In all decisions, we have the choice of exploring (gathering new information) or exploiting (using the information we already have) to make that decision.
A classic example of the explore/exploit dilemma is choosing a meal at your favorite restaurant. You know you love the pizza, so choosing this will not yield any new information but you will be guaranteed a good meal. Or you can choose the special. You’ll gain new information about the world but risk having a meal that’s below your expectations.
New information has value but it costs time, resources, and opportunity. Is it worth it or should you just go with what you know? When the pace of change is high, the explore/exploit dilemma tells us that it makes sense to explore.
There are many ways to solve the explore/exploit dilemma but they tend to share a common tendency. With a new problem, we start out exploring and gradually converge on exploiting. This is because we acquire new knowledge then have less motivation to find new information. Additionally, if we have limited time to solve a problem, then as time runs out we’ll have less opportunity to use the new information we’ve gained.
Over the years, we’ve built up a storehouse of reports, analysis, models, and intuition for a host of problems amenable to analysis. Market forecasting, competitive analysis, and technology adoption rates are all examples of analytical challenges.
Whenever we are faced with a new analytical challenge, we start with a rough evaluation of the explore/exploit dilemma. In a market analysis for a wellness app, our first question was, what do we already know? Wellness was new so we knew very little. We had to run fast and hard.
We gathered as much new data as we possibly could. The number of apps considered to be wellness apps. Download stats. Product features. Reports on the industry and estimates of market growth. Reviews. Price. Freemium versus paid. Analysis of in-app advertising revenue.
An LLM can help you consider the tradeoffs or approaches of each in a given context. Here are some useful prompts:
In our analysis case, ChatGPT’s answers were useful to help us balance the priority of exploring versus exploiting. For us, the key benefit in prioritizing exploration was to find new features while the most important drawback was the consideration that we might not find good data. On the other hand, prioritizing exploitation would help us move faster on core development. We decided that in this case, we would prioritize exploring.
But we also wanted to consider the balance so ChatGPT’s strategy suggestions helped us consider how we could balance the tradeoff between explore and exploit.
We decided to use a threshold that would trigger the switch from exploration to exploitation. Once we collected data on 10 competitors, and conducted 50 user interviews, we switched to exploiting that data.
ChatGPT pointed us to Spotify.
We devoted ourselves to assimilating everything we could, as fast as possible. Pretty quickly, we had data we could use and useful new knowledge. We built a model to analyze the opportunity for this specific wellness app. We used the model to estimate dependencies between growth rate, customer acquisition cost, and technology cost. It didn’t take long for our knowledge to be sufficient to build realistic scenarios and make decisions.