Innovation is like climbing a mountain in fog. You are moving upward with no evidence you are nearing the summit, and the fog is not empty. It is full of risks, customer complaints, competitor one-pagers, three-letter acronyms, conflicting advice, self- doubt, and news stories about someone you went to school with.
Because progress is so hard to measure in that fog, we tend to fill the gap with gut instinct. That makes for decisions nobody can defend later, and for a great deal of avoidable anxiety. The alternative is to measure how much you are learning, do it repeatedly, and keep the rate high.
Incremental progress handles uncertainty
The current operating environment is often summarized as VUCA: volatility, uncertainty, complexity and ambiguity. The volume and velocity of available data keeps climbing while confidence in decisions does not follow; Salesforce has found that understanding customers is a struggle for around two-thirds of businesses. Meanwhile the expectation to innovate faster has not moved.
When uncertainty makes long-range planning unreliable, the answer is to plan and measure in smaller increments. That is the logic behind agile innovation teams: small groups kept close to the customer and able to change direction quickly, which research associates with higher productivity, faster time to market and lower risk than conventional approaches.
The core idea underpins iterative optimization in mathematics. Coordinate descent solves for one variable after another in small steps. Gradient descent moves in whichever direction the slope is currently steepest. In machine learning, the size of each step is simply called the learning rate.
The lesson transfers cleanly: an iterative approach is useless without iterative measurement. You have to stop and check where you are.
Not all measurements are equal
If we are measuring progress in small chunks, the central question is what counts as progress. There is no shortage of candidates. Most organizations already collect far more data than they convert into decisions.
So choose the measure carefully, rather than defaulting to whatever your dashboard already displays. Be especially wary of vanity metrics. Dollars raised is the plainest example: the fastest way to build a billion-dollar business is to have someone write you a billion-dollar cheque. Even DAU/MAU and MRR/ARR can mislead depending on the business model, since none of them show you whether the customers you are adding are profitable.
Learning as the goal
If the core problem is uncertainty, then reducing uncertainty is the goal. That leaves two options: change the environment so it is less uncertain, or learn. Unless your business model is organized crime, it is bound to be learning.
Learning is simply feedback stripped down to its insights, and feedback is what corrects a false hypothesis. The idea recurs everywhere under different names. Control theory calls it feedback control, machine learning calls it learning, and every customer-facing team calls it customer feedback.
There is a strategic argument too. The experience curve showed decades ago that companies learning through scale and repetition drive their costs down quickly. What has changed is the pace: as data and change accelerate, the advantage shifts toward whoever can learn fastest. Framed that way, an innovating team is doing the same job whether it is hunting for product-market fit, building a repeatable sales motion, or entering a new market. All of it is cutting through fog.
Two ways to measure it
Method 1: weekly reflection
Ask, every Friday, what you learned this week that was useful. Share it with your team and your manager in any meeting you already have. The filter matters: count only what was meaningful to a customer or to your unit's objective, and let others push back. Done honestly, it becomes obvious very quickly whether you moved up the mountain or spent the week in a muddy hole.
Teams that get clear value from this tend to formalize it: a weekly meeting to surface learnings from fresh notes and data, and a monthly one to pull two or three themes across a department or the whole company. Learning is not the research team's job; it applies to everyone.
Method 2: count experiments
Treat your activities as what they already are: experiments that test your hypotheses. The step is to make them explicit. Rather than emailing a few investors, email thirty and measure the response rate. Rather than announcing a feature, announce it and measure the click rate.
The rate of learning is then just how many experiments you ran this week, a number you can track without any new tools.
Where to start
Measure learning often, and commit yourself to keeping the rate up. Once measuring it is routine, you can start working on increasing it. Mastery of the learning rate is going to separate the companies that compound from the ones that merely stay busy.
