Watching a demo of predictive analytics can feel like winning that first game of pool of a newcomer. I just saw an easy way to make a little money for my company and perhaps make a bit of a name for myself. Then the vendor with the shifty eyes says, “Hey kid, you’re pretty good at this game. What say we double the stakes?” Are the Werewolves of London swelling from the background?
It’s in the Way That You Use Analytics for IT
The key to surviving this encounter is to realize what is really being offered. Predictive analytics exist, and more than likely, they can help you do what you do, perhaps even provide you with a sustainable competitive advantage. It is also true that a person can use geometry to strike a ball such that it hits three rails and drops in the opposite corner pocket. It’s in the way that you use it.
It’s not in the model
When somebody wants to promote a truly crazy idea to an individual, they often start by describing quantum physics. Did you know that there can be two electrons divided by the breadth of the universe that will flip upside down if the other flips right side up? [True] Did you also know that the Reptile elite worked with Joe Montana to fix the ’88 election? [Less True]
When a vendor wants to push an expensive solution to an enterprise, they often start by describing predictive analytics. Did you know that the same math that describes the development of the human brain can be used to improve mailing campaigns? [True] This model will make all of your business processes issues a thing of the past! [Less True]
Wouldn’t it be amazing if we could turn loose Genetic Algorithms, Neural Networks, Multiple Logarithmic Regressions, Decision Trees, or for that matter, Random Decision Forests to make IT Operations peer into the future? Would you feel any different if I told you the youngest of the models listed above was born in 1975? The oldest was born 71 years before a canon in a symphony hall was the latest novelty.
The mouth agape feeling that we have upon witnessing the wonders of the future makes us all vulnerable. These capabilities are so close. If we could just harness them, and then appears the guy that could get us there. As we mature as managers, we have to know that when we hear exactly what we want to hear, it may not be the truth. Say, kid, you’re pretty good at this game.
It’s not in the technology
Well, it is technology. The models listed above have been impractical at scale until the relatively recent explosion in computing power. However, it doesn’t matter so much which technology you choose. Nearly every software manufacturer has built or bought a capability to incorporate machine learning. Capability is not execution.
If my company purchases Excel, it does not mean that they have acquired a finance department capable of valuing projects based upon the net present value of their projected future cash flows discounted by risk. It is, however, required for well-trained and experienced professionals to execute their complex work. Or Google Sheets, or any equivalent spreadsheet software. A capable software is required but not sufficient to achieve the task.
It’s in the Use Case
There is a growing set (but not enough) of well-established use cases for AI/ML. Optimizing mailing/email campaigns, Basket Analysis for retail optimization, Customer Clustering for marketing strategy, fraud detection, Recommendation engines, etc. It’s appropriate to have the mouth agape feeling again. These achievements are mind-blowing.
Bear in mind, however, that these achievements were each hard-won, often after decades of experience. What if your situation does not fit one of these established use cases? Would you say that you are in year one or year ten of that process? Are you interested in being the pioneer? Years of advantage may accrue to your organization, but there is cost and risk involved.
It’s in the Data Gathering
We discussed in our last article the process preparations that are necessary before embarking on the Predictive Analytics path. However, we should state that it is necessary to implement sufficient monitoring and process tooling in order to generate the data upon which AI/ML models can be fed. There is no AI/ML without data. There are current trends in the area that are reducing sample size, but that is not to be confused with obviating the need for sufficient data feeds.
It’s in the Data Wrangling
One of the fantasies that AI/ML indulges is the ability to perform next-generation analytics without people that have next-generation skills. Unfortunately, every one of the established use cases was born out of the deep experience with people familiar with the business processes, application data, and often the psychology of the constituent parties.
The classic basket case analysis involves an AI/ML obtained observation that two of the most common item combinations within a shopping basket are diapers and beer. Our imaginations can jump to a very lazy weekend and chuckle, but consider the decades of data structure that enabled tracking individual items within individual baskets for individual customers. No model or technology gets you there without industry knowledge and painstakingly minute knowledge of the data available in the organization and industry.
It’s in the Way That You Use It
The key is to find out from your vendor if they have the use case that fits you. Do you have the tooling in place to support the use case? Has the vendor credibly demonstrated the business process and data inventory mastery in your industry or function to firmly establish the use case under review? If you are in a situation where you are going to be the pioneer, do you have the people you need to succeed, and can you trust the vendor to be a reliable partner?
Northcraft has spent over a decade working with clients to identify valuable use cases and using AI/ML models to help them extract value from IT services and Operations. If you’re going down that road with another partner, remember that it’s ok to ask for examples and references. The rewards are vast, but don’t let yourself get hustled.
If you got an area of excellence . . . you’re the best at something, anything . . . then rich can be arranged. Rich can come fairly easy.
– Paul Newman as Fast Eddie Felson