To get a data architect I know worked up, just ask him about how customers end up buying the wrong tools.

How about sales people who push federation tools on those who actually need data warehouses?

“It all sounds extremely sexy,” says my source, who works for a major business intelligence vendor and whom I can’t identify. “You have a lot of people who exaggerate their ability to combine data to provide business solutions. … They don’t prototype, they don’t profile, they don’t actually think about the problem or do testing or even send some high school data analyst out with Excel to put something together that [the customer] might want. They don’t do that.”

Many sales people tout EII because that’s what they have to sell, he says. “The EII tools give you your data, warts and all,” he says. It’ll work fine as a data warehouse substitute “if the data’s pretty clean to start with, if it has a somewhat similar structure, if you can define the data you need, if the data’s relatively common across all the sources, and if there’s not much duplication.”

Even if the salesperson has a more appropriate tool than what the customer asks for, the customer may never hear about it. “‘Fine!,’” thinks the salesperson. “‘If you want to buy a hammer, that’s fine. If you want to buy a wrench, that’s fine. It’s not like I care. It’s just sales to me.’”

Just once, says my source, he’d like to hear one of these questions: “How long does it take for a novice to become OK at this task?” Or, “How long would it take for an expert to become proficient at these two things?” Or, “If I have a failure, what is your tool’s usual process for recovery, and what gives your tool more integrity than others?”

Mark Madsen, meanwhile, has been been thinking about similar problems but from a different perspective. He’s research director at the Third Nature consultancy and a keynote speaker at this month’s TDWI conference in Las Vegas.

One source of problems he sees is vendor marketing. “It’s all about ‘our tool does this’ or ‘has these features,’” he writes in email. “A lot of people don’t think about them that way. They think about them as ‘what this tool is for.’” People end up using an ETL tool for real-time synchronization, for example, or a federation tool in place of a data warehouse.

Even product documentation can lead users down dark paths. “All those docs that say what the features are help when you know what feature you want,” he writes. “When you’re trying to accomplish a task, you’re thinking in a different way.” A common result: convoluted solutions.

“I once did something in an ETL tool,” he writes, “and the product developer said, ‘That’s not how you do that.’ They had built around an improper conception of how users apply it.”

Design schools tell you that every user has a theory of how anything works, he writes, which determines their approach to it. Wrong theories explain why people push on doors that need to be pulled, for example. He says that this insight has made him change his approach to teaching his courses or showing clients.

“I’ve realized that I need to start with the ‘what this thing is for’ and move into what you do with it, and how it works.”

Mark may go into this more in his keynote at this month’s TDWI World Conference in Las Vegas. His long-running “Clues to the Future of Business Intelligence” — perhaps the “Cats” of tech presentations — has been one of the most interesting I’ve seen in any tech industry. I expect “Stop Paving the Cowpath” to be worthwhile.

Everybody knows what “retention” means until they have to design a metric. Ken Rudin, once of LucidEra and now general manager of analytics at the games site Zynga, thought that he and his team could “put something together” quickly — but it actually took “four solid weeks of discussion and debate.”

About 50 million people play Zynga games every day. It’s the leading online social gaming platform, according to Ken, and it’s grown from zero in 2007 to revenues of “a few” hundred million dollars annual revenue. Every day, the company captures 20 to 30 billion records of data, and Ken and his team use that data to improve revenue, viral marketing — and customer retention.

Zynga players play free. The revenue comes in a few dollars at a time for “virtual goods.” In the popular game FarmVille, for example, a player might get tired of the old-fashioned plow. The tractor upgrade costs $2.

“There are tons of different ways you can think about retention,” he laughs, “and which one should we use?”

How do you know when a customer has left? “Unless we don’t get a note saying, ‘Hi, we’re no longer playing,’ how do we know?”

Of course, no player’s going to make it that easy, so how long should Zynga wait before considering the player gone? A week? A man could have dropped his virtual pitchfork for a real vacation — or he could have plowed the last row.

Ken dealt with analytics all the time at LucidEra, but games were new to him. He’s learned a few things.

“It turns out, as you might imagine, that it depends on the game,” he says. The average simulation-game player tends to visit frequently, for example. Poker players, though, are much more likely to come back after, say, a three-month gap.

The retention curve also varies by the length of each player’s tenure. A new player who stays away 30 days is much less likely to return than a player who’s been at Zynga for years. Ken now puts users in three basic tenure buckets: “new,” “mature,” and “elder.”

Whatever question you try to answer, it has to be actionable. “There are metrics, and there are metrics that matter,” he says. If volume plunges, were the missing players mostly new ones? If so, it could indicate frustration; perhaps the games need better tutorials or less functionality at the beginning. Or were most of the missing the long-term customers? If so, perhaps the games haven’t offered enough challenge.

Ken expects growth when the economy improves. “When we look at what happens over holidays, such as July Fourth and Thanksgiving, usage really drops. Then it picks up as people go back to work,” he says. “[The games] are part of their routine. On vacation, players break their routines. They sleep late and spend more time with family. They don’t play the game.”

“It’s fascinating,” says Ken. “In analytics, so much of the problem is figuring out what the question really is.”

I think he means that it’s a great game.