Month: March 2015

Ask for stories, not advice

You’re better off with a story than the advice the story derived from.
Every bit of advice is a mere summary of the story that the advice derived from. I came across this advice — OK, it lacks a story — back in December as I looked for background on a new source for a report I wrote in December.

In a Forbes article by Clowdflower founder Lukas Biewald, “My board fired me, here’s what I learned,” Biewald tells why.

Ask for stories, not advice. Every piece of advice has a more memorable and interesting story behind it. In Silicon Valley, there’s no shortage of smart, successful people wanting to give you advice and, as a first time CEO, I didn’t have the experience yet to distinguish smart advice from empty advice.

Slowly, I learned to find the story behind every lesson. For example, one advisor wanted us to invest heavily in intellectual property. The reason was because his startup was bought for its valuable patents. Another advisor told me to never raise money. Why? Merely because he once had a bad experience. And yet, another advisor insisted I raise as much money as possible. It turns out he had turned down investors and eventually ran out of cash.

When someone starts giving you generic advice, ask them for the real story behind it. Then, you can decide if the advice is relevant to you.

Stupid analytics gets them talking

If data analysis somehow ends up supporting an unwise choice, should we blame the data? Of course not, no more than we should blame a stethoscope for a bad medical diagnosis or fingerprints for a bad courtroom verdict. Common sense says the only ones responsible are the humans involved.

AllAnalytics contributing editor James M. Connolly wrote last week in defense of common sense in a good blog post, “Analytics won’t cure stupidity.”

Let’s admit it, we know the limitations of data analysis. The obvious truth whispers to us in the quiet of our own little minds, and that truth can also find a voice in private business conversations over a beer or lunch.

And that truth of the limitations of data analysis also finds voice in online comments, such as in those responding to Connolly’s post. Such conversations go around and around, often failing to convince anyone. They may have no effect at all on, say, the foolish executive who insists that data must has a predetermined result. But at least some people test arguments, let others know when they’re really off, refute facts, and even make a few people think again. I find what might be thought of as “campfire behavior” — known these days as collaboration and storytelling — exhausting, often annoying, but also crucial. This is where supposedly hard facts, including data analysis, assume the rightful position: subordinate to review by smart people.

Such dangerous ideas! The first comment in the string following the post sounded like a slap at Connolly himself: “This article shows you have the diplomacy and tact in handling issues from the viewpoint of a toastmaster.” The same person commented again a few minutes later with a myopia that’s useful for close examination of data but only obstructs decision making: “You’ve given me a choice: cursing the data or cursing the stupidity. Neither. I look at the facts and draw a conclusion very carefully. I am mathematically-oriented.”

“Toastmaster” Connolly replied, “No, Judith, you don’t have to choose between data and stupidity. Just be aware that even the best data project could place the results in the hands of the very fallible human…” She replied, seeming to have missed his point again.

Others joined in. One told a story about presenting analysis to executives only to discover that they didn’t care about the analysis, they only wanted a certain conclusion. He wrote, “The endpoint was already decided.”

Another asked what if the analysts are the “stupid ones”? Still another told about the eroded trust that results from predetermined results. And another commenter wrote, “I’ve heard that technology will never truly change the world because the problem isn’t technology, it’s humans.  I guess the same could be said of analytics.”

That’s right, the problem — and the promise — is with humans.

The best remedy for stupidity is a traditional one: conversation. It’s not perfect, and often it takes a long succession of campfires ignited and exhausted before sense emerges. And sometimes the result is still stupid. But it’s always smarter than leaving everything to the data, no matter how “careful” any data analysis may have been.

People tell stories anyway

People tell stories wherever they are, in every medium, day or night, in business or not. What can software do to help tell data stories?

Give them templates? Force suddenly mute would-be storytellers into a best practice?

Probaby not. You can imagine the stories about the stories. “I knew it’d be another god damn story” — “damn” being for the stale formula.

“People tell stories anyway,” says Donald Farmer, the Qlik vice president of innovation. What can software do? He says maybe the best thing is to just stay out of the way.

