Big Game and Agtech (about that McKinsey chart)

Anyone who has spent time in digital agriculture knows the chart I’m talking about. It’s the one that shows that agriculture is the least digitized industry, on par with hunting. Hunting? I suppose there’s camera traps and GPS apps nowadays, but my memories of hunting were all horses, boots, rifle, and walking.* Come to think of it, agriculture is mostly pickup, boots, clipboard, and walking. So I guess they are not far off.

The points people make with this chart are varied, but usually it amounts to “gosh, ag seems like it’s poised for digital transformation.” But you could also wonder: Why is this chart true in the first place? Why isn’t ag up there at the very top?

I had a conversation recently that helped me understand this chart better. In the desk behind me sits Mamadou Diao. He designs auction systems for online advertising, first at Facebook, and now under his own banner. He writes code that for any given moment will determine what the ROI is for an ad placement, and adjust strategy such that only profitable ads are placed. Imagine that. You could place dozens or millions of different ads, in different venues, addressing different search terms, and Darwinian selection would leave you with a rigorous online marketing strategy that only made money. A hammer that only hit nails. That, in essence is what Google and Facebook do: by enabling a direct calculation of ROI, they give you what amounts to a license to print money.

And thus, we find that ICT & Media, aka online advertising, are at the very top of the list of digitization.

In ag, by contrast, you can use a product for years and not know whether it works. I met an Iowan farmer earlier this summer who had done variable rate seeding for at least a decade, and he still wasn’t sure if it paid off. Some years better, some years worse, it seemed like the signal was within the noise. At our company, we could supply you with irrigation advisory that has been scientifically shown to boost product quality, and thus price per ton of product, but that’s no guarantee that the user’s irrigators will follow the advice, and receive the benefits. And thus, it is hard for farm owners to know what a dollar of investment in digital ag gives them in return. Ten bucks? Ten cents?

Adam Bergmann, most recently of Wells Fargo Cleantech, likened this to the dawn of the solar industry, where the performance was so poorly examined or understood that an entrepreneur pitching an innovation couldn’t make headway because none of their audience knew how their performance was different from the potential. And so, prospects couldn’t know whether they were being offered snake oil or the real deal. Twenty years later, a company could pitch an anti-dust coating that improves performance by fractions of a percent and the customers know what the problem is, and can readily calculate ROI and time to payback.

It’s a problem that has bedeviled the adoption of digital tools in agriculture, and one that must be taken seriously. It’s not simply that there must be an ROI but that the customer can immediately calculate the ROI. Some have called this “the sales-ready product.”

A couple thoughts on what this means for product strategy in our industry:

• One strategy is for agtech to focus on smaller value propositions that have a more immediate verification. If I tell you right now a pump is broken, you can verify that right now. If I predict that this inversion should break up in an hour so you can spray, that can be verified in an hour. If I predict the sugar in these grapes will be 2 units higher tomorrow, you can verify that tomorrow. If I can tell you a heat wave is coming this week and you should get out ahead of it by turning on the drip lines now, you can verify that soon enough. Note that these don’t pay dividends in cash per se, but in time/attention or reduced risk. This is essentially the strategy that Silicon Valley has pursued for making products that immediately pay off, gradually overcome your skepticism, and ultimately entrain your full reliance. I never question anymore whether Google Maps is giving me a good route forecast, because the emotional memories of being on time (while cutting it close) are etched in my brain. One timely example of this strategy is WeFarm, a knowledge sharing network, which returns a response to a farmer query on average in 13 minutes. It’s not hard to understand why they are allegedly the “fastest growing agtech startup ever.”
• Another strategic direction for agtech is in benchmarking. Some of the most powerful tools I have seen allow users to play “what if” they had chosen the right variety and seeding rate instead of what they actually planted. It’s a kind of prediction of the past with immediate interpretation. Farmers Edge’s lets a user compare their yields against other users in comparable soils and weather. Assuming operator error is not a factor, knowing the grass really is greener on the other side of your fence is powerful stuff. Climate Corp has a tool that enables users to select an area of interest in a field, such as where they planted a different variety or different density, and directly know the outcome of that experiment. By making it easy to learn the results of the experiment, the potential gain or loss of revenue from the alternative strategy can be calculated: an ROI.

I’d argue that these are both forms of predictive analytics, and also note that both make use of in-field data: a ground sensor in the first case and a tractor plug in the second case. What I think is happening is that agtech is catching up, 10 years later, with the smartphone revolution that ushered in location-aware predictive analytics for consumers. When I search for coffee, I should get back a search result of coffee shops, nearby, which are open now, in essence making a prediction for where I might walk to next. In agtech, we are soon going to be in an era where similarly terse queries, say “harvest timing” or “best maize seed” or “spray now” returns something from our apps that knows the day in the season, the time of year, our place on earth, and returns an answer that is close to exact. Given the reality that IOT based predictive analytics in ag is coming out of its squirrely adolescence into respectable maturity, I am really excited about the user experience and business model innovations to come out of it, to truly drive adoption of digital agriculture at scale.

Postscript:

* Jeff Reed had me convinced, and it may even be true, that he used the AWS DeepLens AI-ready camera to identify deer walking across his lawn and only send an alert if it was 6 points or higher. It’s at least as sophisticated as anything I’ve ever seen in agtech, or Google Lens for that matter!