It’s necessary to think about what analytics help the business do, in their most abstract form. All that analytics can do is help you make a decision better. That’s it! So the value of analytics relates directly to how well you are making the decision before the analysis, and whether the information that you would need to make a perfect decision is available the first place. For example, if there is just one supplier for a part that you’re interested in, no amount of analytics will help you optimize a decision. In a similar vein, if you’re receiving almost no information about the visitors who are coming to your website, no amount of machine learning will help you design a better pitch for them. You need to have both more information and more options before you, and then the computer brain will improve your business.
All that said, there are certainly plenty of situations in which human decision-makers are being drowned out by either the quantity/speed of decisions that they need to make, the amount of information that they need to consider when making them, or the number of options that they need to weigh. Let’s take a look at each of those in turn:
- There are some situations where the speed of decision making is clearly too much for a person. The most commonly cited example is credit card fraud detection, where thousands of decisions need to be made every second. But there are also more pedestrian examples in many businesses. It’s worth remembering the famous study of judges granting parole in Israel. In this study, researchers found that judges were more likely to deny parole to prisoners that they reviewed later in the day. Granting parole is a risky move, because the judge could face ridicule if the prisoner immediately commits a new crime upon release. There is no such risk in denying parole, making it the default option when a judge cannot make up their mind. The researchers concluded that as the day wore on, the judges’ reserve of mental energy was depleted, and they fell back to the simpler, less risky option much more frequently.
- There are also situations where the amount of information that needs to come together to make one good decision is overwhelming for a single person. People naturally gravitate towards using rules of thumb, which generally rely on just one or two pieces of information. This works really well for some sorts of decisions, where just one or two pieces of information gets us most of the way to making an accurate decision, but there are plenty of decisions where each of a large number of factors makes a small difference, and there’s no one or two that can be used to get most of the way there. Managing this sort of complexity is where computers excel, to the point that their results can be overwhelming and confusing to decision-makers. There are techniques that can be used to reduce the number of factors that need to be considered,. But when the decision truly is complex, don’t allow your human tendency to prefer simplicity to overshadow the need for accuracy.
Finally, if there are many options to choose from, it is simply unwieldy for a human to sort through the massive list of possibilities. An obvious example of this is Facebook’s friends suggestions, where there are at least 10,00 2nd-degree connections between the logged-in user and other users on Facebook. Though a human could look through all of those profiles and probably make an even better suggestion than the algorithm can, it clearly doesn’t make sense for humans to choose between all the options. Honestly, a lot of information that really matters doesn’t exist on Facebook, so having a system that can take a decently good shot and try try try again will outperform a meticulously researching matchmaker.
These are all situations in which it does make sense to implement an automated decision-making system. Remember, though, these systems are necessarily expensive to build and require ongoing maintenance. If a person can do it, have them do it! You can always implement a system later, and then he’ll be doing it with much more on- the-ground knowledge
Photo credit: https://www.flickr.com/photos/gagilas/