The future of feeds: Understanding structured data

September 2015

iProspect Head of Product Strategy Alistair Dent argues that all marketing is becoming search, where context parameters are the new queries. His advice is to envisage feeds as ‘structured data’ that can inform advertising platforms. The job of marketers is to provide a platform with targeting criteria and appropriate responses, so the platform can then search among all responses based on a user's context to determine the most appropriate results.

"The future is already here. It's just not evenly distributed." William Gibson’s quote appeals to the way I'd like to see the world: that we have solutions to all of our problems, we just need determination and logistics.

Unfortunately I don't agree. The future is still coming. And it's going to be awesome.

The future of digital marketing means talking about feeds. But whether we're using CSV, XML or JSON, a file or an API, whether we're pushing feeds into a platform or arranging a platform to pull a feed from our server... The mechanics are boring. The sexy stuff is what we do with them.

As digital marketing gets more complex, our reliance on automation increases. The inevitable result is that the systems and platforms that help us manage our campaigns get more sophisticated and can do more of the low level work on their own. But they need more input from us at a strategic level: more input about what targets we want to use, what ad messages we need, and why.

A feed is quite simply a way to tell somebody something. Vague, right? Actually, ‘feed’ is too restrictive a term. What we're using is ‘structured data’ from a variety of places to tell somebody something. Let's look at a few examples.

  • We use structured data to talk directly to users. Whether using RSS to feed into people’s readers or direct data links from our website into our Twitter feed, we're uploading structured data to a platform that sends that data directly to users.
  • Sometimes we use data to personalise how we talk to a user. Dynamic creative is a great example. If a user has bought raspberries in 50% of their visits to a supermarket’s website, the supermarket can use that database output to customise emails to the user so that the top content block features a special offer on raspberries. The email template specifies the layout, but it's that structured data file output from the CRM system that customises how the supermarket talks to that user.
  • Apart from talking directly to people, we might use structured data to keep our media up to date with current status. Google Shopping lets us use a file to point their platform to latest images, stock status or current price.
  • Structured data is also a mechanism for determining ad targeting. If you create an AdWords campaign as a CSV file and import it into AdWords Editor or straight into the interface, you've created a structured data file to determine ad targeting: IF a user searches for this keyword, THEN bid this amount, AND show this ad.
  • Frankly, most of your existing website is based on structured data. HTML stands for HyperText Markup Language, and the ‘markup’ portion of that means structure. We use structured data to tell people something: in this case you might be telling Google what snippet to display or what deeplinks to show in results. Even your sitemap reveals the structure of your site, so Google knows the difference between a category page and a product page.
  • A user might receive structured data that indicates what's happening in the real world: weather reports, traffic reports, competitor pricing, the stock market... the list goes on. Lots of services exist that are designed to take real world information and turn it into structured data. Query that data and synchronise campaigns to it!

There are many places where we might get structured data and different people we'd use it to talk to, but typically we're talking to advertising platforms. The better we communicate our information to these platforms, the better they can perform for us.

All marketing is becoming search

We're used to thinking of search as a list of websites that might answer a query we have entered. But there's no reason search results should be limited to websites, or even information pulled from those websites (like the knowledge graph). Google Now is a great example of this change in action.

Google Now will look at every piece of information it knows about me and determine what’s relevant to me at that moment. Sources include websites, weather feeds, local listings and even my own emails. Without having to type in a query, it has searched all these sources based on what it thinks I want. The search is happening based on my context. It might include:

  • The content of pages I've visited
  • My current location
  • The time of day and day of the week
  • What's coming up in my diary
  • My search history

Context parameters are the new search queries. We need a new definition of search engines to work in this world. We can think of a search engine as a platform to choose resources from structured data based on user context. Any marketing channel where we're using structured data to define what we want to show a user and in what circumstances can probably be considered search. That's almost all of them.

Our job as marketers is to tell the platform our targeting criteria and the appropriate responses, so the platform can search amongst all responses based on a user's context to determine the appropriate results.

Search results aren't just destinations

We categorise searches as navigational, informational or transactional: 

  • Navigational searches are when a user looks for a destination on the web (or in an app). They know where they want to get to and need a search engine to help them find it.
  • Informational searches are performed by users looking for an answer. Will it rain? What time does Asda close? Who provides the cheapest energy?
  • Transactional searches are conducted when trying to complete an action. That might be a purchase, signing up for a new service or choosing an estate agent. In this case, a search engine helps the user complete a task.

Ads are traditionally destinations, pointing a user to a new place on the web. Destinations as search results work well on desktops, okay on mobiles, and poorly on wearables. As the world gets more accustomed to increased contextual data and new devices, we can expect to see ads expand from those that point users to destinations to those that provide information and actions. Structured data lets us define the information or action and the criteria when it is relevant, so the search engine can look through the user's context and the competing bidders in order to determine what to serve to the user.

A typical use case for serving information might be when I detect that an existing customer is within 100 metres of my store. If I know her buying habits, then I can determine a special offer that might interest her and push that to her device (or wearable) via Google Now. So I'm inserting criteria based on user history and current location, and serving a single simple message.

Actions are even simpler. If it's raining at the time a person normally leaves work, he might want his watch to present a one-click button to order a cab. The cab company can insert criteria based on a user's habits (which the device will know, but the company might never see) and the weather, and serve a button linked to the customer’s installed app.

Talking about structured data is talking about the future of advertising. The platform knows the user's context, I know my targeting criteria, and I tell the platform the destination, information or action to present.

The business of ideas