Data industry’s breakthrough starts with a smile

The tweet stream running from Tableau’s presentation to the Boulder BI Brain Trust on Friday morning reflected one dominant reaction: awe. That’s easy when the product is a new shiny thing, but not so easy for number nine. As I said in one tweet, Tableau demos are always thrilling — no less than it was seven years ago. It’s magic, and it’s about time we give magic its due.

The question is what kind of magic? Is it the magic of a car salesman and his “new car smell”? Or is it an enduring magic that “takes a licking and keeps on ticking”?

The demos, at least, keep ticking. So far, now nine versions along, so does the product. How else can we explain the apparent, sustained cheer among the legions of seasoned users? Go ahead, tell me it’s all a fake. The walk on the Moon was a fake, too.

If there’s any sleight of hand in the Tableau show — and Qlik’s, too, by the way — it’s that it’s all about data. Listen to any Tableau presenter, and you hear them tell about data discovery in the context of a story. Storytelling is where the magic happens. I had this data, I looked at it this way, then that way, and I flipped over into this view, and then over to this one, and by god I saw this… It’s presentation magic, but unlike “new car smell,” this magic works even months or years after purchase.

Tableau has its doubters, of course. One bitter agent of the “just the facts, Ma’am” creed sat next to me last September during Tableau CEO Christian Chabot’s keynote and grumbled, “He’s telling them they’re artists.” Remember how that old song goes, “Do You Believe in Magic”: “I’ll tell you about the magic and it’ll free your soul, but it’s like tryin’ to tell a stranger ’bout rock and roll.”

Sure, “You’re an artist” wins over the rubes in the cheap seats. But there’s more to it. In fact, many are a kind of artist, data storytellers. Forget Tableau’s “story points,” the cave painting kit for beginners, and look instead at what users produce. Look at not just Tableau’s stories about data discovery but at stories of discovery from users.

Perhaps the Other 80 Percent — the estimated mass of business users who’ve shied from the cold grip of traditional data tools — needs to feel swept up in discovery and data stories. That could just be the data industry’s salvation.

Impressions of Strata Conference

Strata buzzes. Other events go to sleep for long stretches. But Strata+Hadoop World, at least the one in Silicon Valley if not those held in New York and London, is the only event I’ve seen with buzz that comes close to the buzziest of all, the Tableau conference. And like Tableau, Strata is growing. It switched venues this year from the Santa Clara Convention Center to the much bigger San Jose Convention Center, and it still sold out.

As usual in the tech world, the tagline “make data work” merely implies the many elephants in the room, the humans. Who does the work? The lazy data sure doesn’t, and the tools are like the golden retriever that forgets basic training. The humans gather at events like Strata to train themselves to make technology do what they need it to do.

These are my impressions from a full day as reporter and industry analyst.

Mark Madsen’s storytelling session

Mark Madsen, the man many find remarkable for his “intellectual energy,” spoke for 20 minutes Wednesday morning about storytelling. How to begin a story? The usual advice, he says, is wrong. Don’t start with the data. And don’t start with “the story.” No, instead start with your intent. Business stories should always aim for action, whether it’s to explore and understand, to change behavior, or to just change minds. His intense, stimulating presentations — which he’s often made final just a few hours before, sometimes late, late at night — never wrap up without goring a sacred cow or two. The cows this time were visualization bulls Edward Tufte and Stephen Few, who advocate data density. But that’s not always appropriate, says Madsen. It’s more effective, he says, to ask whether the audience wants to see the details. Or do they just want “the basic interpretation so they can get on with the decision making and go out to play a round of golf”? If Florence Nightingale had gone by today’s dogma, she would have shown detailed charts to doctors — and persuaded fewer of them to sanitize their surgical instruments, thus killing thousands more from infection.

Teradata’s AppCenter

I had listened to the Teradata senior director of Aster marketing Manan Goel and vice president of UDA product marketing Chad Meley explain the new Teradata Aster AppCenter and its “build, deploy, consume” theme for self-service with big data. I had heard all about how it lets users fit functions together. I had heard about user-made apps that monitor a retail customer’s path to purchase or a patient’s path to surgery. It all sounded like fun, actually, and I said so. Their faces brightened. Then we went on to Loom, Teradata’s tool for weaving value from Hadoop’s snarl.

