Let’s not dilute the significance of data: it’s really important. Data gives us insights into content performance that print editors of two decades ago would – to paraphrase Jimmy Stewart’s journalist character in The Philadelphia Story – have sold their grandmothers for. There’s a caveat of course: over reliance on data can be more trouble than it’s worth, so what’s an editor to do?
“Data is the sceptic’s best weapon” says business analyst Andrew Chen and it’s hard to argue the contrary. Data is quantifiable. It’s usually massive in scope and has the potential to reveal so much about so much of our organisations. Who wouldn’t want to wield it?
And, as we’re asking questions, why is it so important to recognise the distinction between data-driven and data-informed?
As we’re increasingly operating in an environment of information overload, we’re grappling with the twin concerns of financial viability and the quest for ‘remarkable’ content (remarkable as Seth Godin reminds us: that is, something is different and therefore worth remarking upon).
That’s what we’re all looking for, isn’t it? Those indicators of optimal performance which provide us with nuggets of gold which we can hopefully replicate. The bigger the data set, the more chance we have of finding that, right?
The data-driven environment is an attractive sounding one but “no matter how good the data, it has its limitations,” says Casper Sermsuksan. “It is only a snapshot of reality that doesn’t paint the full picture of our customer journey and behaviour”.
We’ve witnessed a transition in newsrooms, from the vogue for data-driven workflows, to a more data-informed approach. It’s no coincidence that this evolution sits neatly alongside the industry’s move towards embracing more meaningful analytics. Where we are overwhelmingly beholden to data, there’s a real danger that content becomes stymied by what ‘works’. Data-driven newsrooms know that judicious use of appropriate analytics is more valuable – and more cost-effective.
When we think back to the problematic viral content of three, four or five years ago, it’s fair to say much of the motivation to replicate those kind of ‘high performing’ stories stemmed from reliance on metrics that placed a high value on easy numbers without really considering what those easy numbers actually meant.
Now we know differently of course.
Local differentiates. Great content differentiates. Viral is not a differentiator.
That’s Troy Young speaking at the FIPP conference earlier in the year. When he spoke about the importance of standing apart from the competition and extolling the virtues of unique content, it’s interesting to think how far we’ve come.
But, as Chen says, ‘not everything is an optimization problem’. Insights from data are only as good as the data sets which inform them, and they can’t possibly reveal everything worth revealing. In fact, the way that metrics are set up can only really reveal reflections of the existing strategy, of the established patterns. It shouldn’t be confused with being a prophetic tool. Thankfully, successful newsrooms are demonstrating that working on and investing in data analysis doesn’t create data dependency as it does reinforce editorial gut instinct.
Some decisions simply can’t be made without human interaction or decoding.
"Believing you can run on one algorithm is like having a hammer and believing everything is a nail" – a lesson from @boosc that has value even beyond the subject of AI#DISummit #publishing pic.twitter.com/e7DnYBkfut
— Content Insights (@InsightsPeople) March 20, 2018
Take note from the world of AI. When Syllabs or Arago talk about data and robot journalism, they all take great pains to stress that it’s not a universal solution to a labour shortage. Neither is it always a feasible undertaking. The sectors in which it works – in reporting sports, local elections and even – in the case of Sweden’s Mittmedia – in the real estate sector, it’s because the data sets which are available are sufficient to be able to create useful content. Try to use data to generate a story about the crisis in Yemen and you’d come unstuck. Some decisions can be made without human involvement, but many can’t – and shouldn’t.
‘Use data as a drunk uses a lamppost: for support, not illumination’. There’s that brilliant quote again, as told to us by Juan Senor reminded us back at the FIPP conference in Berlin in March. That’s the essence of the data informed approach: using data to hone, refine and confirm editorial decision making. To use data not just as the sceptics’ best weapon, but everyone else’s too. It’s just great practice.
When we spoke to Christopher Pramstaller of Sueddeutsche Zeitung at Digital Zeitung last month he reiterated this point: if editors find that data contradicts their own gut instinct more than around eight times out of ten, they’ll find it very hard to trust that technological solution. This isn’t to say that editors only use data when it suits them, but rather that they’re only likely to use data when they can see it has legitimacy and has value in their workflows.
Sueddeutsche Zeitung also made the decision to call time on the use of real time analytics in the newsroom – and for the same reason: it was data that didn’t add anything to the workflows of its staff. Seeing what was trending at that precise moment didn’t help productivity or efficiency. They made a strong editorial and technological decision to cut back on that aspect of their analytics, and it’s worked for them. Other newsrooms will have their own tales of success and development.
So find that lamppost, but don’t worship at it. Let it guide the way, but don’t look to it for prophetic insights. It is only a lamppost, after all.