Last week, I wrote about why we decided to pursue a new direction. Now I want to give you a behind-the-scenes look at how we did it. Strap yourselves in, this one’s a bit longer than usual.
Our “Aha!” moment was liberating, but it introduced a new set of challenges. Almond pollination only happens once a year, and at the time of our “Aha!” moment, the next pollination season was less than 10 months away. We had to get something out there, otherwise we’d wait 22 months until bringing new revenue in the door.
With less than a year to build something our target market was willing to pay for, we had to focus. As much as we wanted to develop that beautiful product we envisioned, complete with all the fancy bells and whistles, there simply wasn’t enough time.
We couldn’t afford to be good at all the things—we had to be exceptional at one or two important things. We narrowed in on two guiding principles: the product had to be accurate and it had to be faster than manual inspections.
Much of our time during the development phase was spent on customer discovery. We found a handful of almond growers who weren’t annoyed by our monthly phone calls. These folks were our sounding board. They would tell us if we were still on the right track or whether a course-correction was needed. Ultimately, we had to strike a balance between what they told us and what was achievable.
Accuracy is our first guiding principle for a good purpose. If it’s not accurate, it doesn’t matter how fast the product works. Before we began to think about building speedy software, we started outlining the predictive hive strength model.
The first step towards building an accurate model is to gather data. From May to November, several times a week, we’d set out at the crack of dawn to image hives under infrared. Over the course of this time, we’d end up capturing tens of thousands of infrared images. But it wasn’t as simple as taking a photo and moving on. There were a lot of variables to test in order to find what produced the best results.
That leads into a story about one of my wildest experiences in my 3+ years with The Bee Corp. In order to test the accuracy of our model over time, we conducted a handful of "marathon" data collection studies. This entailed capturing IR images of the same hives every few minutes over the course of several hours.
One muggy summer night, I was tasked with doing a marathon study all by myself from dusk to dawn. I loaded up Ellie’s car with enough sugary snacks and Red Bull to cause a heart murmur and set out to our bee yard tucked away in the rolling hills of Southern Indiana. No cell reception, no WiFi signal, no escape from the dense humidity; just me and all the bugs you can imagine.
I wish I could say it was eventful. It wasn’t. It became a dreary routine: pause my podcast, turn on the camera, grab my flashlight, exit the car, take photo 1, photo 2… get back in the car, shut off the camera, play my podcast for 10 minutes and repeat. By the time the sun came up, there was a visible rut in the grass from tracing the same path 50 plus times that night.
The worst part honestly wasn’t even all that bad. Running on 20+ hours without sleep and with tablespoons of sugar and caffeine coursing through my veins, paranoia set in rapidly. Anytime I heard a strange noise in the woods, I was certain an axe murderer or a grizzly bear was just steps away. But I had a secret weapon: my infrared camera. Equipped with the superpower of night vision, I had confidence that I’d be able to get the jump on any bears looking for a quick snack.
A fun note: I ended up doing 2 overnight marathon studies, once in August and again in October. The second time was during Game 3 of the ALDS—and I’m a HUGE Red Sox fan. In the middle of the woods I was somehow able to tune into Nate Eovaldi and the Sox dish out the most lopsided pounding in Yankees postseason history 😊.
Thanks for sticking through to the end on this one. I hope you enjoyed it. Check back next week to read about Ellie’s takeaways from the February launch.