance companies face directly from the inside.

Hint: their problems were not a result of the lack of blockchain in their day-to-day lives.

Fast forward a month

Addenda was accepted into Dubai’s Fintech Hive Accelerator programme in August 2018, where we gathered a small team of blockchain experts and decided to make a big dent in insurance, regardless of how. By the end of 2018, Karim and I began to deplete what little savings we had left. Unsurprisingly, building the blockchain was not the hard part - finding the use case was!

As we progressed, we realised we had no specific focus. Which insurance line of business were we trying to disrupt? Were we after reducing fraud in life insurance? Did we want to speed up payouts in travel insurance claims? Or was Addenda all about transacting payments between brokers, insurers, and re-insurance companies?

To say that we pivoted a few times while we were still a B2B company would be an understatement. Below are some samples of unsuccessful prototypes that we later identified as “solutions looking for a problem”:

This was the first time we used the Hyperledger Fabric distributed ledger

What it was: A blockchain prototype where encrypted data representing life insurance policies was shared in real-time between insurance companies. With this tool, underwriters would have been able to check the history of policy applicants via their Emirates ID. For life insurance, a minimum viable product (MVP) would have prevented problems that plague the industry: from anti-selection, non-disclosure, to over exposure.

What we wish we made: We really wanted to use this for medical insurance. A decentralised layer of medical history would have really reduced the onboarding process time and underwriting cost due to policy applicant history being available pseudonymously on a distributed ledger. To tackle privacy concerns, we wanted to use over the top players (OTP). To look up an Emirates ID, the policyholder would need to authorise the insurer’s search via a four-digit SMS. We thought we could get into medical insurance through a life insurance prototype, but most insurers were worried about privacy and regulatory constraints.

Claim process blockchain

Proposed screen for filing a car accident on the Dubai Police app

What it was: This was a really cool idea that I wish actually worked out. Along with our insurance partners, we wanted to push this prototype to the Dubai police. The idea was to have a dashboard for all of the people involved in a car accident, from the policyholder and insurer, to the repair shops on the backend. You would know where your car was, whether a claim was accepted, what the status of repair was, and when to pick it up again.

Why it failed: Barely any garages or repair shops were willing to use another layer of software (extra operation cost) to report car repair statuses to insurers. We also never really made proper contact with the Dubai police.

Facultative (FAC) close-out blockchain

What it was: This prototype was another solution looking for a problem. We built this with the intention of speeding up VAT payment reconciliation between brokers, insurers, and reinsurers. This pivot never really took off beyond proof of concept (PoC).

Why it failed: it was a stupid idea.

Car accident history blockchain

What it was: We called this the “Addenda Engine”. This proof of concept involved eight or nine UAE insurance companies and was used to track the history of a car’s accidents via its chassis. I genuinely thought this would take off. We started with a sample size of 400,000 accidents and identified several cases where the same car had been totalled several times, indicating a serious fraudulent case. For reasons I do not want to disclose publicly (if you know, you know), we never proceeded with this PoC.

You’d think at this point, we would have let go of blockchain PoCs, right?

Wrong. Here’s what happened instead: