Learnings from my failed startup

I built a Shopify app. Made mistakes I didn’t see coming. Here’s what I’d do differently if I started today.

What I was building and why

MagicSell is an AI upselling and cross-selling app for Shopify stores. The idea was real — most e-commerce checkouts leave money on the table. Cart abandonment is massive. I knew this was a problem worth solving.

So I started building. I hired a team. We moved fast.

That was mistake number one.

The 3 mistakes that actually killed momentum

I built before I sold. I spent months on the product before talking to a single potential customer properly. Not a quick call — a real conversation where I tried to close them on something that didn’t exist yet. I assumed the problem was obvious enough.

The co-founder situation. I picked up someone I thought I needed. Shared equity. Things moved in different directions. I had to unwind it — slowly, awkwardly, and later than I should have. That’s a whole other post.

I hired too fast. At peak I had 8 people on the team. I hadn’t validated the product with paying customers yet. The overhead — salaries, coordination, context-switching — slowed everything down when we needed to be moving light.

What I’d validate before writing a line of code

If I started again, here’s what I’d do in week one:

  • Write a one-paragraph pitch and DM 20 Shopify store owners on Twitter or LinkedIn
  • Send them to a landing page with a waitlist
  • Try to get 5 people to say “I’d pay £X for this”

If I can’t get 5 “I’ll pay” responses before building, I’m not ready to build.

The fake door test is underrated. A simple landing page describing the product — before the product exists — tells you more than 3 months of building.

What I kept: skills, network, clarity

I’m not complaining. Everything I learned building MagicSell the wrong way is why I’m rebuilding it the right way.

I know the Shopify ecosystem now. I know what store owners actually care about. I know how to scope a feature without a team debate. And I know the exact failure modes to avoid.

I’m rebuilding it solo. Leaner, faster, more focused on distribution than features.

The takeaway

Don’t hire before you have paying customers. Don’t pick a co-founder out of anxiety. Don’t build for 3 months without trying to sell. The validation comes first — everything else follows.

Building something? Follow me on Instagram and Twitter — I document everything.