Do one thing well

It was another day hunched over a small desk in our budget WeWork. Revenue: <$5k ARR. I had a prototype on screen share with a customer that has taken a liking to what we were building.

This was 2020, AI SDRs were not a thing, and little did we know we were trying to build one. On the left hand side of the prototype, we had some email automation capability; on the right, we had what we now know as “Deep Research”: plugging a search API into a large language model. The idea was to automate prospect research, then incorporate that research into a better, more personalised sales email.

Customer says, “I am so glad you showed me this. I am really only interested in the right hand side, the deep research. This left hand side thing I already have and its not a differentiator.”

Being the first time founders we were, we were really just building a startup based on Medium articles, HN threads, and tweets. So, we heard that, pattern matched it with all the advice out there to “do one thing well”, ditched the email automation features, and ended up that same quarter with more than $100,000 in ARR.

Happily ever after, right?

Or so we thought. We rode this wave for a few more quarters, but the reality started to creep back up on us: none of our customers wanted to JUST buy a deep research capability. They wanted an application. We needed a car, but we only had the engine.

What people think doing one thing well is

As we struggled, doing one thing well became the hot topic. Are we the best at a certain type of data? A certain type of technology? A certain type of customer? How can we be more focused? If we are not focused, how can we go fast enough?

In hindsight, I made many wrong assumptions. For starters, I was thinking about focus in the wrong frame of mind. Implicitly, I believed that focusing on just one thing would unlock everything. Turns out, there is a term of this kind of thinking: people called it looking for “magic bullets” or “silver bullets”.

The idea of a magic bullet really attracted me: can we build in a technology area where we can accumulate valuable IP so good it sells itself? VMware 2.0? I may not have articulated it that way, but that was certainly the driving force behind many of my decisions. While I know at the back of my head that a magic bullet doesn’t exist, and I did not confront this assumption/habit fast enough.

This idea gave us too much comfort, because you just need to focus on building a thing and “revenue is guaranteed”. It was satisfying to be building fast and see new things come to life without “boring” things like analysing customer feedback, or watching users use your product (which, as we were not focused on customer experience, was an uncomfortable thing to do.)

To ground our search for the silver bullet a bit more, we looked at it in many ways with our advisors and our customers. Terms like “unique selling point” or “killer feature” became daily lingo.

This is how I learned, during this period, that a “killer feature” is a misnomer. We got a lot of advice to find our killer feature, and in some sense, we found various iterations on what should our killer feature should be. But customers don’t buy features, they buy solutions. “Killer feature” was an adage from a time when selling features and feeds was enough. It did more harm than good to our product. So the fact that we were discussing which “feature” was killer by default was wasting time.

So I went around in circles, unable to decide what our killer feature is/should be. A pattern started to emerge in sales calls: the prospect would be very interested in the few killer features we did offer (a good signal) – only to then pick a competitor that may not have what we have, or maybe not to the same quality, but had other “features” we deemed “undifferentiated“.

Every time this happened, I thought to myself, “ugh, vapourware wins another round.”

What people think tackling an end to end workflow looks like

The other aspect that created a lot of friction in our will to solve this problem was how HARD the opposite seemed to be. Who starts off building everything? Boiling the ocean seemed even less practical.

While we were well funded, it’s nowhere near the amount needed to build everything. Plus the risk was very high: if you tried to do everything all at the same time, you are taking a large amount of risk with very little validated return.

Frustratingly, whenever I was in conferences, other startups and products of similar vintage always seem to have a ‘bigger’ product. It looks more complete, there is more to demo, etc.

They would have prettier looking screens. There is a lot to demo. In our internal meetings, we would comment on other products in that they ‘demoed well’.

I tried to find solutions to this problem, but we seem to never be able to get past how we can innovate everything that’s required to build the holistic product surface we needed.

What “doing one thing well” actually is

In the next role, I was fortunate to work with other visionary founders. I also was part of the M&A committee, which meant that we debated and evaluated a LOT of startups. With a new perspective, I gained a more nuanced understanding of what we can do better.

TL;DR – doing one thing well = you are solving a problem in a way where one aspect of your solution is 10x better.

For example, Linear is a new project management tool. It solves the problem of managing complex software projects. A lot of Linear is very similar to other project management tools: issues, issue lifecycle, comments etc. But it’s faster to use. 10x faster.

The point is, it’s not whether you have a killer feature or not. Your tool needs to solve a problem entirely for a customer, in a way where they can migrate from their current solution to yours. Then, the motivation for them to move comes from that one aspect that you are better at. You doing one thing well, best in the world in fact, but you also solve the rest of the problem.

So your customers don’t need to rely on other solutions AND THEN yours, they can replace their existing tool with yours.

Well, how do we do this?

If you look at the graphic for this section, the first thing to notice is that the X axis has changed from feature to value. This is (hopefully) product management 101, where the emphasis is “jobs to be done” over “features”; But I think we need to go beyond this: I find that in early stage, 0 → 1 situations, you need to be even more zoomed out to ensure that you are thinking about customer value, then jobs to be done, then features. So,

1. What value do you want the customer to gain?
2. How are they delivering that value today? What are the “jobs” that they perform to deliver that value?
3. What features in your product enable those jobs to be done?

The second thing is to focus on a type of value that customers really care about. This gives you differentiation. However, you still need to defend your moat, so you will want a focus that’s easy to convey (enabling more viral distribution) but hard to execute on. This lets you generate moats, momentum and excitement the deeper you go into that direction.