Reader — How I Make Hard Decisions ×

At hexafarms, some customers asked for video processing. They thought it would solve a problem they had, so I built it. The feature worked well technically, but customers rarely used it and keeping it running took engineering effort. I decided to drop it. What stayed with me was how quickly I’d gone from hearing a request to building a feature. I still needed to understand the customer’s problem well enough to tell whether video processing would help.

I read Decisive by Chip Heath and Dan Heath because I wanted to get better at making these calls. I like to move fast, and sometimes I make a bet before I’ve looked closely at what I’m betting on. The book gave me a few ways to think through those choices, whether I’m deciding what to build or what to do next in my career. Decisive on Amazon

Look for a third option

I tend to frame decisions as yes-or-no questions. Should we build video processing? Should I stay at this company? The book suggests imagining that the options in front of you disappear and asking what you would do then.

I saw what a wider search could do when we were deciding where to take hexafarms. We started with vertical farming. Companies were interested, but that interest wasn’t turning into buying decisions. When we talked to more growers, we heard a problem worth pursuing beyond vertical farming. Growers needed more reliable ways to predict how much they would harvest. We broadened our focus to horticulture.

I wish I’d applied the same thinking to video processing. If that feature were off the table, I’d have had to ask the customer to show me what they were trying to get done and where they got stuck. That might have surfaced a simpler way to help.

Get some distance

At VOIDS, we spent a lot of time discussing how to make our agent more useful to customers. I kept coming back to what the choice would look like in three years. I expected customers to have an easier time connecting tools from different providers and building their own solutions. If our product handled just one step, they could connect that step themselves and carry on with the rest of their work elsewhere.

Once I started interviewing customers and asking them to walk me through their work, I could see which steps still sent them back to Excel. That changed what I suggested we build. I proposed covering those missing steps so customers could finish more of the workflow in one place. When I’d been looking at the agent on its own, all I could see were ways to improve the step it already handled.

The book suggests asking what advice I’d give my best friend in the same position. I use a similar question for career decisions, like when I left Hyundai Mobis to study robotics. When I look back in three to seven years, will I be glad I spent those years on this work, and what choices will I have then?

Run a small test

This was the idea that hit me hardest. With big decisions, like changing companies or buying a house, my instinct is often to make a bet and see how it goes. If there’s a small experiment that can tell me something before I commit, I think it’s worth doing.

Early at hexafarms, an agronomist on our team switched between weather forecasts, sensor data, and crop images to assess yield. Over a weekend, I built a prototype that put outside weather, sensor readings, and fruit counts in one view. She tried it and told me the charts were too crowded to read easily. I put my time into improving the view. She went on to use it regularly, and we eventually integrated it into the main product.

If a customer asks for a feature like video processing again, I could try delivering the result manually for one customer before building the system. Do they use the result in their work? Do they ask for it again? That would tell me something about whether it’s worth engineering.

Decide when to revisit the bet

The book’s fourth idea is to prepare to be wrong. If I go ahead with a feature, I can decide early what would count as a sign that it’s helping and when to check. At hexafarms, I would have wanted to see customers use the output in their regular work. I could have asked what they were doing with it before committing more engineering time.

I’ll still make bets. Before the next big one, I want to ask what I need to learn and whether there’s a small way to learn it now.

I shared my experiences and reflections in an interview, and AI helped turn that conversation into this post.

Keep reading

More in Decision-Making

  1. Consistency vs The Power of Knowing When to Walk Away

    · 1 min read

    Two startup truisms collide: 'consistency wins' vs 'know when to quit.' What survivorship bias hides about both, and how I've come to read the difference.

  2. AI is everywhere, even in interview

    · 3 min read

    It is expected to use AI in the technical interview, but it is even used when not allowed.

  3. SteelMan

    · 1 min read

    People think in a similar manner

  4. Relation of Ignorance and Evaluation

    · 2 min read

    It is not necessarily bad, even if you couldn’t understand it.

📂 All 6 posts in Decision-Making