8 Smart Ways To Use Prescriptive Analytics - InformationWeek

InformationWeek is part of the Informa Tech Division of Informa PLC

This site is operated by a business or businesses owned by Informa PLC and all copyright resides with them.Informa PLC's registered office is 5 Howick Place, London SW1P 1WG. Registered in England and Wales. Number 8860726.

IoT
IoT
Cloud // Platform as a Service
News
6/28/2016
07:06 AM
Lisa Morgan
Lisa Morgan
Slideshows
Connect Directly
Twitter
RSS
E-Mail

8 Smart Ways To Use Prescriptive Analytics

Somewhere between blind faith and skepticism is the world of prescriptive analytics. Here, machine-generated action items and potential outcomes meet human decision-making. Finding the right balance between algorithms and common sense can be tricky, so consider these tips.
4 of 9

Avoid Blind Faith 

Not all prescriptive analytics platforms and software are created alike, although it can sometimes be hard to tell the difference. Some options are broad, some narrow, and some have been training on certain types of data for considerably longer than others. Despite what a vendor may claim, don't simply buy software or a platform subscription and hope it's going to provide the magical answers out of the box to meet your unique business needs.
'There's a risk that a prescriptive system might be blind to your real challenges,' said Doug Henschen of Constellation Research. 'Does the system have a sufficiently holistic take on the dynamics of the decision?'
'For example, a prescriptive system might draw certain conclusions about customer spending patterns, or a propensity to buy certain goods, if the data is limited to their purchasing habits with a single retailer. If the view is broadened to include social sentiment, demographic and lifestyle information, and perhaps third-party data on spending with other retailers, the recommendations might be very different,' Henschen said.
(Image: DasWortgewand via Pixabay)

Avoid Blind Faith

Not all prescriptive analytics platforms and software are created alike, although it can sometimes be hard to tell the difference. Some options are broad, some narrow, and some have been training on certain types of data for considerably longer than others. Despite what a vendor may claim, don't simply buy software or a platform subscription and hope it's going to provide the magical answers out of the box to meet your unique business needs.

(Image: DasWortgewand via Pixabay)

4 of 9
Comment  | 
Print  | 
Comments
Newest First  |  Oldest First  |  Threaded View
LisaMorgan
50%
50%
LisaMorgan,
User Rank: Moderator
7/8/2016 | 4:23:17 PM
Re: Prescriptive analytics: A thought experiment
I worry about the accuracy of prescriptive analytics too because I think the problem is being oversimplified in a lot of cases.
LisaMorgan
50%
50%
LisaMorgan,
User Rank: Moderator
7/8/2016 | 4:15:31 PM
Re: Prescriptive analytics: A thought experiment
Thanks for mentioning it.
GaryB790
50%
50%
GaryB790,
User Rank: Strategist
7/5/2016 | 11:05:30 AM
Re: Prescriptive analytics: A thought experiment
@StevenL530,

I think that you may have missed some information somewhere. The Prescriptive Analytics data is not generated in a vacuum and the data is not static it is a function of accumulated data over some time period.  If need be, there could be a [pick your time frame] window to ensure that the trending isn't under reported by having a massive dataset showing the opposite of what is actually happenning.

Within the Prescriptive Analytics algorithm there should be a "feedback loop"...for example, page hits, or some socitial trending information gathered from somewhere else noting the trending of deserts, sweetners (artificial vs natural), baked or fried or frozed....etc.

Like all prediction algorithms, it takes time [for data sample collection] for the algrithm to optimize and predictive type formulation....and the correct data being sampled.
SteveR329
50%
50%
SteveR329,
User Rank: Apprentice
6/29/2016 | 6:26:29 PM
A journey not an event
Great article.  I agree with most points.  Data normalization is critical (garbage in, garbage out) as is the human factor.  People are fickle which makes trying to predict the decisions they will make a very big challenge.  It is still a journey worth taking, but as you point out, expect it to be a journey not an event.
StevenL530
50%
50%
StevenL530,
User Rank: Strategist
6/29/2016 | 10:52:12 AM
Prescriptive analytics: A thought experiment
Your algorithm prescribes adding the Fruity Pebbles cheesecake recipe to your web content. The counter-intuitive recommendation is only counter-intuitive the first time. The recommendation now becomes intuitive, routine. So each year, in late spring, you run another recipe just like this, Cap'n Crunch Cupcakes, Fruit Loops Parfait, and so on. But something changes. One year, the recipe no longer garners clicks. Visits are down. Will the prescriptive algorithm double down on its advice, run more breakfast cereal no-bake recipes? Or will it detect the root of the new trend? If it is just telling me what to do and not why, how does the managing editor know when to change course and in which direction? With dozens of statistical models forecasting every possibility, how to choose? Weather prediction has been banging up against this challenge for over a century. In business, we are always on the hunt for the disruptive change. Not only will prescriptive analytics miss the disruptive change, it will reinforce management's reliance on these now misleading analytics far longer than they otherwise might.
ShannonK584
50%
50%
ShannonK584,
User Rank: Apprentice
6/28/2016 | 1:53:15 PM
A bit of a limited review of the potential of prescriptive
Many points brought up here are valid. However, I worry that the view of potential use cases / growth with prescriptive are quite limited to the idea of heuristics (rules-based) decision automation that people (including Gartner) are incorrectly calling prescriptive analytics. 

If you view prescriptive analytics as optimization (linear programming), the use cases and benefit are much wider reaching. Also, you don't run into the probably of having misinformation from prescriptive (e.g., infeasible plans or plans that are not optimized to your business objectives). 

I recommend reviewing this article and considering redefining your idea of "prescriptive analytics." The use cases and benefits of what you refer to here are quite different from the use cases and benefits of true optimization. Furthermore, there are far fewer problems, and the benefits make attacking change management issues a no-brainer.
Slideshows
7 Technologies You Need to Know for Artificial Intelligence
Jessica Davis, Senior Editor, Enterprise Apps,  7/1/2019
Commentary
A Practical Guide to DevOps: It's Not that Scary
Cathleen Gagne, Managing Editor, InformationWeek,  7/5/2019
Commentary
Diversity in IT: The Business and Moral Reasons
James M. Connolly, Editorial Director, InformationWeek and Network Computing,  6/20/2019
White Papers
Register for InformationWeek Newsletters
Video
Current Issue
Data Science and AI in the Fast Lane
This IT Trend Report will help you gain insight into how quickly and dramatically data science is influencing how enterprises are managed and where they will derive business success. Read the report today!
Slideshows
Flash Poll