Prescriptive analytics is the final stage of business analytics. Learn more and read tips on how to get started with prescriptive analytics.
Technology has given us the ability to forecast enterprise trends and predict success in ways the business leaders of yesterday couldn’t fathom. In the past, successful businesses had to rely on small sample sizes, simple questionnaires, and other ways of gathering of data to predict general trends, but not anymore.
The modern business world is inundated with data. Everywhere you turn, some website or app is asking for your data or gathering it quietly in the background, but why?
All that data has to go somewhere, and it should have a purpose. Sure, lots of it sits in data lakes or other forms of data storage, and plenty of it ends up being sold for profit.
In an ideal world, your data wouldn’t be used for quick gains, but would go to serve a better cause that many businesses already use it for: To make the best possible business decisions.
All of that data being amassed by businesses can be used to describe current trends, predict what’s going to happen next, and most importantly, prescribe the proper course of action a business should take to ensure success in the most efficient way possible through the process of prescriptive analytics.
Launching a prescriptive analytics initiative is no small undertaking, but the results can be transformative. If there’s uncertainty in your organization’s future, you can do your best to eliminate it with the right prescription. There’s a lot to know before you start, and this guide will help you understand what needs to be considered before jumping into the analytics deep end.
SEE: 60 ways to get the most value from your big data initiatives (free PDF) (TechRepublic)
What is business analytics?
To understand prescriptive analytics, it’s important to have a basic working knowledge of the larger world of business analytics.
Business analytics is a multi-stage process. Each step involves the analysis of data to reach a particular type of conclusion, the ultimate goal of which is to build the best possible strategy for optimized organizational action.
There are typically three parts described in business analytics:
- Descriptive analytics is the kind of analysis that is performed to describe an organization’s current circumstances. The data used in this instance can include customer feedback, sales numbers, website traffic—essentially any data that is a record of past events that can be used to analyze business up to the present.
- Predictive analytics uses the same type of data, and sometimes the descriptive outcomes, to predict what will happen given the current circumstances. Businesses often employ machine learning and various forms of predictive modeling to make predictions. Think of predictive analytics as what will happen if current organizational practices and habits remain the same.
- Prescriptive analytics is less fortune teller and more medical doctor. Instead of simply predicting what will happen, prescriptive analysis tweaks certain variables to achieve the best possible outcome, and then prescribes that course of action.
Businesses can employ one or all of these forms of analytics, but not necessarily out of order. In order to predict the future, you need to know what has already happened, and in order to change course, you have to know what’s likely to happen without that course correction.
It’s entirely possible to stop after getting an accurate picture of the present and what led up to it, but most organizations would be short-sighted if they stopped at that point. Sticking only to descriptive analysis leaves the future a mass of uncertainty that is likely to surprise—and not in a good way.
What is prescriptive analytics?
Prescriptive analytics is the third and final stage of business analytics; it builds on predictions about the future and descriptions of the present to determine the best possible course of action.
At the core of prescriptive analytics is the idea of optimization, which means every little factor has to be taken into account when building a prescriptive model. Supply chain, labor costs, scheduling of workers, energy costs, potential machine failure—everything that could possibly be a factor is included in making a prescriptive model.
The term prescriptive analytics was coined by IBM and described in detail in a 2010 piece an IBM team wrote for Analytics Magazine. The article breaks down the three types of business analytics into greater detail, including how IBM conceives of prescriptive analytics as consisting of two elements:
- Optimization, or how to achieve the best outcome, and
- Stochastic optimization, or how to achieve the best outcome and make better decisions by accounting for uncertainty in existing data.
The authors of the Analytics Magazine article also point out an essential (and obvious, once you think about it) fact about prescriptive analysis: It isn’t a new concept. What is new, they say, is the computing power that makes comprehensive prescriptions possible.
“With improvements in the speed and memory size of computers, as well as the significant progress in the performance of the underlying mathematical algorithms, similar computations can be performed in minutes. While this kind of information might have been used in the past to shape policy and offer guidance on action in a class of situations, assessments can now be completed in real time to support decisions to modify actions, assign resources, and so on.”
What also sets modern prescriptive analytics apart is the speed at which we can update prescriptions. Now a hitch in the system, a change in vendors, an error in accounting, or the loss of an employee can be responded to in near real time and with a depth of knowledge not possible in the past.
What tech goes into prescriptive analytics?
Prescriptive analytics relies on big data collection. All of the data an organization gathers, structured or unstructured, can be used to make prescriptive analyses.
Gartner’s definition of prescriptive analytics mentions a number of different tools that could go into making prescriptive analytics happen, including:
- Graph analysis;
- complex event processing, which involves combining data from multiple sources to infer patterns and model complex circumstances;
- neural networks, or combinations of various machine learning algorithms designed to process complex data;
- recommendation engines, which are computer algorithms designed to predict positive or negative preference based on what users have chosen in the past;
- heuristics, or alternative methods of problem solving that can approximate an answer when finding a definite one fails; and
- machine learning.
Machine learning and artificial intelligence are the driving forces behind the growth of prescriptive analytics. One of the largest prescriptive analytics firms, Ayata, has built its entire prescriptive system around AI and machine learning, which it says is built on “AI controlling and combining the science of predictions with the science of decision making.”
