Measure for Measure

I read recently that the CEO of a very well-known UK foods brand said she felt that her business was suffering from “paralysis by analysis”.  What they needed, she suggested, was the return of a more entrepreneurial spirit.  It wasn’t clear exactly what was meant by “entrepreneurial spirit” but, reading between the lines, it seemed that the suggestion was promoting a bit more gut-feel and risk-taking and, perhaps, less emphasis being placed on seemingly endless number-crunching.

Innovation is, of course, one business activity that clearly carries associate risk.  Innovation is about the future and doing things that haven’t been done before.  This is the stuff of risk and uncertainty.  So, what role should “number-crunching” play in the innovation process?  Should innovation teams be encouraged to rely only on the entrepreneurial instincts, or can some numerical analysis actually help them on their way?

In this article we will present the case that sensible use of analysis and numerical modelling can, in fact, significantly speed up the innovation process as well as reduce some of the risk inherent in this kind of activity.

What colour are numbers?

The great physicist and engineer William Thomson, 1st Baron Kelvin, made the point wonderfully well as follows; “When you can measure what you are speaking about, and express it in numbers, you know something about it.  When you cannot express it in numbers your knowledge is of a meagre and unsatisfactory kind.  It may be the beginning of knowledge, but you have scarcely in your thoughts advanced to the state of science.”

To us, good use of numbers is like having a few more colours in your paint pallet. Some kind of “numbers model” as we call it adds significantly to the definition of the idea.  Our view is that an idea isn’t properly defined until there is some kind of numbers model, related to the purpose of the idea, included alongside the written definition.  Any new idea should be about doing something better or, least, “less bad”.  Perhaps our idea is to speed up a machine, or reduce cost, or maybe it’s for a product that we expect to contribute to our profits.  In every case, it is reasonable to ask the question “how much”.

In the case of the machine, “significantly faster” clearly still leaves a high level of vagueness, whereas “50% faster” adds so much more to our understanding of the idea.  This quantification would let a customer know how excited they should get. It also helps the technical people assess the proposed solution more objectively and get to grips with the logic of the idea more easily.

Perhaps most importantly, however, quantifying an idea this way enables much better identification of the assumptions that lie behind the idea.  These assumptions usually contain uncertainties and these uncertainties, in turn, can become the focus of a disciplined exploration using rapid application of a “plan, do, study, act” cycle as described in earlier articles.  In this way, using a numbers model of some kind helps reduce some of the risk involved in the new venture.

Numerical psychology – how to measure anything

Notwithstanding what we think are the incontrovertible benefits of building in a numbers model at the point of idea definition, we find that many idea inventors and innovation teams are reluctant to commit to a numbers model from the start.  We know that at least some of this reluctance relates to what they see as the purpose of the numbers model.  It seems clear that even when ideas are at their infancy – when we know least about them – many people actually fear committing numbers to paper in the belief that they will be somehow be held accountable for these numbers at some future date.  They anticipate a conversation that sounds a bit like “…but you said we would sell 1,000 of these in the first year…”, and the conversation taking place when only 10 of the damned things had been bought!

It’s important to emphasise that the numbers model at this stage is an estimate and not a commitment.  Its purpose is to add colour and clarity to the idea, and, as we have said, allow critical assumptions to be exposed so that they might be put to the test.  Nothing is surer than that aspects of the numbers model will change as the innovation team learns more as it conducts the activities required for exploration.

In his book, “How to Measure Anything”, Douglas Hubbard notes that there is a limiting belief in many organisations that some things just cannot be measured.  As a result, he says, “…resources are misallocated, good ideas are rejected, and bad ideas are accepted.” Hubbard notes that many people are prone to hide behind the term “intangible” to avoid the effort of having to think about how something can indeed be quantified.  Hubbard is adamant that anything can be measured.  As he says “…if it can be observed at all then it can be measured.  No matter how fuzzy the measurement is, it’s still a measurement if it told you more than you knew before.”

We wholeheartedly agree with Hubbard’s assertions and we strongly encourage inventors and innovation teams ensure that they make some useful effort to quantify the purpose of their ideas, even if their first instincts are that the item in question is entirely “intangible”.

