Too many cooks spoil the decision?
This is the first installation in a series of scientific research blogs from Fingertip’s Data Scientist Lasse Winter. In these blog posts we dive into data acquired from our own use of the Fingertip application to discover curiosities around decision making.
Decision making is an inherently social process which in an organizational setting usually involves a group of people. And where there is a group, there are everything that makes us human; tensions, sparks, emotions and feelings. The participants debate on the matter, then perhaps vote on the way forward and make a decision. Some people are excited about the decision, others less so. If the decision turns out to be a poor one, the overall mood sours. Or if the decision turns out the be a success, even initial naysayers can get a big smile on their faces.
The beauty of Fingertip is that it digitizes the entire decision making process, from drafting the idea all the way to the execution phase. Fingertip also asks the user to evaluate a multitude of aspects about the decision, starting from his/her mood. These aspects create a steady supply of “modern oil”: data. This gives us a unique vantage point for exploring decision making process in a totally new and exciting way using data analytics. The possible questions waiting for an answer are limitless.
For example: does the mood of people involved in a decision vary as a function of the number of people involved?
The more the merrier?
The Condorcet’s jury theorem famously states that if the individual probability to be correct of each participant in a decision is over 0.5, then adding more people to the jury would increase the chances of ending up in a correct ruling. Wisdom of the crowd is also known to produce superior estimates to experts in a number of tasks, such as guessing the correct number of balls in a jar.
The more the merrier then right? (Assuming people involved are slightly better than average chumps). Likely not! The wisdom of the crowd tends to function best when there is an objective answer to be had. However, decision making in usual workplace environment tends not concern estimates of a number of balls or a verdict in a trial. The decisions often concern blurry issues with no correct answers and what’s worst your amazing decision still ends up leading to lackluster results.
Speaking against increasing the number of people involved in a decision towards infinity is the increasing cost to decision making process. Just witness the world’s greatest show of democracy when India votes in parliamentary elections for several weeks on end, in contrast to the quick and efficient process in tiny Finland. One-man teams need not concern themselves with opposing opinions and can move quickly and decisively where larger teams need to spend significant time debating and gathering opinions.
But the smaller the team, by rule the smaller the pool of opinions and knowledge base that is required for making an informed decision. And if your decision affects other people, it might be difficult to gain their approval the thinner the pool of decision-makers (it might be tempting to forgo the process of deciding what to watch with your spouse on Netflix but on average I would advise against it!)
Quite obviously the sweet spot is somewhere between the extremes.
Finding the sweet spot
Below I have plotted the number of people in a decision against the average mood given by users. The data is limited to decisions with a minimum of three people evaluating their mood so a single grumpy person evaluating a decision does not speak for the entire group of people involved in a decision and thus increase the reliability of the mood metric.
From the scatter plot we can see that there is a weak but not insignificant relationship between the number of people in a decision and the average mood, which is quadratic rather than linear in nature. Within the plot is embedded a quadratic regression curve from a model fit on the data. The regression model turns out statistically significant, giving weight to our hypothesis, although the overall variance explained by the model remains minuscule.
However, not every person involved in a decision is created equal. Some people are deeply involved in the process and might even have the honor to carry out the decision. Some people are required for their expert opinion but will shy away from the repercussions while others just need to be kept in the loop. Fingertip employs RACI-model to make a theoretical distinction between the different roles involved in the decision making process: Accountables, Responsibles, Consulted and Informed. There is no space here to go deep into the details, but go ahead and read more in our blog.
Realizing the differing roles of people, what follows is a natural follow-up question. Is the function for optimal number of people different for differing roles? Theoretically it looks like there should be no upper bound to the amount of Informeds in a decision, besides the obvious limitations to transparency by security issues. Same applies to a lesser extent on the number of Consulted’s as more varied perspectives usually lead to better decisions. However, most obviously the number of Responsibles should see diminishing returns after certain point, for reasons spelled out above.
The relationship for the count of Responsibles vs average mood follows the expected pattern. You are better of in the middle rather than having very few or a whole bunch of people in Responsible role. However, for the count of Consulted the relationship is wholly different. Here adding more people to Consulted role initially decreases the average mood, until the mood again picks up. Although, it needs be said that the quadratic relationship is driven by only couple observations where count of Consulted is above 15. Both these relationships turn out statistically significant.
A small foray into data provided by Fingertip already provided us with some interesting insights. However, it is a data scientist’s ethical responsibility to issue a word of warning. The data from a single organization which is Fingertip itself. While there are reasonably grounds to believe the hypothesis of quadratic relationship between the number of people and the average mood stands also in huge multinationals as well as medium-sized family enterprises, the results cannot be justifiably generalized, much less invoked as a law of nature. However, thanks to Fingertip, any organization wishing to find their “sweet spot” has the possibility to do so!
Lasse Winter is the leading Data Scientist at Fingertip with a background in social sciences and a specific interest in text analytics. He is passionate about gaining exciting insights from data. During his free time Lasse loves sports and reading, with a specific passion for football.
Fingertip is an online decision-focused business management solution designed to substantially improve efficiency, effectiveness, and empowerment in large complex organizations in the digital age.