Ayata describes its prescriptive software as using operations research, which involves making better operational decisions using various analytic methods, and metaheuristics, which are heuristic models designed to choose the best heuristics to use to simplify and speed up the rate of solving a particular kind of problem.
All of the technology that goes into prescriptive analytics is designed to make models more accurate by using a wider range of data types, relate different forms of analysis to each other to create a web of knowledge, and decrease the amount of time needed to deliver results by making heuristic decisions based on all the data and analysis that has been performed.
What are prescriptive analytics use cases and real-world examples?
The use cases for prescriptive analytics are vast. Any business with an eye on optimizing its performance, and the budget to spend on prescriptive analytics software and the man power needed to operate it, can benefit from some form of prescriptive analysis.
From a marketing and sales perspective, prescriptive analytics can be used to:
- Optimize the assortment of products in a retail store;
- optimally price items and services;
- find the best mix of marketing methods (online, print, radio, etc.); and
- negotiate a better contract with customers and vendors.
- Improve driver retention to reduce training costs;
- eliminate unnecessary driving, flight, and sea transportation miles;
- increase driver productivity by improving routes and eliminating wait times to load/unload;
- increase speeds and reduce costs by optimizing distribution networks; and
- eliminate nearly all warehouse packing errors (companies in the case study were 99.5% error free).
The oil and gas industry makes extensive use of prescriptive analysis to:
- Improve drilling completion rate by training machine learning models to recognize the most beneficial places to set up field operations;
- determine the best possible locations in a particular field to drill first;
- optimize equipment configuration to eliminate downtime due to breakage and maintenance;
- improve operational safety and eliminate potential environmental disasters; and
- establish the best possible pricing by predicting the rise and fall of fuel markets.
- Decrease transaction processing time;
- lower transaction costs;
- increase the total amount of possible transactions processed in a particular time period;
- create better portfolios for financial investment;
- optimize financial decisions like when to invest, how much to invest, etc.; and
- reduce investment risk (in the IBM case study, prescriptive analysis reduced risk by 30% while maintaining similar yields).
If your business collects data and could feasibly use that data to model the present, predict the future, and find the best of all possible outcomes, then prescriptive analytics probably has a use case in your industry, too.
How can my business use prescriptive analytics?
Getting started in prescriptive analytics can be challenging, especially if your organization hasn’t done much with business analytics up to the present.
If you have a lot of data that could be used to build prescriptive models, you have a good starting point; without data, you’ll have to start from scratch and begin gathering and compiling the data you need to make a good analysis.
SEE: Big data: More must-read coverage (TechRepublic on Flipboard)
Logistics analytics firm River Logic has an excellent guide on how to get started with prescriptive analytics, which it breaks down into three parts:
- Figuring out what you want to get out of prescriptive analysis;
- outlining the steps it will take to get there; and
- determining what kind of employee skills you’ll need to get the job done.
What is your goal for prescriptive analytics?
Determining what you want to do with prescriptive analysis is essential for formulating a successful plan.
- Does your organization need to reassess its entire approach to a particular issue?
- Are you losing ground to competition?
- Is there a particular goal you want to meet in the future?
- Or do you want a rolling analysis of your current state, the possible futures, and how to optimize for the best outcome whenever reassessment is needed?
These complicated questions inform the next two steps that River Logic recommends.
What will it take to accomplish your goal?
Prescriptive analysis isn’t something you can just plug into your organization and expect it to spit out results—you’re going to need a lot of framework in place to be effective. River Logic breaks this step down into six sub-steps.
- Build a team: You’ll need the same roles as you would have in any transformation initiative.
- Conduct a value discovery: Workshop and have meetings to define the problem, think of a solution, and get an idea of how much value you can get out of it.
- Build a proof of concept: Tackle a subset of your overall goal, like developing a business implementation model, and work it out to see if you’re moving in the right direction.
- Implement: Design the entire model for your prescriptive analysis and its surrounding business needs, and flesh out the rest of your proof of concept model. Validate the entire model through small scenarios and tests.
- Configure: With the structure in place, your team will know what data is needed and be able to address all the technical elements of a predictive model.
- Roll out: Start full prescriptive operations, and put your initial findings into practice.
What skills does a team need to run a successful prescriptive analytics program?
There is a lot of mathematics, programming, analysis, and data science that goes into a successful prescriptive analytics program. If you don’t already have qualified people on board, you’ll want to consider finding the following types of professionals.
- People who can configure your analytics model: This could include data scientists, business analysts with good spreadsheet and database skills, or anyone with an engineering, math, actuarial, or accounting background.
- People who can define the problem you want to solve: Subject matter experts like financial planning and analysis professionals are ideal for this role.
- People who can embed analysis into business planning: Anyone with skills in data integration, data management, user interface configuration, and business intelligence professionals fit here.
What prescriptive analytics tools are available?
Part of this total process of getting started with prescriptive analytics will be figuring out what sort of software you want to use to conduct your prescriptive analyses. IBM, NGDATA, River Logic, FICO, and SAS are just some of the organizations that offer optimization modeling and optimization solving software.