On the future of publishing
David Boyle has led analytics initiatives at some of the biggest TV, music, and print publishers in the world. From work in political campaigns, to music label EMI, to HarperCollins, and in his current role at the BBC, David stands out not only for his deep understanding of analytical systems and exploration of emerging technology, but also for his ability to teach and win over his executive colleagues so that his insights lead to results.
Leading up to his talk at Pandemonio in February, we asked David for his thoughts on the future of publishing in general, and whether the industry can save itself.
Pandemonio: Will the book publishing industry reinvent itself?
David Boyle: No. Not from the inside. My heart sank at one of my first meetings inside the industry when we were told that digital disruption wasn’t going to be as bad as expected — print sales were holding up as were eBook prices, and kindle sales were slowing. Sure enough, from that moment onwards, people were way less interested in even simple, cheap, proven opportunities to make data-driven decisions.
Publishing is a labor of love for most in the industry, and they retreated back to using ‘skills and judgement’ without any checks, balances or inspiration from analytics. They foresaw a digital crisis which got me hired and gave me a mandate, but when the crisis shrank away in their eyes, it left me without wind in the sails of the transformation I’d come on to lead.
This is different from the music business or the TV business, where the disruption from digitization is much clearer to those involved. In both I’ve seen even the most creative and least data-savvy people embrace transformation via the new powers and capabilities that data-driven decision-making offered.
So, for publishing, I think the more interesting question becomes: What transformation is possible from outside of the big publishing houses? Clearly Amazon has led disruption on the consumption side, so that leaves the creation side. How can technology bypass publishers to help authors create great work?
P What differences are there between entertainment businesses that focus on content production versus those that focus on monetizing content through ad sales or subscriptions?
D Production versus monetization are two very different worlds. The publishing houses, record labels and TV production/distribution businesses I’ve spent time in are very different businesses from the book stores/services, music stores/services and TV channels/services that monetize the content with consumers.
In many ways the world is a lot simpler for those monetizing the content with consumers. They have:
- A clear set of costs (content acquisition, licensing)
- A clear set of consumers they want to acquire, retain and develop
- Direct data on how those consumers browse, consume and spend
- Control of the app/service/product, pricing, marketing etc.
Because of this, they can rely heavily on the data they generate in-house and the data science fu they can do on it. It’s data they control so it’s as clean and comprehensive it covers all content producers. Lovely!
In a production businesses, we have hardly any of those luxuries:
- We rarely acquire finished products, instead having to bet on and develop content from nothing more than a person’s talent and vision.
- We are given very little data from the monetization companies on consumer behaviour, wants and needs — and when we are given scraps, it’s only our content, so we can’t see how our content is performing relative to content from other producers.
And so we spend a lot more time joining patchy data from many different sources and teasing out what we can from it. Opportunities for proper data science are far harder to engineer and realize. I can’t count the number of data science projects I’ve started but that have ultimately failed from a lack of data or scale.
In my experience, this makes the monetizers better at what I call operational optimization — things like recommendation engines and CRM segmentation — because that’s easy and drives instant revenue for them. But because production businesses don’t have the data or the use cases for operational optimization, we spend our analytics time on bigger-picture strategic issues. In my experience it makes the monetizers more operationally smart and the producers more strategically wise.
P How important is an “omnichannel” analytics strategy for media?
D I think that,for the production businesses that I’ve spent my time in, whether they are record labels, book publishers or TV studios, it’s a distraction.
For most strategy-type decisions production businesses are destined to remain forever in a world where multiple data sets each need independently analysing and then comparing by a human to tease out the story. Our content simply plays out across too many different services. We’re never going to get decent data from most of those services. So we should do what limited insight work we can with each data set and then ‘join’ the results in the head of a smart analyst at the end of the process to draw relevant conclusions.
That isn’t to say that smart data science and even AI isn’t important to production companies — it’s actually vital to how we create and analyse several of those individual data feeds. But I don’t see a world in the near- or medium-term where the individual feeds come together in any kind of automated fashion.
P When management is reluctant to grant access to data sources, how do you persuade them to let you analyze it?
D I think this stopped being a problem a couple of years ago. Boy, did it used to be a problem! I can remember flying across the country and getting into the office before anyone else, and sitting at a colleague’s desk with a memory stick to get access to data I otherwise wasn’t given access to (although I won’t tell you where I was employed at the time!)
Access can still be a problem, but it’s more likely to be a database admin that doesn’t know how to automate exports or build APIs than a permissions issue these days.
Today, the reverse is often true. Companies are drowning their analysts. I think all executives now believe there is “gold in them thar data” and they are throwing data at analysts they sometimes hire on that basis. But without clear business questions and without any clear guidance or management. There is often a middle layer of people missing that are knowledgeable enough about both the data/analytical techniques and the business questions to be able guide effective and efficient analytical solutions to having impact on business decisions.
I interviewed for a data science job once only to be told I was better suited to that kind of ‘man in the middle’ role. I took it as an insult for a long while, but in the end it was an important lesson learned — I’ve seen that playing the middleman role can truly transform cultures in a way you just can’t do as a data scientist alone. And that projects that don’t have that middle person often fail to get traction or sustain as a result.
P What is a big analytical mistake executives frequently make?
D One very simple error is the almost complete lack of familiarity with and application of the scientific method by executives.
For example, I often see a lack of attention to detail combined with a reluctance to appear to undermine a colleague by questioning their work. As a mathematician at heart, it’s amazing to watch it play out time and time again. Anyone familiar with the scientific method knows that details matter.
Of course one tiny false assumption or error early in a process can fundamentally change the answer at the end of the process! And of course finding errors in a piece of work should be encouraged in order to make the work stronger!
Another simple mistake is the lack of formulation of hypothesis at the start of analytical projects. Too often the analysis is completed and then the executive decides in a very unclear way if they like or don’t like the conclusions. It’s then never very clear if they’re playing along or if the analysis genuinely confirmed their expectations.
If executives had the confidence to articulate their hypothesis at the start, not only would it be far easier to understand the impact of an analysis — there would be a much clearer process for evolving the executives’ understanding of the world.