For most companies, "Big Data" will be the latest tech-related shuck-and-jive-- a way to waste enormous amounts of money and get little or nothing for it. It'll be another one of those things like "social media" or "CRM" or "knowledge management" or "customer protal" or "intranet".
And this message won't do a thing to help, because it already begins with a bunch of assumptions. that usually aren't true-- and will sabotage the results. Here are the prerequisites that companies need to address.
1. We will make decisions based on the data. When I do a report saying that 74% of customers have mobile phones, 63% of those run Android and 52% of those order at least $1,000 a month from other companies via their phones, are you prepared to say "Damn... let's get an Android app done ASAP"?
Or will you say "Let's put that on the agenda for 2014"? Or "Let's ask marketing to do a study"?
If anyone on the leadership is fond of quoting Mark Twain about "Lies, damned lies and statistics" or of saying "Figures lie-- and liars figure", save your money.
2. We will make decisions even if we don't like what it tells us. When the data shows that we're better off wooing Pinterest users and shutting down Bob's $20 jillion Facebook likes campaign, can Bob use his pull to bury that finding?
What if I give you the finding from my last point, but the head of the web division (which has mobile) is an Apple fanboi or Windows groupie? Can they kill the plan dead in its tracks because they hate Samsung?
Do we have a corporate culture of consensus building-- meaning never implement anything that someone with clout doesn't like?
People do this work because they want to see their findings put into effect and see what happens. The third time you kill something for political reasons, they'll stop working hard, figuring that it doesn't matter what they find.
3. We will recognize that this is R&D and it can't be 'managed' and "structured". Most of the research I do leads me down blind alleys. Many studies are not conclusive. Obviously, if I never find anything useful, that's an issue with me.
But if you consider that time "wasted" because I didn't come back with "Do this and make thirty quazillion dollars selling to Gay Hungarian Women", you're a bad fit. You're like the people who want to implement Total Quality Management, and want the program to cut expenses by 30% and improve customer satisfaction by 25% in the first six weeks.
4. We will share information. Here's another common thing. You ask me to research the buyer behavior of men, because you want to sell them tools. In doing the work about what men buy and how, I come up with a new insight about men and their dogs. I do a little poking and realize the same behavior is true for women and cats.
Pet food is in a different silo than chain saws-- so does this mean I don't give it to them? Or do I not do it until you sell them my time under your "shared services" model?
(Which they probably won't buy, because they didn't commission it and have no idea whether it is valid. What they're going to assume is that you're trying to offload $50K of man-hours that produced nithing for you. So of course they resist.)
5. We will devote X% of our work week to asking questions and framing them in an objective way. I can hire all the geeks in the world, but you have to give me time to work with your operations people to figure out what we can do.
And "work with them" doesn't mean "Send e-mails that they sometimes answer." It means "sit around and brainstorm and talk and tell stupid jokes and then get an idea." The more we talk, the more some odd idea is likely to take root after casual conversation.
6. We already have complete and clean data. Garbage in, garbage out, you know. You have all that stuff ready for me to analyze? If you have one system for franchises, one for independents and one for company-owned stores, we probably have a problem. Especially if the indys don't have the same rules to follow or have to share the data.
You don't have all of those things, I don't care how good your data scientists are. You won't get insights back and you might as well shove your money up your Hadoop.
