What I learned from seven years as the Guardian’s audience editor

Seven years ago I started working full time at the Guardian in the glamorous role of SEO editorial executive. For five of those years I’ve spent every working moment plugged into the richest realtime data source of any news organisation in the world. It’s been very like being in the Matrix but with less kung fu and more Polly Toynbee and Nigel Farage. This week I started in a new role, so here are a few of the things I’ve learned…

1. Data belongs in the newsroom

Whatever role you play in a news organisation, data should be critical to you. If you want to understand the ever-changing medium we’re working in — and ideally become an expert in it — there’s simply no alternative. One of the first challenges I faced at the Guardian was persuading editorial staff that Twitter wasn’t the internet. The fact that most people had no idea that Google sent us 10 times more referral than our favourite social network meant that we were making really bad decisions on a daily basis. Anyone involved in the process of commissioning, producing, publishing and promoting journalism needs to understand our audience and how they come to us.

2. Data isn’t everything

It’s odd to even need to argue this considering Facebook’s recent difficulties in detecting and displaying news trends, but I’m constantly fascinated by the news industry’s credulous view of automation. This BBC article from earlier this year is a great example of our occasional tendency to believe anything that tech companies tell us. The fact is that editing is a purely human activity. With a bit of machine learning, some natural language processing and a lot of scraping news sites to pick up on editorial signals generated by humans, it’s possible to make a page that looks like it’s been edited. But looking like it isn’t the point: all the value lies in the individual choices and the nuances around them. The kind of data we can get about how people are clicking on and then going on to read pieces is an essential factor in how we choose to commission and promote journalism. But the fact that it’s easy to come up with thousands of reasons why it would be legitimate or essential to do the opposite of the most obvious response to the data means that we cannot pass responsibility on to an algorithm with any confidence. We are always at our best when data is used to inform rather than lead the editorial process.

3. Sometimes you don’t need data

When I first started at the Guardian the idea of introducing data into the newsroom was greeted as warmly as the idea of bringing in a bucket of nuclear waste. The whole news industry has made real strides since then. But we also need to develop our relationship with data and start to understand the complexities a little more. Data isn’t always the answer to a specific question. We need to be careful about framing the questions we ask in a way that skews the data. We need to be clear about where our data is partial. We can’t expect to use historical data to predict a very different future. Data can lead to horrible decisions and pointless delays if it isn’t used judiciously. Throwing data at any problem is seductive, especially in a world where big tech companies often succeed because of their data. But going back to Facebook’s problems with fake news and the trending bar is a perfect way to illustrate that it’s not the solution to everything.

4. We should all be spending more time looking at time spent

We know that time spent with a piece of journalism can’t be relied on perfectly to reflect the quality of a piece. There are just too many other factors at play. But we still need to look more carefully at these numbers, particularly when we see clear patterns emerging. We’re still learning what the internet is capable of as a medium. But our perception of the best way to approach tentpole digital journalism still seems stuck on a model first established more than five years ago which involves throwing pretty much everything but the kitchen sink into the mix. Snowfall was a critical moment in the development of digital journalism. But this industry has allowed itself to ignore pretty strong signals and instead make convenient assumptions about what makes people read for longer. Sometimes text alone is the best digital format. Sometimes a picture and a caption is enough. Personally I’d love to see any body handing out awards for interactive formats or visual treatments demanding attention time analytics to be handed over as part of the submission.

5. Don’t say something’s an experiment unless you have a target

Don’t fall into the trap of allowing yourself to use words like experiment or test when you really mean that you just want to do something. If you’re not setting a target, measuring and then drawing conclusions you’re not engaged in an experiment.

6. Data that you don’t like is more important than the data you do like

If you’re aiming for a situation where data informs editorial judgment and editors respond with nuance then you have to be honest about it. Every time someone ignores a consistent trend in data, fudges a data point to avoid it spoiling a ‘success’ story, leaves out some data that would disprove a personal theory, makes up data on the hoof to avoid looking ignorant or generalises off the back of a single data point, we do huge damage to ourselves. It undermines faith in data of any kind. And it almost certainly leads to or reinforces behaviours that are at best pointless and at worst counterproductive to editorial aims. This is the single biggest problem I see in organisations that use data to feed into a subjective or creative process. Being absolutely clear about what we know, what we assume, what we personally believe and what we don’t know should be a golden rule.

7. Don’t treat global audiences in a lump because they happen to use the same platform

Facebook has 1.79bn monthly active users. Not all of them like cats. Some of them are interested in social work. Or flower arranging. Google processes over 40k search queries every second on average. Most of those aren’t for Justin Bieber.

8. When is the best time to publish something?

When the audience is awake and you have the room to promote it.

9. We should be thinking about who will read something and how

The internet may be of infinite size but news organisations’ resources are not, nor is our promotional bandwidth. Asking pretty simple questions about the audience for a piece and the way it will reach them should be at the heart of every commissioning decision.

10. There is no reality in which this is a good headline

None. At all.

And finally…

People really, really do want to read great journalism on the internet. If this weren’t true there’s no way I’d have been able to do this job for seven years.