Two Statements and Four Questions About Big Data
This series appeared on the Eyes on Analytics blog the week of May 14. It’s consolidated here in part because it was popular.
Consider the following two, distilled, points of view:
Statement 1:
“Big Data Analytics is going to change the way we do business. Sure, a lot of it will be routine “I’m okay!” status updates from sensors, but making sense of the key parts of it, like “help me, I’m failing”, will be extremely useful. Companies that were previously exempt from competing on analytics will be disrupted by new entrants who will compete better, either by being more effective or being more efficient. Big Data Analytics is already having a disruptive impact in marketing, where it never used to before, and is gaining huge traction in medicine. There is reason to believe that Big Data Analytics will cause better decision making in the organizations that chose to invest both in the physical infrastructure and in the cultural infrastructure that’s required to truly succeed.”
Statement 2:
“All the big industries that rely on data already have Big Data. Airlines, casinos, Internet arbitrage firms, logistics firms and especially finance already have all the data they need. Indeed, all of that data has made them dumber, not smarter. In fact, even in those sectors, there’s little evidence that executives use all of that data to make substantially better decisions, especially when it comes to big strategic decision making. Did anybody see the economy after 2008? This is all just a second wind of hype coming from the Business Intelligence industry, which has so far failed to make anybody smarter. Don’t buy the hype. The companies that have long competed on analytics, since the 1960’s, have nothing new to learn from this next wave of Big Data Analytics. Just ask the line managers what they need, they’ll tell you.”
Game of Trolls
It’s not fair to label those who hold statement 1 to be true as blind Gartner Hype Cycle finger clicking optimists out to make twelve points on the next deal.
And, it’s not fair to label those who hold statement 2 to be true as See-I-Told-You-So get off my lawn here we go again curmudgeons.
It is fair to say that some among us are trolling.
Let’s not play the troll the game, at least, not for this week.
Let’s assume that both statements contain truth.
Four Questions:
- Did anything really go wrong with Business Intelligence generally and Web Analytics specifically?
- Where does the assumption that better data causes better decisions come from?
- Is that assumption credible?
- What questions really matter that would cause statement 1 to come true and mitigate the concerns expressed in statement 2?
Two statements, four questions.
Which is really right, and under which circumstances?
Did anything go wrong in Business Intelligence generally and Web Analytics specifically?
The extreme skeptics towards Big Data Analytics argue that Business Intelligence failed. Others point to cases of success. Who’s right?
Today, we see:
- Major decisions about pricing, like those recently at Netflix, continue to be made without any analytical support (or, the analytics were completely ignored).
- Few people, to this day, really understand the numbers they’re looking at. (Few can explain the accurate definition of ‘time spent on site’, for instance).
- Hundreds of thousands of firms continue to compete just fine without any analytics at all, without any consequences or the cost!
Specifically:
- Business Intelligence systems are expensive and relatively hard to implement
- The world moves very quickly, so, frequently, by the end of the third year of a major integration the infrastructure is out of date (Latency)
- Most people experience data through Crystal Reports or SPSS viewers, which hardly inspire, and generally speaking, it’s a bad user experience
So, there’s good reason to say that BI failed, in the context of the expectations that were initially set.
The high expectations set for such BI systems frequently fail to materialize
And yet, BI is software engineering. And, failure is common in software engineering. Why would we expect 100% success in BI when the success rate in software engineering is so low? Why was this sold as a sure thing?
(Because people buy sure things.)
There have been failures. There have been expectations. There are plenty of scars to go around.
Moreover, everybody, including the people funding these projects, believed that a better dashboard would make them a better driver of a car.
The high expectations set for teams of people frequently fail to materialize
All too often, we expect that we only have to explain a concept once, and that an entire team of people will understand and retain that knowledge.
How many times have you explained the difference between a visit and a daily unique visitor? Or, what time spent on site really means?
It’s not that everybody is stupid or ignorant. Those traits tend to be normally distributed and they tend to cluster in areas of the economy where stupidity and ignorance thrives.
It’s certainly the case that learning takes effort and many people are lazy. It’s also the case that most people don’t spend all day working with this material. Is it any wonder that people forget? It’s not in their job description to remember anything specific.
We expected people to improve their numeracy. To be just as comfortable with a trend line as they are with a word processor. Most analytics professionals expected much better collective decision making.
We expected so much more of people.
What went right?
BI and web analytics made certain individuals a hell of a lot smarter. While the plural of anecdote isn’t evidence, I can say that it made Scott, a line manager I worked with in 1999, much smarter. He used analytics, in real time, to optimize the price of drink specials at the night club he managed. That’s right – contrary to the popular Strata laugh line – certain managers are really capable of making decisions in real time.
Centralized BI practices caused a massive reduction in the bullwhip effect in supply chain logistics. It also led to much more efficient use of warehousing space.
A lot went right. And it made a whole bunch of individuals a whole lot smarter.
What went wrong?
The promise of a new technology didn’t deliver all the benefits as expected. The possibility of failure wasn’t discussed, and the expected results – both in terms of a change in performance and decision making – wasn’t fully realized.
Even those who are hyping Big Data Analytics, and those who are playing down Big Data Analytics, could agree on that.
Where does the assumption that better data causes better decisions come from?
