Every since leaving Retail Week, I’ve been mulling a series of posts about what online publishers can learn from online retailers, particularly in the area of personalising content.

Today I see Patrick Smith has beat me to it by posting an interview with John Butler, global head of media for Dunnhumby, the firm that manages the Tesco Clubcard and other retail loyalty schemes. The interview raises many of the issues I’ve been thinking about.

As Steve Outing pointed out a few months ago, supermarkets know a lot more about their customers than newspapers do. It’s firms like Butler’s that are largely responsible for that.

While retailers have in recent years learned to mine their purchasing data to build up a detailed picture of their customers’ behaviour, news organisations are still working much like pre-Clubcard retailers; editors (like retail buyers) can make no more than educated guesses about the aggregate demand of their perceived audience. Even armed with online analytics data, editors pick stories and their placement largely on the basis of little more than what they presume their audience – as an aggregate – is interested in, maybe backed by a little reader research. But there’s the rub, as Patrick reports:

What people say they think about something can only take you so far, Butler argued, what’s more important is what people do.

What strikes me about Butler’s comments was that loyalty programmes like Clubcard allow supermarkets to treat their customers not just as a statistical aggregate — a mass audience — but as an aggreate of millions of individuals with specific interests known by their actual behaviour. This allows them to better market their products to suit not just a median or mode interest of their customers, but to suit the complex overlapping interests of each individual customer.

It has always puzzled me why Amazon can send me an individually-targeted email offering me a product I’m likely to be interested in, but even the most sophisticated online news sites cannot. Retailers’ personalised emails are often uncannily accurate, while news sites’ emails, RSS feeds and Tweets merely play the percentages, bombarding readers with stories until they eventually click on one. It’s usually up to unbundling and filtering mechanisms like search and social media sharing networks to make that distribution approach anything close to personalised news. Publishers leave it to other firms technology to ensure that “if the news is that important, it will find me“.

Assuming I’ve opted in to having my personal data used in this way, every news site that requires registration should know what I’m reading and commenting on, and should be able to use its taxonomy categories to make educated (algorithmic) guesses of which future stories I am likely to be interested in. Local sites should additionally know where I live and work and should know how a geocoded story’s proximity to places that matter to me affects my likelihood of being interested in those stories.

Rather than sending me a breaking news alert by email, RSS feed or Tweet when an editor assumes a story is important to the whole audience (or some pre-defined segment of it), it should be possible to send an alert about every single story that is published – to only that subset of readers likely to have an urgent need or interest in that information. Some firms are working on this, of course, but it’s not an obvious feature on most news site.

As news sites seek to increase user engagement – particularly where they are attempting to convince readers to pay for their bundle of content rather than relying on the personalisation offered by third-party filters – attempting to develop Clubcard-like customer intelligence would be good place to start .