Predictive Analytics

Predictive Analytics Is No Crystal Ball

Predictive Analytics has long been touted as the “crystal ball” of Fortune 1000 companies. Companies typically turn to the promise of predictive analytics to try to better manage risk, improve return on investment; and increase yield. They are essentially trying to figure out what they don’t know and predict the future by searching for outliers and trying to find patterns and relationships.

Yet, predictive analytics is still largely a hammer looking for nails deeply shrouded in hype. Here’s why:

1.    Selective successes
While there have been some highly publicized successes, they can be broadly categorized under two scenarios (a) personalized recommendations in a universe of information and (i.e. Amazon product recommendations) (2) Predicting failures – be it churn, equipment, etc. But history cannot always predict future. Most operational business scenarios involve people and lots of unknowns. And this is next to impossible to factor into to predictive models. For example, a sales model will include accounts and analytics around their buying propensity, but what the model cannot include is the knowledge of the sales reps, changing market volatility, and the repercussions of a competitor launching a new a product and the subsequent changes in market share.

2.    PhDs have to do the job
To generate extrapolations you have to be a data scientist. These expensive and hard to find people are experts in data, not experts in your business. They see the numbers in black and white and speak in a different language—r square, gradient boosting, random forest and coefficients. They’re certainly not talking in terms of sales velocity. The nuanced tribal knowledge of business users gets lost in translation to the data scientists, if it is not tightly managed.

3.    The process isn’t yet seamless with business users
You may want to put critical insights into the hands of the business user. However, operationalizing them can be quite difficult. The result is that you’re left with a future prediction horizon while the market is being disrupted in the present. The result is a disjointed system where the people making the decisions (the field) are far separated from those building the models (data scientists).

4.    Don’t take data for granted
The market may wax poetic about the potential of big data, but the reality is that it simply isn’t at anyone’s fingertips. Business users have to rely on IT expertise to get their hands on quality information. The data itself needs to be mapped, de-duped, cleansed etc. before it becomes complete and accurate for data scientists to even engage. And the business user? They are left waiting for insights to be delivered while they must execute now.

So what’s the solution? At Anaplan, we believe it is to empower business users to complete all mainstream modeling while relying on data scientists for specific problems. The idea is that the business user should be able build their own information hub and models using simple business syntax and add their own nuanced expertise. Within that scenario, the business users tackle the toughest modeling problems — the ones that have a direct impact on their world.

Anaplan isn’t a replacement for predictive analytics; it’s a partner. Used in parallel, Anaplan arms you with the ability to:

  • Gather information from the field, execute an analysis of a data set, and provide recommendations
  • Collaborate on models with your data scientists
  • Seamlessly couple knowledge from the field with results from predictive analytics

Sounds pretty great, right? Anaplan gives companies the ability to put the business user, analyst, and the data scientist in the same car, have the same conversation, and drive towards the same goal. A business user can analyze their model in real time to compare versions or see what happens when certain drivers are changed, thus testing new hypotheses. This enables course correction and optimization like never before. For instance, a beverage company can analyze their current promotion and determine when to double down their spend, or stop the campaign and reallocate the remaining dollars.

Related Posts

Eliminating the strategy-execution gap Most businesses fail to re-plan and re-align quickly enough to capitalize on (and let alone deal with) change. The constant torrent of changes that take them of...
Anaplan implementation tips: What to consider when... You’ve spent hours, days, and weeks (maybe months) pulling data from siloed point solutions into massive spreadsheets, hoping to consolidate and analyze the dat...
Unlock Supply Chain: Louis Vuitton’s hackathon wit... The mission: Develop a world-class supply chain application for the globally renowned luxury brand Louis Vuitton. This challenge was accepted by 50 students who...

Share this Story

Leave a Reply