top of page
  • Thomas Thurston

Businesses are like ants

Updated: Jul 24, 2023

Business is complicated. Sometimes businesses do everything right, but still flop. Other times they seem to do everything wrong, but succeed anyway. Predicting if new businesses will survive or fail has turned out to be one of the most elusive tasks in human history. So it might be a surprise to learn how computer-based research and modeling suggests businesses are more predictable than we think.

Businesses are run by humans, and humans can be hard to predict. Yet rather than modeling businesses as smart, creative, multifaceted entities, the most accurate computer models thus far have been those which assume businesses are single-minded drones.

We’re starting to learn that businesses are less like humans, and a lot more like ants.

ants - licensed

Ants aren’t stupid, they’re focused. Foraging ants are divided between “scout” and “gatherer” ants, with complex feed-search processes. The zig-zag pattern used by scouting ants is highly efficient, and their phermones create a robust communications system.

Jurgen Kurths, physicist, mathematician and professor of nonlinear dynamics at Humboldt University said “while ants can appear chaotic and random-like, they very quickly become an ordered line of ants crossing the woodland floor in search for food… These insects are, without doubt, more efficient than Google in processing information about their surroundings.”

While ants are far more sophisticated than they appear, the goal is simple: find food, bring it home.

Similarly, despite the chaotic-seeming behavior of new businesses, the goal is simple: find profits, bring them home.

The strategies used by businesses to pursue profits can zig-zag. Tactics range from straight-forward to elaborate. But if you know where the profits are, and assume businesses will mindlessly pursue them, you might be surprised by how predictable company behavior can be. For example, you can better predict which competitive battles they’ll win or lose, which growth initiatives they’ll prioritize or pull the plug on, whether they’ll grow margins or end up commoditized, and a variety of other insights.

Of course our ant-like models aren’t always right. Sometimes luck intervenes or the businesses simply don’t do what the models predict. Like any predictive model, there’s a margin of error. Yet despite the exceptions, ant-like behavior seems to be the rule – which is a good place to start.

Using computer models to predict business behavior is controversial for some people, but it’s worth noting how traditional, subjective, qualitative predictions have an abysmal track record. Roughly 70% – 80% of new businesses fail before their tenth birthdays, regardless of whether they’re launched by multinationals, venture capital dollars, startups or even micro-businesses.

In fact, studies on the accuracy of expert predictions find business managers and stock forecasters to be among the least accurate of all experts sampled (less accurate than livestock judges, grain inspectors, photo interpreters, soil judges, nurses, physicians, student admissions officers, parole officers, clinical psychologists, psychiatrists, etc.).[i] Mechanical models of company behavior are simply proving more consistent and accurate than old-fashioned subjective guesswork.

In business innovation, perhaps the biggest contributor to poor expert predictions has been the failure to recognize it as a signal versus noise issue.

New business innovation has historically been analyzed through descriptive approaches – storytelling – that rewards one for painting as complete a picture of the past as possible. In other words, the study of innovation has focused on completeness; noise and everything. It’s about adding variables, not eliminating them. Did Bill Gates have butter or margarine on his toast the morning before he talked IBM into licensing DOS? Did he have toast at all? How might those details have made a difference?

You get the idea.

Rather than valuing completeness, predictive modeling is obsessed with accuracy. Completeness only matters to the extent that it impacts accuracy. In a predictive mode, the goal is to eliminate noise to focus on the fewest variables with the most predictive weight.

In a perfect world, it’s best to be both complete (consider everything) and accurate (predict everything). Both are important, and there’s a virtuous relationship between them. Yet, if I’m being honest, most of the time description and prediction are – practically speaking – opposite ends of a spectrum. They’re almost always at each others’ throats.

While “knowing where the food is” helps predict ant behavior, there’s another critical piece of the puzzle. In Sciences of the Artificial, Herb Simon says “viewed as a geometric figure, the ant’s path is irregular, complex, hard to describe. But its complexity is really a complexity in the surface of the beach, not a complexity in the ant.” In other words, the ant’s path is largely determined by the twigs, pebbles, holes, hills, spiders and other obstacles it must traverse on the beach.

So if you want to predict what the ant will do next, focus on the beach – not the ant itself.

We’ve found the same to be true when modeling businesses. Around 80% of the predictive value seems to come from market externalities (ex. customers, competitors, technology trends) rather than details internal to the company itself. In business we’re used to obsessing over internal minutia, including every detail of a company’s strategy, tactics and leadership. While this matters, to be sure, we’re finding it only accounts for around 20% of the equation.

Trying to model and predict new business behavior has been a surprising journey, with all sorts of unexpected and counter-intuitive lessons along the way. One of these surprises (at least for me) was how little business predictions have to do with the things I used to think were all-important such as leadership, teamwork, personality and internal nuance. While these things matter, statistically the bulk seems to lie outward – with understanding profits, competitors, customers and external trends.

When trying to predict business behavior, our research suggests you should:

First assume businesses are single-minded, profit seeking drones.

Secondly, it’s more predictive to focus on the market landscape a business must traverse, rather than minutia about the company itself.

Businesses, it seems, are more like ants than I ever would have guessed – and maybe that’s a compliment.


[i] Shanteau, What Does It Mean When Experts Disagree, Kansas State University


bottom of page