26 March 2025 • Ian Reppel • 8 min

The Drake Equation for Product Management

What does the Drake equation, which deals with estimating the number of extraterrestrial civilizations, have to do with product management? Quite a bit, actually!

The Drake equation calculates the number of active extraterrestrial civilizations in the Milky Way. Frank Drake proposed it in 1961 to stimulate scientific discussion around the search for extraterrestrial intelligence (SETI). A related concept is the Fermi paradox: Why, despite billions of potentially habitable planets in our galactic neighbourhood, are there so few signs of intelligent life beyond our solar system?

We can adapt it to product management to compute the number of successful products in the market \(N\) as follows:

\[N = R_{\mathrm{i}}\, f_{\mathrm{v}}\, n_{\mathrm{s}}\, f_{\mathrm{d}}\, \underbrace{f_{\mathrm{a}}\, f_{\mathrm{r}}\, f_{\mathrm{s}}}_{f_{\mathrm{PMF}}}L.\tag{D4P}\]

What do these variables mean?

\(R_{\mathrm{i}}\) is the rate at which ideas are generated (in the same units as \(L\))
\(f_{\mathrm{v}}\) is the fraction of ideas that are both validated and feasible to be developed into solutions
\(n_{\mathrm{s}}\) is the number of viable solution options for each validated problem \(p\)
\(f_{\mathrm{d}}\) is the fraction of viable solutions that are developed and launched
\(f_{\mathrm{a}}\) is the fraction of products that are adopted by the market
\(f_{\mathrm{r}}\) is the fraction of products that actually retain a significant portion of users in the market in the long term
\(f_{\mathrm{s}}\) is the fraction of products that are scaled across the market
\(L\) is the typical lifespan of a product that has achieved PMF (in years)

Mathematically, the order is irrelevant, but in reality the equation represents a sequence of consecutive steps. With the Drake equation for products (D4P), we can clearly see why there is a dichotomy between the number of ideas out there and the lack of successful products in the market. Ideas are worthless unless they solve real user problems. Even if there is a problem worth solving, it may not have any solutions that can be turned into profitable products, because every solution must also be adopted and scaled. But prior to scaling, each product needs to retain sufficient users who have adopted it. The fraction of products with product/market fit (PMF), \(f_{\mathrm{PMF}}\), is the product of these factors, as indicated by the equation.

It is often easier to reason with the total number of ideas generated over the lifespan of a product: \(n_{\mathrm{i}} = R_{i}L\). This corresponds to what most people refer to as ideas for market problems.

Embedded in the D4P is the double diamond from design thinking:

Generate ideas to increase the problem space: \(n_{\mathrm{i}}>0\)
Converge onto the most promising problem that has been validated: \(0<f_{\mathrm{v}}\leq 1\)
Generate solutions to increase the solution space: \(n_{\mathrm{s}}>0\)
Develop only the most promising solution(s) to solve the problem: \(0<f_{\mathrm{d}}\leq 1\)

The value of these diverging phases becomes apparent in the equation: they increase the number of successful products in the market.

How many ideas do you need for one successful product?

Let’s use the D4P to estimate the number of ideas required to have a single successful digital product in the market. Note that the figures referenced are for B2B/SaaS and B2C products only. For consumer goods or hardware, the variables have different values.

\(f_{\mathrm{v}} = 0.65\): idea validation rate

A little under two thirds of problems are properly validated. While it sounds obvious to validate ideas prior to implementation, more than a third of startups fail because of the lack of a need from the market. This fraction not only includes validation, but also filtering due to prioritization and feasibility. Note that validation is an ongoing discovery process, not merely a one-off.

\(n_{\mathrm{s}} = 25\): solutions per problem

It is not uncommon for teams to come up with seventeen solution ideas in a brainstorming session. With an improv mindset that might be bumped to 25. Note that in design thinking and design sprints people aim for one hundred ideas in an hour-long session. That matches the notion of 8 ideas in 5 minutes. With 25 solution ideas per validated problem, we have a realistic yet conservative estimate.

\(f_{\mathrm{d}} = 0.90\): development rate

Out of every 7 solutions, 4 are developed, and only 1 is successful. Solutions are dropped either because of the lack of a market need or because of resource constraints, which gives us the constraint: \(f_{\mathrm{v}} f_{\mathrm{d}} = 4/7 \approx 0.57\). With \(f_{\mathrm{v}}=0.65\), we therefore find that \(f_{\mathrm{d}} \approx 0.88\). This is not far off from the fact that a lack of resources is responsible for at least 6 and up to 20% of all project failures. Let’s settle for \(f_{\mathrm{d}} = 0.90\).

\(f_{\mathrm{a}} = 0.20\): adoption rate

In SaaS, the product adoption rate is a bit under 20%. Across industries, it is nearly a quarter. One out of five is therefore a decent approximation for adoption, which, for the purposes of the D4P, includes activation and conversion.

\(f_{\mathrm{r}} = 0.10\): long-term retention rate

Across industries, the eight-week retention rate is 20% For SaaS, that number is 35% after two months and 30% after three months. Mobile apps only retain up to 10% after thirty days. After one year, we can therefore estimate the long-term retention rate for tech products to be around 10%.

Long-term retention figures for media products and games are often closer to 1%.

