But here's the interesting part: Vijayasankar has managed to make the corporate blogging tightrope more like the forum for open expression it was ideally meant to be. What do customers make of a big data critic selling big data solutions? The answer might surprise you.
Last weekend, Vijayasankar joined me on an analytics video hangout to hash the learnings in his new role. The first half of that discussion has also been released as a podcast (see embedded player below).
Having just written my own semi-cynical post on Internet of Things security, it was a good time for a contrast between analytics hype and field realities. Here's a few highlights.
On Vijay's role at IBM - and customer reactions to his blogging habit
We kicked off with a chat about Vijayasankar's return to IBM. Now three months into his new role as a VP of IBM's big data analytics practice, Vijayasankar is knee deep in building a practice and project delivery. But how do customers react to Vijayasankar's public views on big data? Vijayasankar:
It's kind of funny. I have a rather irreverent view on big data and analytics to begin with, and my day job is to sell the same thing that I think is kind of hyped up. But what I find is that a lot of customers trust me, because I don't BS about this. My views are fairly public on what is hype and what is reality. Most of my conversations with customers are actually very easy. We're almost always on the same page when it comes to what's hype and what works.
So I said to Vijayasankar: "I was just reading the Gartner Hype Cycle, and I've got to say the analytics space has more hype and bullshit than any other space combined right now. You take everything from Hadoop to Apache Spark to predictive technologies to the Internet of Things. What's bullshit and what's actually real right now?"
Vijay's answer? Marketing sex appeal is obscuring "unsexy" use cases where genuine value is being achieved:
The vast majority is still in the BS territory. The reason the is because we primarily talk of the marketing use cases, because those are easy to understand, but that's not where the real value is. For example, predictive has a very straightforward way of reducing warranty costs at one of my customers.
The general principle being: if a customer doesn't invoke the warranty, then the manufacturer makes a lot of money, but if the customer needs to use the warranty, then they lose money. It's a very straight win-or-lose situation, so they have all the incentive to do predictive maintenance for the machines they give to a customer. This is how they make profit. This, while simple, is not a sexy use case.
Analytics consulting has been through a "massive change"
OK - but in the process of achieving analytics results, don't we need a new consulting approach? Vijayasankar:
We have been through a massive change. When I left IBM, the vast majority of these big projects were either fixed price or time and materials, but now, we don't do a lot of either. There are, of course, a few projects where the customer wants to do it that way. But the vast majority of projects we sell now are strictly outcomes-based. We'll say things like: if we can improve your inventory turn rate, and/or X amount of improvement in inventory turns, you pay us Y dollars.
That's a much easier conversation with the line of business buyers. With IT this is tougher, but when line of business is the buyer, an outcome-based model is a lot more fun, a lot more easy for them to understand, because that's how they deal with their business partners too.
Beware "actionable BI" unless it leads to outcomes
I thought Vijayasankar was going to talk about analytics-as-a-service (which is kind of Aasy now that I think about it). And we did cover that. But Vijayasankar mostly talked in the language of outcomes. And yes, that includes IT sales, though the specific outcomes might be different:
With IT, we have a different flavor of outcome-based projects. For example, for a certain amount of money, you drive us to an SLA, so if you have priority tickets, we solve it. If you have projects, we do those projects. If you want innovation work, we do that innovation. You just don't need to worry about who is this person doing it, where does this person stay, do I need to see him 9 AM to 5 PM everyday all week?
The customer can completely avoid thinking about those terms, and just hold IBM to a service-level agreement on what they want, and then either pay us when we meet it, or beat us up if we don't. It's a much simpler process. There are no longer these big 500-page contracts. These contracts are much simpler, like four pages of easy information.
During our deconstruction of BI catch phrases, Vijayasankar also singled out "actionable BI" as problematic:
Customers have heard the analysts and bloggers talk about actionable BI a lot, and definitely want in. It's good that people are starting to think about it, but people need to tie their business process to analytics a whole lot more. My favorite example is: I can create an IT system that will tell you in real time if inventory needs to be moved from Texas to California, but knowing that you will make profit is only one part of this equation.
You also need to tie together everything else. Do you have people, do you have trucks ready to move this inventory? Otherwise, what do you do with this knowledge? It's useless. You'll spend a billion dollars creating real-time systems for everybody, but if you can't really have the business process to go with it, you might as well save the money.
On to the Internet of Things, now atop the hype cycle. Is the IoT for real? Planning for IoT security is a given, but is there a signal amidst the noise? For Vijayasankar, it comes down to identifying trends and exceptions. And you'll need statistical savvy to do that:
It is not particularly important that you read every little signal. It's essentially a trend that you spot, amongst a lot of sensor data that tells you whether something is going well, or something is not going well, or if there is an opportunity to do something. To understand that, you need a pretty solid grounding in statistics.
It is not just for sensor data. This is true for all kinds of analytics, where if I come and tell you that my model says there is a 60 percent chance that outcome A will happen, this is not a reason enough for you to go all-in on option A. There's still a 40 percent chance that option B can happen, which means you need a plan B. A lot of people who use predictive analytics just make that leap of mind.
All this talk of cloud, IOT, big data and so on, there is one underlying problem: people still don't appreciate data for what it is trying to tell you. On TV, you will hear daily, with the election season approaching, "So-and-so has a 51 percent lead." It doesn't really mean much.
That's a wrap - but wait, about that agile thing...
The podcast also covered changes in project length and delivery style. Vijayasankar expressed a surprising appreciation for agile methodologies given his past skepticism. Though as he was quick to point out during the podcast and afterwards, he remains skeptical of agile for packaged software implementations.
In fairness to Vijayasankar, his shift on agile is more of an evolution than an about-face. But I enjoyed giving him a hard time about becoming an agile fanboy. On a Twitter exchange with Mico Yuk, another BI expert who is not an agile fan, it became clear that the hyped up language and cure-all expectations around "agile" are the problem. The change in project delivery to shorter, iterative projects based on contractual outcomes is hard to argue with. If there's a better way forward for big data services, skin-in-the-game is a good place to start.
End note: The next video with Vijayasankar, which gets into his career moves and disillusionment with enterprise social media, will be released on my YouTube channel this weekend.
Image credit: Funny racedriver young man driving between clouds concept © ra2 studio - fotolia.com.
Disclosure: diginomica has no financial ties to IBM. Vijayasankar and I have had a spirited dialogue for many years on topics such as these. We both served as SAP Mentors, a volunteer role.