Ben Silbermann quit his job in 2008.
After leaving his job, Silbermann built a shopping app called Tote: a virtual replacement for paper catalogs. But he soon learned that he had included too many features in the app right from the outset, which made it difficult to ship. Also, at the time, mobile payments were difficult. So although the app had many users amassing large collections of their favorite items and sharing them with their friends, it didn't have many paying users.
But observing this behaviour led to Silbermann launching Pinterest with his two cofounders in 2010. His initial growth strategy was to target his techie friends and let them know of this new website that allowed them to pin, organize, and share things they found on the internet that they loved.
"What's the point?" was their reply. No one in their tech social circles seemed to be interested in such a thing.
But as any experienced qualitative user researcher would tell you, they were still very polite about it.
"This looks interesting. Very interesting," they’d say.
Silbermann then went to every Apple store and switched the default home page on every MacBook there to Pinterest.com. This still didn't seem to work.
Slowly, he came to the unappetizing realization that maybe tech people were the wrong customers to target. Because while he faced ample rejection, he also observed a small group of people who seemed to love it.
“There was a small group of people who were really enjoying it. Those folks would not be who you’d think of stereotypically as early adopters. They were folks I grew up with. People that were using it for regular life stuff: What is my house going to look like? What kind of food do I want to eat?”
To find more of this kind of people and build an initial community, he decided to attend a large women-focused design and blogging conference in Salt Lake City. It was here that he met Victoria Smith who then ran a blog called SFGirlByBay. She quickly hopped on the platform and invited other bloggers to do the same. Just a year later, Pinterest was at over a million users — with everyone using it for creating mood boards around their hobbies and interests.
"We didn't have an engineering problem. We had a design and community problem."
Now, while this case study can be about how Pinterest found its target customer persona, I want to highlight something quite different about it.
Which is, the difference between gentle and harsh environments; the difference between simple domains and complex ones.
A game of chess is a simple domain. It has a fixed set of rules. No moves are possible or allowed outside of its fixed and limited set of rules. All the variables affecting the gameplay at any point in the game are known. There are no unknown variables. The map is the territory.
A real-world endeavour on the other hand, like launching your own business or charting your career trajectory, lies in a complex domain.
In a complex domain, not only are there many unknown variables, there are variables which you don't know that you don't know: unknown unknowns.
This simple distinction changes the entire nature and way in which you approach a simple game like chess versus a complex domain like business. Let me illustrate some major differences.
Simple domains have tight feedback loops. Complex domains have loose feedback loops.
A tight feedback loop means that effects can be easily mapped to isolated causes and the effect follows the cause quickly. It's like solving a sudoku puzzle. Everything is straightforward and logical. And it is simple to map your way to the outcome. Results of actions are immediate and measurable.
To achieve outcome C, you take step A followed by step B. Learning in simple domains has tight feedback loops.
In contrast, complex domains have multiple causes leading to a single effect. It is often not possible to attribute a single cause or set of causes that led to the effect.
For example, a startup is concurrently running multiple marketing experiments at once. If in a week, it suddenly sees a spike in number of leads or sign-ups, it may be hard for it to attribute exactly what marketing initiative or branding activity led to the spike.
Last-mile attribution using UTM links only tells you what link the user clicked to sign up. It doesn't tell you what series of actions led them to click that link or what bunch of factors convinced them to click the link in the first place. It is only after a long period of sustained and consistent effort over a few months do you realize which marketing activities are effective and which aren't.
In short, marketing is a complex domain with loose feedback loops. Learning accrues slowly. And to not derive false conclusions from mistaken causes, you need to take time to observe all the variables in the system at play — to learn how they interact with each other.
Tight feedback loops are good in gentle environments with simple rules, where outcomes are clearly specified and simple frameworks work well and provide immediate feedback.
But they are not good in harsh and complex environments where following simple rules can lead to measurebating. Using tight feedback loops in harsh environments makes people focus on short-termism instead of taking long-term approaches.
Imagine a situation where you are a manager, your team is understaffed, and you are behind on your quarterly goals. It is extremely tempting in such a scenario to temporarily drop your hiring standards, in order to get people in quickly so you can hit your goals. A temporary hiring frenzy can create tight feedback loops and put numbers on reports, but did you make a good decision in the long run?
