Disclaimer: This is not a political post. However, the recent twists and turns in our national politics have inspired this blog. Specifically, we saw an outcome that was assumed to be highly unlikely by almost all predictive data models, even ones that were crunched by the most sophisticated data scientists in the world. Countless businesses around the world made important assumptions based on these predictive models, and within 24 hours, those assumptions were all rendered useless at best. With this recent example, I’m reminded of my healthy skepticism of claims that predictive technology is a ‘silver bullet’ to solve a wide range of enterprise problems. Put succinctly—while predictive obviously has value (particularly in demand forecasting and pricing strategy in retail, for example), does predictive really deliver the highest and most immediate value for decision-making in the enterprise in other contexts? How reliable are predictive models based on currently available data, and are there more tangible (and obvious) ways to improve an organization’s revenue and profitability?
At Pramata, we provide solutions that allow companies to digitize their most valuable enterprise customer relationships, by extracting core data from existing contractual relationships and selectively drawing in CRM and billing data to deliver intelligence that answers questions such as …
- What has this customer bought?
- What price points are active for this customer?
- Are there any non-standard operational commitments?
- When can I increase prices and by how much?
… and hundreds of other important customer details that drive actions and decisions in sales, finance and operations.
When you look at the types of questions our solutions address, the first thing you will notice is the vast majority are not open-ended questions, but rather concrete and tangible intelligence about current customer commitments, or decisions that need to be made. It’s fascinating that in most companies, the status quo is to collect and disseminate this critical customer information using a combination of tribal knowledge, ad hoc CRM data and a lot of spreadsheets with data from disparate systems. It may provide you with a partial picture, but data fragmentation, incomplete info and inaccuracy leave out many details that can leave a lot of value and money on the table.
Compare the concrete intelligence available from Pramata's approach to predictive approaches that point to potential risk or potential value. What is more tangible—the ability to systematically manage price increases and boost profitability by 3-5%, or an opportunity score with limited context saying that this customer might be of interest? One equals a guaranteed ROI if executed on swiftly, while the other one has a lot of 'maybes' hidden in it, particularly if there are gaps in your base customer data (which will reduce the accuracy of the prediction).
Does this mean that predictive has no place in digitization strategy for enterprise customers? Not at all. Predictive has its purpose, and the possibilities from data science are truly exciting. However, we believe in (because we’ve seen) immediate results gained from getting accurate, complete and actionable information about current customers into the right hands at the right time! In fact, if you don't have this precondition in place, the ability to actually execute an effective predictive strategy may be greatly hindered, if not downright impossible.
The upshot? Once you have accurate and actionable customer relationship intelligence available throughout the organization, executing on predictive gets a lot easier. And with better data available to data scientists, results get much more … predictable.
So in closing, recent events gave us a huge reminder that predictive models still have great limitations in their accuracy. But executing on what you can truly know and using concrete customer data to drive your business decisions today, that approach provides immediate and repeatable value while you leverage the exciting but still unpredictable frontier of data science.