Big data needs to be as easy to use as Facebook, suggests Tulloch, which means employing ordinary people to do the testing — until it is
Companies looking to roll out big data platforms to their staff in order to ‘democratise’ access to data need to rethink their employment strategies when it comes to testing their platforms, according to Walnut Data co-founder and technical director Bob Tulloch.
Instead of employing the smartest graduates that money can buy they need, instead, to recruit non-computing people who will be better able to advise on interface design, and whether the applications they are building will truly work.
Speaking at this week’s Computing IT Leaders’ Forum in London, ‘Data Strategy: Building a Framework for Success ‘, Tulloch suggested that the big data revolution would effectively be over once the fast-expanding role of the chief data officer died out — and that chief data officers themselves ought to play the lead role in this process by building platforms that anyone can use.
«The users have got to be connected directly to the data with no intermediaries. Let them create their own content. Let them own the data. Unless people think of the data as theirs, you won’t get anywhere. They have got to own it. And the only way they’ll own it is to use what people are familiar with: web-based interfaces, » said Tulloch.
That, though, will require organisations to build and iterate platforms not targeted at data scientists or other technical specialists, but ordinary people.
«You’ve got make those web-style interfaces so that people are familiar with them. But how do you test them?
«The only way of testing is to recruit the wrong sort of people — not bright and shiny graduates, » said Tulloch.
He continued: «Bright and shiny graduates are far too bright and shiny. They’re eager to please.
«You want the older, the crankier, the awkward, the people with little or no patience. You want the mum, whose got two kids at schools, a house to run and a husband that’s off sick. You want the bloke, who’s six-foot four, has hairy knuckles and is slowly recovering from the rugby club ‘do’ the night before.
«You want those sort of people. These are your testers. They must have no programming skills whatsoever. They must be able to test your interfaces so that if they can’t get what they want, it’s your fault, not theirs.
«So, how do you do it?» asked Tulloch.
«My test case is a friend, who only dresses in bright orange, because his main job is ‘tree surgeon’. He’s a nice bloke, and he’s very good at hedging and tree surgery. But he’s not very good at computing. And he’s a really good person to test out my interfaces. If he can get it, I’ve got a reasonably chance that other people might be able to do it.
«Build, test, rinse and repeat. It’s never one shot — you’re going to do this many, many times.
«So implementation policy is very simple: eliminate data scientists, » he said.
In the process, he admitted, chief data officers would effectively put themselves out of a job, in the same way that chief electricity officers in the 1890s also got squeezed out: with the development of standards and greater ease of use, meaning that they were no longer needed for something that had become simple enough for anyone to safely use.
Computing’s DevOps Summit returns on 19 September. Attendance is free to qualifying IT leaders and other senior IT professionals, but places will go fast, so secure yours now.
