8 ways to avoid being run over by analytics
8 ways to avoid being run over by analytics
My sister saw a man hit by a car in Bremen, Germany, where she lived. Having checked, in German, that someone had called an ambulance, she bent down and talked to the man in English. Bruised and surprised, he asked how she knew he spoke English. “It was obvious, you looked the wrong way crossing the road,” she said.
This is how I feel about a lot of analytics – we’re looking the way somehow and overlooking more obvious clues about intent, preferences, desire and what we should do next. Here’s some advice from two analytics heavyweights, Amir Orad, CEO, Sisense, a business analytics software company, and Sid Sijbrandij, Co-founder and CEO, Gitlab an application to code, test, and deploy code all in one go.
It’s taken from their panel a couple of weeks ago at WebSummit, moderated by Ron Miller of TechCrunch.
1. Maybe the most extreme example of looking the wrong way is a failure to realize that data is not, and will not, simply be a byproduct of doing business – rather every company’s business will be based on data. Your company is producing data that is potentially very useful to your customers. (Miller quoting Carl Bass, President & CEO, Autodesk – a maker of professional 3D design software and consumer applications).
Miller said that while a company can’t give customers specific information about their competitors, who are also customers, you could, for instance, highlight things successful companies – or the not so successful – do. Obviously, this relies on customers’ giving permission for data about them to be used anonymously and in aggregate.
2. Orad argued that if people understand that they and everyone else benefits from sharing data from as wide a pool as possible, they would be willing to share – it’s how well you explain and how trustworthy you are perceived to be. He predicted the imminent arrival of insights as a service and dashboard as a service; “Sooner or later you will turn [that information] into a service by itself and either monetize it or use it to benchmark.” And especially if you combine it with machine learning that guides best practice, such as provided by Marketo.
Sijbrandij reckoned the biggest mistake most companies make with the data they collect is that they don’t ask, “Is it actionable…There’s no point collecting data if your developers go down the wrong path with it.”
3. Think hard about the best way of finding out what you want to know. A/B testing is fine for simple things, but not where there are a lot of variables – like a national election, for example (the debate took place on the day after the US election). More complex issues require a greater depth and granularity of data, and as many second sources of data as possible, Orad said.
4. Integrate big data analytics into products to gain more insights because tools to enable that are much better than they were. Also, look to startups in this area as they tend to specialize in data and analytics for a specific vertical, advised Sijbrandij.
5. Another facet of integration, as Sijbrandij observed, is although it is common practice to use a different tool for customer experience than deployment, say, it is easier and more efficient to use the same product throughout. Indeed he described this as a “game-changer” in terms of efficiency and interoperability.
6. Orad pointed to the fact that there is so much more data out there, and a lot of it is free – so make use of it. [Editor’s note: you can read a short report about the economy of data in smart cities here, although the same principles hold good in other contexts.]
7. Sometimes big data is the last thing you need. Miller cited a recent conversation with Jean-Pascal Tricoire, CEO and Chairman, Schneider Electric, who’d commented that that when you have to act fast, you want small data immediately – when something goes wrong, you need to know the specific thing to do right at that moment.
8. Analytics should be all about timeliness, speed and agility, not archives. Orad said this is where technology comes into play, to mine data in seconds: “The trick is not to build monster data warehouses, but to use data to gain agility. Huge data is irrelevant – who cares about data from a month ago?” Sijbrandij added that data should be “where you need it and you also need an open source monitoring tool”.
This is my two pence worth, which is pinched and adapted from George Orwell – break any of these rules rather than do anything outright stupid. By that I mean anything that is looked at with technology as the first consideration, not customers and actions. If you don’t you’ll find yourself under the bus.