The development of analytics as a function is nascent, with the top management of many firms often ignorant about how to implement the data in business. This is why it’s vital to keep certain principles in place. But independent consultant and data and CRM expert Mike Sherman has some tips. He was previously head of L!ving Analytics, under Singtel’s group digital life division, and spoke at the recent Analytics 2015 Singapore event.
1. Be modest.
Making projections can be hard, so be modest when doing so, said Sherman. Showing how people in history have made vastly inadequate predictions, such as then chairman of IBM Thomas Watson predicting in 1943 that there would be a world market for “maybe five computers”, he advised executives in analytics to do so with a touch of humility.
2. Focus on outputs, not tools
“One of the things you must remember when doing analysis is that the drill is the tool to get to the hole. The focus is the hole. Us in analytics, we love to talk about the drill.”
“That creates problems. You can have a really great drill that doesn’t create a ripple. You need to think what you are using it for. Start with the intended outputs.”
3. Business significance over statistical significance
Sherman summarised the concept of statistical significance, often used in research: “If I repeat an analysis, how likely is it that the difference will be in the same direction as before?” But he posited it was vital that the data needed to have business significance over statistical significance. He defined it this way: “I have to make a decision; If I repeat the analysis, how likely is it that I’ll reach the same conclusion about what decision to make?”
Focus your analysis on what matters, he advised. What decision do you have to make? What information will change that decision? Data can help to reduce decision-making risk, he said.
4. Porpoise, don’t boil the ocean
Here’s why a researcher is compared to a porpoise – the porpoise comes up to breathe, goes back down to swim and does this repeatedly. This is how researchers should look at data. Don’t look at it as conclusive.
“Start with a hypothesis, go down deep and look at the data, and come back up and see if you’re right,” he said.
5. Correlation is not causality
Sharing the example of this Dilbert cartoon to illustrate the point, Sherman also gave an example.
While at McKinsey working on a client, he said it had a dilemma. For a supermarket, does shelf space for a brand drive sales or does sales drive shelf space? Everyone thought that it was the former. But it wasn’t the cause. “Sales drove shelf space, not the other way around, showing a clear misunderstanding of correlation.”
6. Close the loop …
Sherman talked about sites such as Amazon as a perfect example of being able to close the loop.
“It knows what you browse, what you buy, and what you review. For instance, if one buys a book on Amazon, the search results you get will change. If you write a negative review on Amazon, the suggestions you get will also change. That’s closing the loop and using that to drive suggestions,” he said, suggesting that this was a good way for brands to look at holistic data.
7. Beware of useless numbers
A chart may not mean anything, and the numbers may not mean much. This happens if you’re not asking the question right, said Sherman.
Also, there’s a flaw of averages. For example, if you survey both cardiologists and psychologists on what the top drug companies are, and take the average of that, this will be meaningless.
Sherman debunked the concept of the “average person” as well. “What is the average person? Is he half male and half female? This misconception drives a lot of conclusions.”
8. Demographics are better than nothing (but only just)
Too much of the targeting today is based on generic demographics and can be misleading. For instance, we know gamers may be young, but what if you have players like a grandfather who likes to play with his grandchildren? Sherman asked.
9. Ask the right question
Sherman gave instances of wrong ways to question, for example, asking people if they were competitive about pricing. “Who’s going to say no, charge me higher?” But the truth is, few customers know the competitive pricing about certain goods or services.
10. Communicate clearly
Finally, one important thing for those in data and analytics is being able to tell a good story. “For data, you must be able to communicate clearly, think of it as an elevator speech. Think of it that way when communicating to the CEO. You must be able to do it in a few seconds. Do it in a one-page executive summary.”
He told of his experience at McKinsey, where everyone was taught to be able to do an elevator speech of a project.
“If the client is in the lift with you and asks, how’s the project going, you don’t say ‘good’. You need to be able to say – so here’s what I think you need to do,” he said.
“The key to a good elevator speech – synthesise, not summarise. Don’t give facts, give the meaning behind them.”