Pentaho: the big question that went unanswered

I arrived at my meeting with Pentaho with a question. Because no industry observer I asked could make sense of Hitachi’s just-announced purchase of this open-source warhorse, I wondered if their action was the other shoe dropping after JasperSoft dropped into Tibco?

Is Hitachi thinking about the Internet of things? No one knew. I asked three Pentaho representatives about one reaction I repeated one particularly provocative comment: “I’m surprised anyone would see value in a company that controls so little intellectual property.” The three paused, sighed, and looked away from the table. Finally, director of big data product marketing Chuck Yarbrough began to explain his position, and director of corporate communications Rebecca Shomair got up to close the door. We control “lots” of IP, they argued. True, 80 percent of the IP contained in Pentaho’s enterprise edition is, indeed, open source.

However, what one of them called technology’s “shifting sands”—the normal advances and adjustments all business software undergoes all the time—requires regular updates of the other, proprietary 20 percent. None of them would discuss this point. I suspect that the real value is the knowledge that comes with the workforce. The whole gang is staying, which at least ensures that customer support of Pentaho Business Analytics will continue without interruption. As if considering the broad landscape ahead, one of them stared into the distance and concluded, “It’ll be interesting.” Indeed, it will be.

Four disruptions that paved the way for Paxata

I asked Paxata co-founder Nenshad Bardoliwalla about the long-dead rumor that Tableau was about to buy Paxata — pronounced pax-AH-ta — which was just an appetizer for what I knew would be a meaty conversation. He had that cheerful laugh of someone who was happy with what he has: by all accounts, an interesting and disruptive success. Paxata is a self-service tool meant for ordinary users to prepare data for analysis. I wanted to hear more of a story spun by VP of marketing Cari Jaquet: the four disruptions that paved the way for Amazon, eBay, Netflix, and now Paxata. There was the mature browser; also machine learning at scale; also distributed computing; and also the infinitely flexible cloud that expands and contracts on demand. Paxata’s now got 35 customers, with seven new ones in the last quarter alone. So I asked him when he would buy Tableau, and he gave the best laugh I heard all day.

Get-to-know-you with Trifacta

Much later, it was get-to-know-you time at my first-ever meeting with Trifacta, a Paxata rival. CEO Adam Wilson and the officially no-title marketing chief Michael Hiskey told me about “data wrangling,” the cost and security advantages of all-on-premises data, the advantages of sampling versus the all-data approach. One more thing: “It’s all about Hadoop”; Trifacta requires it. Thirty customers have signed on.

O’Reilly report on active learning

You may not know yet what active learning can do for you, but let me tell you something right now: It’s coming to an analytics tool near you if it’s not there already. Read all about active learning — essentially, a human layer over machine learning that dramatically improves accuracy — in a new O’Reilly Media report that showed up in the O’Reilly booth. Download it here.

Two new sources

I met two people I’ll call for updates and opinion: WebAction director of marketing Jonathan Geraci and Metanautix CEO Theodore Vassilakis.

Articulating buzz

“What did you hear?” is everyone’s question to those who go out to wander through buzz. What I heard or overheard is what I saw in the buffet: self-service with excellent features along with inevitable shortcomings that show up only with close inspection. You wanted things to work exactly as you expected? Silly you! At the Strata buffet, the salmon was good, the eggplant parmesan was tasty and filling, the string beans had a little bit of garlic and were not overcooked, and the green salad was exactly like the salad I’ve had in dozens of other events in each of the last three decades. But there was no dessert, not even a bowl of fortune cookies. Those who saw silver chafing dishes and expected ceramic plates and silverware rolled in cloth napkins instead found paper plates, plastic forks, and napkins that tore on mere 15-hour stubble. Will the update for Lunch Buffet have cloth napkins and metal forks? Let’s put that on the wish list.