These days, there are lots of screaming fans on the analytical side of Data Science… AI, Machine Learning, IoT…. Scores, ranks and notifications aplenty. VCs chasing AI start-ups. AI beer fests and bake-offs. Commercials on TV. Wanna buy a tulip bulb? How about a no-money-down mortgage? Whoa. The hype cycle takes one’s breath away.
Here’s what I’ve learned over the last 10 years in the Data Science space:
Data Science is worthless without PROCESS.
Ugh. “Sounds boring… Let’s keep finding sexy INSIGHTS!”
Hold on there, Cowboy.
Everyone should care about PROCESS. Why?
PROCESS is how things actually happen.
Here’s the lifecycle of data-driven process:
- Process creates Action
- Action creates Results (and Data)
- Data creates Insights
- Insights improve Process (repeat)
Let’s walk through these with a real life scenario. An inside sales team of 15 people is focused on qualifying leads. They have been trained and conventional wisdom points to a sales process of 4 calls and 4 emails over 3 weeks. Simple. Good starting point.
Except… that most of these inside reps don’t follow that process. They sometimes call once and abandon the chase, sometimes they have a hunch and call and email 20 times. Sometimes they don’t call or email at all. And sometimes leads are still open after 425 days (no joke… you know who you are). Some reps like Insurance and others like Tech. Others like Lead Source B and Chicago Metro more than NYC Metro. The unsuspecting data scientist comes along and collects 2 years of data. Call it ~ 25,000 calls per year x 15 reps x 2 years… 750,000 calls, and at least as many emails.
Seems like a meaty data sample. It’s not. The data has some heft, but it was created by a random process. Behaviors (engagement pattern) and attributes (industry, lead source, revenues, etc.) are irregularly over- and under-represented in the data. This defines randomness.
Crappy data breaks down the 1 – 4 lifecycle above. Insights will be worse than crappy… they may seem to exist but are misleading.
And so we’re back to PROCESS from #2 above. Without it there are two issues:
- Data is garbage (already covered)
- Actions can’t be controlled
Human-based processes are the most complex, hard to control and most valuable to a business (customer sales and service). If process can’t be implemented consistently, there are bigger problems and Data Science is irrelevant. The business strategy can’t be implemented efficiently. Full stop. The training that you spent beaucoup dinero on got wasted. Your process is out of control.
If the process can’t be controlled, or said another way, implemented consistently according to a plan, then even if you found an insight about it, you could not change the process to apply that insight.
Related to this, most companies stop analyzing after finding what they believe to be a valid insight. If they have quality data and believe in the analytical approach, and the “smart person” who created it, it must be right… right? “Time to implement!”
But hold on…
Test Insights in the Real World
Analytical results are a product like any other. They need to be tested, and tested systematically. Do diet pills work? Do tiny Bose speakers sound better than the hulking ones in the corner? There’s no way to tell without… broken record… PROCESS. Do scores and ranks drive better business results?
With a consistent, controlled process, it would be easy to evaluate the impact of insights applied to it, and to change the process accordingly.
The hard challenge is changing human behaviors. Compared to this, data science math is a breeze. People don’t like to change, and our current sales and service systems were built before Big Data, so they react, when they should guide.
The Next Wave: Smart Process Automation
The next wave of high value business applications guide people, proactively and intelligently, through a process of behaviors, while realizing that people are not robots and need autonomy and flexibility. With better, human-oriented process control systems in place, all of the Data Science capabilities become feasible, more powerful and easier.
So… don’t believe the hype.
Data Science is a tree falling in the forest without PROCESS.
PROCESS is KING.