Fermi is your friend

We should also point out, however, that we are not advocating trying to achieve a level of precision in quantification that cannot be justified either by the investment required to make the measurement or by the degree of accuracy that can be achieved through the measurement process itself.

In fact, much of the quantification we encourage at the early stages of an idea’s life-cycle can be drawn from the work of Enrico Fermi, a Nobel Prize-winning physicist who was famous for his intuitive, some would even say “casual-sounding” measurements.

Fermi was well-known for teaching his students skills in approximation of potentially baffling and strange quantities.  He would ask, for example, how many drops of water are contained in the Atlantic Ocean, or how many piano tuners would live in Chicago (Fermi taught at the University of Chicago after the Second World War).  His foundation principles included the beliefs that we all know more than we might think we know about a given quantity, and that we can use logic to break down a seemingly unknowable quantity into contributory elements that we can know something about and hence make usable estimates.  As statistician George Box is claimed to have said “…all estimates are wrong, but some are useful.” And this is key.  We don’t need our estimates to be correct at this stage – we just need them to be useful!

For Fermi’s piano tuners example, one way of reaching a useful answer might be to break the quantity down into elements like:

  • The population of Chicago
  • The average number of people living in a household and therefore the number of households
  • The proportion of households likely to have a piano, therefore the number of pianos in Chicago that piano tuners could work on
  • The average frequency with which a piano gets tuned
  • How long it takes on average to tune a piano, therefore the number of hours work available in Chicago for piano tuners
  • The average working hours available per piano tuner and therefore the number of piano tuners required to fulfil the likely demand

Whilst this model clearly needs the estimator to make big assumptions and rely on maybe only sketchy estimates, the numbers that emerge are way more useful than the “how could I possibly know that?” exclamation that would most often be made by someone when first asked the question.  This is the foundation of using Fermi estimates to bring some much-needed quantification to the innovation process.  This process holds whether we are estimating speed increase, costs saved, products that will be sold, new employees we will need to recruit, lives that will be saved, or pain that will be reduced.  And all the time, these estimates give us better clarification of the idea and the assumptions we need to investigate.

Fuzzy front end – introducing confidence intervals

People talk of the fuzzy front-end of innovation with good reason.  The term relates to the high levels of risk, uncertainty, and unknowns that percolate through the whole innovation project, and particularly at the outset.  A concept that helps us deal with all this uncertainty then is the idea of confidence intervals applied to our estimates.  Using a confidence interval (usually a 90% confidence interval for most purposes) helps us persuade reluctant estimators to stick their head above the parapet.  The confidence interval allows the group to give a range for their estimate rather than confine themselves to a single number.

So, in the example of the new product, we would look for the people involved to state something along the lines that they would be “90% confident that sales would be between 300 and 1500 units in the first year”, this estimate usually based on applying some Fermi analysis to break the quantity down into more knowable elements.  In this case, the knowable elements might be things like how many potential customers there are for the product, what the assumed business model will be, how much of this type of product customers typically buy in a given period, etc.

This type of estimate with confidence intervals helps us a lot in that it does give an idea of how confident (or not) a group is in its work.  If what appears to be a very wide range is suggested, then this might indicate that the group are really not confident at all.  As mentioned earlier, the other things this kind of estimate with confidence intervals gives us is access to the thinking and assumptions behind the group’s work – and these can then form the basis for much of the initial exploration that needs to be done to progress the idea.

You do the maths

There is no doubt that, in our experience, having even a rudimentary numbers model in place from the outset greatly increases the speed and confidence of the team tasked with progressing a new idea.  In fact, we’ve noticed almost a tangible burden being lifted when the team understands the true purpose of the numbers model – not a commitment, but rather an essential part of the idea definition and the foundation for much of the early exploration work that must be completed.  Communication of the idea is greatly enhanced by some useful quantification. It is much easier to picture a new idea when you have some notion of the dimensions.  As confidence in the method grows, we find teams become very adept at using Fermi techniques to help them think through an idea in the early stages.  So…next time your team produces an idea for consideration, make sure they’ve used a Calculator!