Note: the websites and services described here no longer work after Gigaom ceased operations in early 2015. I’ve updated the links to point to the Internet Archive.
Gigaom operates a set of interconnected news, events, and research businesses, and though the news business is ad supported, the company’s real revenue depends on moving news consumers through the funnel to our events and research businesses. And of course, the only way that works is to provide value that news readers want, and never appear spammy.
Our most recent effort in that space is alerts. It seems everybody is toying with alerts of some sort, but our research and our own experience tells us that the alerts need to be extremely relevant to the reader to be worth the time and interruptions1.
So, to support targeted alerts, as well as to support recommending content from across our network, we built out a consolidated index of all content. This consolidated index was most important as the foundation for a number of other tools, but we also exposed it as a separate search site for all Gigaom content2. This allows readers to subscribe to alerts based on the topics they wish to follow.
But, enough of that. Here’s how we introduced Alerts:
You are a busy person. You jealously guard your time and you hate distractions. But you know that you are what you read, and that you can’t navigate an ever-changing world without a little help.
We have something for you.
Please meet Gigaom Alerts. We’ve been testing this out for a few weeks now (it’s still a beta-stage thing, to be very clear) and we want to invite folks to help us really put it through the paces. Gigaom Alerts allows you to tell us which stories and topics you need the most, and have freshly published blog posts and Gigaom Research reports delivered right into your inbox.
Alerts can be delivered as content is published, daily (with your choice of morning or evening delivery in your timezone), or weekly on your choice of days. And no matter how many alerts you sign up for, all the alerts for a given delivery time will be sent in a single, consolidated email. And if you want to change your alerts, just go to accounts.gigaom.com/alerts and sign in to do so (there are links to that in the alert emails).
Alerts is built on top of our consolidated search index, where you can preview previous content that would match any alert, and use the facets to filter down to the specific content that matches your interests2.
Finally, Gigaom Alerts are in addition to our many newsletters, though we hope to consolidate those in the future.
Alerts also demonstrate the strength of the identity system we use across all Gigaom properties. None of our users really thinks of having an account at Gigaom, but that’s how we know who to send the alerts and newsletters to, how we know if you’re allowed to read our subscription content, and it ties your comments together across all our properties. We’ve done a lot of work to make creating an account painless, and even more work to make sure that you can get back into it when you realize you need it.
Our identity system supports Twitter, LinkedIn, and Facebook log ins, as well as basic email and password. Users who start with a social network login and then try to add an alert or newsletter will need to give us an email address, so the CTAs for those are all designed to handle that with minimal impact to the user.
Not all of this is perfect, and there are lots of opportunities to make it all even easier and smarter in the future, but this is a solid start that we can build on.
Gigaom has ceased operations and the search and alerts services are no more.
However, I presented our work on this at a media industry meetup. I’ve re-edited the original meetup video to try to make the screen more legible and the audio more, um, tolerable:
The video spends most of the time on Gigaom Search, the public face of our consolidated content index which supported a number of efforts within Gigaom. Jump to about 13:30 to see Gigaom Alerts.
The behavioral characteristics of the reader (watching what they interact with) and their friends behavior (“you’re friends are reading…”) are all interesting and relevant, but such tracking and data mining are not without risks, and they’re being increasingly regulated (at least in Europe).
Additionally, knowing what people previously interacted with doesn’t tell us whether they’d be interested in fresh new content being published right now. At least, not without additional data we can use to relate the new content to the old.
The good news is that not only do most Gigaom authors use tags well, Gigaom’s content is also well-suited to automated term extraction using NLP. That means Gigaom content is well described. A post about Apple or Samsung will reliably have the correct company tag applied, for example.[return]
Search really deserves a post of its own. It’s based on a rich index of full text as well as detailed metadata describing the content of each post. That metadata is generated by the authors themselves, including metadata suggested to the authors by automated term extraction via NLP.
Additional tools allow us to identify new and emerging terms, as well as to consolidate variations of a term (including different expressions of the same item, as well as typos, and other variations) and correct them to the preferred term. This keeps editing easy while allowing us to maximize the quality of the metadata. These processes are ripe for additional automation in time.
All that metadata appears in the search site as facets the user can use to narrow a large number of results down to exactly what they’re looking for:
Users don’t need to use facets to find what they’re looking for (though they are required for Gigaom Alerts). We did lots of testing and fine tuning to make sure our fulltext keyword search returned similar results to using Google with[return]