\(f_{\mathrm{s}} = 0.30\): scaling rate

Out of these adopted products with sufficient stickiness, maybe 30% can be scaled up to be profitable. After all, 70% of startups fail due to premature scaling.

If we look at seed-stage startups, only 30% make it to Series A, which is when they expect to see signals indicative of product/market fit. And of those that do reach that stage, half manage a Series B, which is where they expect to hit PMF and grow the business. So, 30–50% for the scaling rate seems about right.

\(L = 5\) years: product lifespan

On average, digital products last about 5 years. Google manages to shoot products after 4 years, but most companies are not that trigger-happy. Note that SaaS and B2B products tend to have longer lifespans.

Results

With what we have found so far, \(f_{\mathrm{PMF}} = 0.20\cdot 0.10\cdot 0.30 = 0.006\). This suggests that only 0.6% of all product iterations achieves product/market fit! That figure may seem extremely low compared to the oft-cited startup failure rate of 10%, but startups rarely produce a single product variation. In fact, out of the businesses that succeed, 93% pivot at least once, and each pivot naturally consists of at least one new idea, and a batch of fresh solutions. With an estimated two pivots per startup, only two or three ideas per pivot, and perhaps a dozen or so proposed solutions per idea, we can see how a single product variation’s success drops to the PMF fraction shown.

Consequently, the chance of product/market for any product idea (incl. its various iterations) is at most 5%. This follows from the startup failure rate and the fact that most successful startups pivot at least once, so they have at least two distinct product ideas.

Let’s return to the question of how many unvalidated problem ideas we have to come up with (per year) for one success \(N=1\)?

\[\begin{eqnarray} n_{\mathrm{i}} &=& \frac{N}{f_{\mathrm{v}}\, n_{\mathrm{s}}\, f_{\mathrm{d}}\, f_{\mathrm{a}}\, f_{\mathrm{r}}\, f_{\mathrm{s}}} \\ &=& \frac{1}{0.65\cdot 25\cdot 0.90\cdot 0.20\cdot 0.10\cdot 0.30} \\ &\approx & 11. \end{eqnarray}\]

That is not too far off from 17. It is therefore a reasonable number and validates the D4P. That also means that businesses need to generate \(R_{\mathrm{i}} = n_{\mathrm{i}} / L \gtrsim 2\) fresh ideas per year. Product/market fit is not a one-off event!

For some industries (e.g. media and gaming) where the long-term retention is an order of magnitude lower, the PMF fraction is only \(0.0006\), which is close to what we find based on the rule of thumb that 2,000 product ideas are required to have one success, or \(1/2000 = 0.0005\). These product ideas are different from raw problem ideas (\(n_{\mathrm{i}}\)) in that they include every iteration and variation, which inevitably includes the suggested solutions. Product ideas such as these have gone through problem validation and solution design: \(n_{\mathrm{i}}f_{\mathrm{v}}n_{\mathrm{s}}\approx 1900\). That’s in the ballpark.

How can you increase your odds of success?

First, validate all feasible ideas: prioritize your ideas and make sure that everything that moves on to solution design and beyond has been validated through primary and secondary market research. If research is too expensive, do it anyway.

Why am I saying this when it is so obvious? Because to \(100\left(1-f_{\mathrm{v}}\right)\%=35\%\) it appears not obvious at all!

Second, focus on the variable with the most headroom that is essential for PMF: retention. Most fractions in the D4P can be increased only marginally, but the retention rate \(f_{\mathrm{r}}\) can still go up by a factor of 10. High retention rates also increase product lifespan, boosting success both directly and indirectly.

So, what can you do to increase the long-term retention rate from 10% to, say, 50%? Compare this to Spotify’s annual retention rate of nearly 80%. Personalization plays a crucial part in such stellar retention figures, but there are easy fixes, too.

Subscriptions: 3×

If your product requires a subscription, you might want to look at annual instead of monthly subscriptions to boost long-term retention by a factor of three.

Onboarding: 2×

You can get another factor of two by improving the onboarding experience. That’s because companies lose up to 75% of customers in their first day with any app. For SaaS that figure is 60% after one month and 65% after two months. The top-10% of products in the market retain customers almost twice as much as the rest.

Define the retention rate for the \(i\)th day as \(r_{i}\). The retention after one year is \(r = r_{1}r_{2}\ldots r_{365} = r_{1} \rho\), where \(\rho\) is the overall retention for the remainder of the year. If all you can fix is the day-1 retention \(r \mapsto r^{\prime} = r_{1}^{\prime}\rho\), you can boost retention after one year by 100% simply by doubling the day-1 retention: \(r_{1}^{\prime} = 2 r_{1}\). That’s because \(r^{\prime}/r = r_{1}^{\prime}\rho / r_{1}\rho = r_{1}^{\prime} / r_{1} = 2 r_{1} / r_{1} = 2\).

Visualize annual retention as 365 stacked leaky buckets. Plugging the first leak has a disproportionate effect by preventing downstream losses. If you can halve the leakage at the very top, you can halve the leakage at the bottom, because the bottom 364 buckets still leak at the same rate as before, but they receive twice the amount compared to before.

Remember that 9 out of 10 customers who have a bad experience do not intend to do business with you again. A single bad experience with your product or customer support is enough for a third of customers to leave. Permanently.