Why data wonât always have all the answers

I’ve just completed a crossword. Truth be told, it’s taken me a little longer than it should have because I got distracted by one of the clues:

16D: Fact (4 letters)
The answer was ‘data’. And that started me thinking. When it comes to CRM, are the two quite so directly linked? They might be. Data certainly can provide facts. But in order for that to happen, you need four essential elements in place first.

1. Asking questions
Is your data just sitting there? Is it part of an endlessly growing list that you use to feed a monthly email or the occasional voucher offer? Is it impersonal and irrelevant, a blanket approach that doesn’t get results, never has got results and never will get results?
Data needs to be interrogated. Without it, it’s just a list of ‘stuff’.

2. Asking the right questions
Companies collect lots of data, masses of the stuff. If you’re debating a new campaign, a change of strategy or new ways to engage with your consumers, it’s almost inevitable that hidden in your data are the answers that can help you. The trick is teasing those answers out.
The answers you get depend on the questions you ask. Focused, targeted questions help to laser in on the right data. Big, open-ended woolly questions invite big, open-ended masses of woolly data.

What does the right question look like? It doesn’t ask for truths about your entire customer base. Instead, it breaks down that base into segments. The more you can break your segments down the more refinement you can give to the questions you ask, and the answers your data gives back.
We’ve been working with Northern Rail recently to develop a CRM strategy that drives off-peak ticket purchases. To understand more about the company’s passengers we broke them down into segments: commuters, families, students and so on, and then again by value.

Why did we do that? Because asking the data how the huge variety of people that make up Northern Rail’s customers use its services is far too broad a question to generate meaningful answers.

But when you ask how high-spend commuters are using the service, then you can start to find real understanding.
And if you don’t already have the data piling up somewhere on a server, asking the right questions can help inform what data you gather, which leads us nicely to…

3. Gathering the right data
In 1982, Steven Spielberg’s ET was everywhere.

Keen to cash in on the merchandise frenzy, video game manufacturer Atari created the ET video game. Then two things happened. The bottom fell out of the home video game market. And the game, released a year after the film, sucked. Estimates suggest 700,000 copies of the game now sit buried beneath the New Mexico desert.

Wrong product, wrong market, wrong time. The point of course, is this: bad data is also the wrong ‘product’ for the wrong market at the wrong time. And like Atari’s video game, it too is little more than intellectual landfill.

So how do you know the good from the bad? As with so many things there’s a balance to be struck between collecting too much data and too little. You don’t always know the value of what you’re collecting until the data analyst has had the opportunity to mine your information and found relationships you could never have anticipated. So there can be value in collecting data that doesn’t yet have a specific purpose – because it might not stay that way.

In the case of our work with skin cancer, for example, we knew a number of ‘body level’ characteristics would be relevant, but didn’t know which ones until we’d analysed the data. So restricting the data collection at the outset could have sunk the whole project before we’d begun.

Conversely, we’ve all seen satisfaction surveys that ask the pointless, the irrelevant or the so enormously broad that answers could never be of practical use.

The data you collect has to have the potential to build understanding. You may not always know in what way, but you can usually identify the questions and categories that aren’t yielding useable information. Cull them, and you leave yourself with more manageable levels of better data.

4. Analysing the data in the right way
In an ideal world data and fact would always be linked. The former would always lead directly to the latter. In reality, of course, that tends not to happen. Usually, causality remains out of reach because proving A causes B needs more than data. It needs context and controlled environments – never easy to achieve when you’re dealing with human beings.

So we need to recognise that data almost inevitably won’t lead to anything as certain as proof. What data can do, though, is establish relationships and connections. The key is knowing when the data is sufficient to allow you to draw conclusions based on those connections.

That means not jumping to conclusions when the analysis is half done. It means not assuming option B is ‘fact’ just because you’ve disproved option A. And it means not over-analysing until you start seeing relationships that aren’t really there, or under-analysing to the point where a blip becomes a trend.

There’s a reason scientists work in clean rooms: so they don’t contaminate the materials they’re working with and skew the results. We need to examine our data under a similar virtual cleanroom – because it’s all too easy to find conclusions contaminated by outside forces.

Data is fact? I’d say it’s nothing so definitive – data is inference and trends, relationships and conclusions. But that would make for a really long crossword question.