I don’t know for sure.
But I can point to two possible sources:
- The enlightenment and the scientific revolution
- Robert McNamara and the Whiz Kid movement
The entire scientific method is predicated on data. A hypothesis is either accepted as truth or rejected as false based on the data. There is no other arbiter. Faith or strength of opinion has nothing to do with it. The data decides.
The assumption that greater knowledge causes greater outcomes flows from that fact. That, if you’re honest about being wrong, that everybody benefits. (Disturbingly, that trend may be reversing in The West, as negative findings have been disappearing from most disciplines.)
You might not be aware of it, but much of what we call Business Intelligence today really started taking off when Robert McNamara and the other whiz kids got back from the. Rule #6 from McNamara is “Get The Data”:
“Even if you don’t have the resources to access everything you need, start with what you have, even that data will show you where to go.”
The great grandfather of the discipline of Operations Research, the great trunk from which marketing science and information management branched off, is based off the assumption that data causes better decisions.
Is that assumption credible?
After all, that does seem to be at the root of those staunchly against Big Data Analytics.
Is the assumption that better data causes better decisions credible?
It depends.
If evidence to the contrary of an individuals aspiration comes to light, and that individual refuses to update their expectations or aspiration, then even the most pristine, accurate, precise and real time data will fail to change their mind.
If evidence to the contrary of an individuals aspiration comes to light, and that individual updates their expectations or aspirations accordingly, then it will be effective at changing their mind.
The key element that decides the effectiveness of data is the human.
Great data can cause great managers to make better decisions.
Great data doesn’t cure ignorance.
Maybe the broader commentary on the value of Big Data has more to do with optimism and pessimism about how our human systems change than it does with the ability of the technology to deliver.
Consider the following two, distilled, points of view:
Statement 1:
“Big Data Analytics is going to change the way we do business. Sure, a lot of it will be routine “I’m okay!” status updates from sensors, but making sense of the key parts of it, like “help me, I’m failing”, will be extremely useful. Companies that were previously exempt from competing on analytics will be disrupted by new entrants who will compete better, either by being more effective or being more efficient. Big Data Analytics is already having a disruptive impact in marketing, where it never used to before, and is gaining huge traction in medicine. There is reason to believe that Big Data Analytics will cause better decision making in the organizations that chose to invest both in the physical infrastructure and in the cultural infrastructure that’s required to truly succeed.”
Statement 2:
“All the big industries that rely on data already have Big Data. Airlines, casinos, Internet arbitrage firms, logistics firms and especially finance already have all the data they need. Indeed, all of that data has made them dumber, not smarter. In fact, even in those sectors, there’s little evidence that executives use all of that data to make substantially better decisions, especially when it comes to big strategic decision making. Did anybody see the economy after 2008? This is all just a second wind of hype coming from the Business Intelligence industry, which has so far failed to make anybody smarter. Don’t buy the hype. The companies that have long competed on analytics, since the 1960’s, have nothing new to learn from this next wave of Big Data Analytics. Just ask the line managers what they need, they’ll tell you.”
Four Questions
- Did anything really go wrong with Business Intelligence generally and Web Analytics specifically?
It didn’t meet expectations. The technology failed often. The people failed often. Read part 2 for the expanded version.
- Where does the assumption that better data causes better decisions come from?
It’s hard wired into the scientific method, and, more recently, into Operations Research. Read part 3 for the expanded version.
- Is that assumption credible?
That depends on the people. Good evidence on good managers makes a difference. Good evidence on willfully ignorant managers is a waste. Read part 4 for the expanded version.
- What questions really matter that would cause statement 1 to come true and mitigate the concerns expressed in statement 2?
It’s the attitude of the people using the data.
There are those that view Big Data Analytics as a tool for advancing their personal aspirations. For instance, “I really want to go to big fashion shows for free, so, I need to find evidence that a co-sponsorship with big fashion shows are really going to move our bottom line. Go find me that evidence and don’t come back with an answer to the contrary.”
There are those that view Big Data Analytics as a tool for advancing their personal aspirations. For instance, “I really want to increase gross revenue by 10%, so, I need to find pathway and evidence to support that objective. I have a few questions about how the firm really makes money and from who – go find me that evidence.”
It’s greatest barrier isn’t really the technology. It’s the people.
What’s really different this time
This is the third effort in several to express this point of view. Here it is:
In 2000, to build a data warehouse to mine all the IRC and ICQ chat logs, you would be looking at a $30 million investment.
In 2012, to build a cloud to mine all the IRC and ICQ chat logs, you would be looking at a $300,000 investment.
The cloud, plus open source distributed computing technologies like Hadoop, plus the rise of a generation of data scientists who understand the power of decentralization and know how to use it, is what has changed. It has reduced the costs, increased the imagination, and is making possible a Cambrian explosion in startups.
There are big things happening on the technology side.
If North America is old enough to remember BI and really won’t change its attitude towards data and the way it makes decisions, if that’s what people who are in favor of Statement 2 are really saying, then that’s sad.
There’s a whole bunch of people in China, Brazil, Poland and India are too young to remember.
Thanks for reading this five part series on Big Data Analytics. If you want to leave a comment, challenge an assertion, or raise a point, you can do that right below.