Likewise, in the case of careers, life-changing decisions are full of uncertainty. While taking the leap, you do not know where you may end up. There are too many unknowns. When you're starting out, you might still be able to learn simple skills that are easy to measure progress on and work on tight feedback loops. But once you achieve a certain level of expertise and are working with large complex systems, you only have loose feedback loops with learning that occurs in spikes versus in a linear upward graph.
In gentle environments, optimize for efficiency. In harsh environments, optimize for resilience and robustness.
When Silbermann started Tote, he approached the problem top-down and formalized the system too quickly by adding too many needless features that made the product hard to ship.
But in harsh environments — the market being one of the harshest and most complex ones — formalizing too quickly without understanding the territory can lead to adverse consequences. When you plan to operationalize and scale the system, first make sure you understand the problem space well enough to know how things actually play out in practice.
Because formalizing and scaling up the wrong thing can be expensive and irreversible and lead to huge amounts of wasted time and capital. You can get caught in headwinds created by your own rigid processes and systems that don't gel well with reality.
So, in harsh environments, it makes sense to pursue things bottom-up and organically, starting from simple systems and eventually growing them into large and complex ones over time, as you accrue learning from experiments on the way.
It is not a game of chess, where you can have detailed strategies and pre-calculated plays for situations. The map is not the territory; heck, you don't even know what the territory is going to look like a few months or years down the line, let alone have a high-resolution map for it.
Structure is a great tool to help reduce complexity and enhance predictability. However, humans crave clarity and detail, which means we tend to apply structure too early while dealing with harsh environments.
Too much conviction and process in your product even before you've put it out in the market and tested its demand prevents you from being nimble enough to pivot.
In gentle environments, plan and execute meticulously. In harsh environments, be open, observe, and follow trial and error.
What would have happened if Silbermann had been too closed off to the idea of pursuing a different market segment? What would have happened if he overoptimized Pinterest for techies and invested in features that his real target consumers didn't care about? We likely wouldn't have seen Pinterest be the product it is today.
For a game like chess, planning is more about optimizing and pursuing efficiency.
For a complex system like a market, planning and strategizing is more about creating more space to run even more experiments and understand the territory and its incentive structures better.
A grandmaster isn't experimenting in a world championship. All her strategies are planned and memorized. They only need to observe the board and decide which strategies to apply when.
But an entrepreneur like Jeff Bezos would say that it's still Day 1 and that he doesn't fully understand the territory.
In such cases, simple trial and error can reveal more information about the territory and unknown variables versus approaching problems top-down with an aim to formalize.
In gentle environments, centralize for efficiency. In harsh environments, decentralize for absorbing local context.
"People closest to problems were usually in the best position to solve them."
— Jeff Bezos
In a simple system, all the variables are known and can be played with accordingly . But in a complex one, understanding of local context can be the difference between something that worked versus something that failed miserably.
For a good example, think about all the startups who blindly copied and launched something that worked in the west, only to find out that Indians had salient differences in behaviour and incentive structures which meant that those products would never work here.
Failure to work with local context and enforcing processes top-down will often lead to employees working only for the sake of following processes and taking decisions aligned with that objective rather than those that work in the long-term interests of the company.
Conclusion and summary
In gentle environments
- All rules and variables are known: The map is the territory
- Simple strategies, techniques, and frameworks do the job when executed precisely
- Allow for clear metrics
- Root cause analysis is easy: isolating and attributing causes is possible
- Optimization and efficiency are the end goal
- Have Tight Feedback Loops and track metrics often
- Centralization works
- Top-down protocols and processes work from the very start
In harsh environments
- Many variables are unknown
- Trial and error works better than a top-down strategy
- Formalizing too early may lead to adverse consequences
- Bottom-up strategies that grow organically with the system work better
- Root cause analysis is hard: no isolated causes can be mapped, there are often multiple contributors
- Chase resilience and robustness to operate in uncertainty'Have Loose Feedback Loops, measuring the right things correctly takes time
- Have Loose Feedback Loops, measuring the right things correctly takes time
- Centralized decision-making is hazardous, local context is crucial
Treat this piece as your introduction to thinking in complex systems if you're used to thinking in simple hacks and mindless planning. They often do not work. And the next time someone uses chess as an example, tell them it's a bad analogy for thinking about businesses and careers.