On Google Reader's Demise

It would seem odd if I didn’t have something to say about Google Reader, so here goes.

As regular readers might recall, one of my longest-lasting quests has been regarding managing RSS feeds.

I started using newspipe and IMAP to read news back in 2005, and it covered most of my needs:

• Read/unread state was synced across all my machines
• Flagging, searching and archiving were trivially handled
• I had an easy-to-use, relatively speedy mobile interface for it¹.

The reading experience wasn’t perfect, but even considering that these days we’re spoiled with smooth, purpose-built UIs it worked pretty well for me.

Google Reader changed all that for two reasons:

• It was fast. Seriously, awesomely, fast, in every way.
• The social features provided implicit ranking of news items, which despite highly susceptive to hype and astro-turfing, made it a lot easier to spot important news.

The second bit is, I think, the most relevant here. Before Google Reader came along (and again after it removed its sharing features), the one thing I wanted from an RSS reader was filtering. Or, rather, relevance scoring, and spotting relationships between items.

A lot of that can be done with Bayesian techniques and natural language processing, and it worked fairly well for me.

But you need a fairly large population (and datasets) to do this properly even in relatively focused knowledge domains.

As I see it, Google built the perfect environment for collaborative ranking (and many other things you get for free with such a large user base), and then proceeded to maim and ignore it for the next few years.

Given their laser-like focus on Google+ (which I still consider more of a mirage than a coherent vision), it’s hardly a surprise that they starved off Reader. But from a knowledge management and information retrieval standpoint, they basically threw away an entire lab crammed with potential regarding content classification, ranking, and ontological studies.

So yeah, there’s plenty of opportunity here, but definitely not in terms of APIs (despite a lot of enthusiasm, Reader’s API isn’t really something you want to duplicate in this day and age), or even user experience – there’s a lot of ground to cover there, but the basic mechanics of feed reading are well-known, and Reader emigrants will likely be more concerned with finding a comfortably familiar experience than jarring differences.

No, what people ought to be focusing on is improving the relevance of what an RSS reader gives to readers – whether it be based on their personal preferences, or on other criteria, but, overall, by adhering to the less is more mantra: less information, perhaps, but more useful and/or informative.

Sadly, I think it’ll take a while for people to get there in the rush to build the next best thing to Reader – and simply making things “more social” won’t cut it, for the simple reason that the human element isn’t always an improvement.

After all, what would you prefer: “Social” news where hyped items and pictures of cute kittens bubble to the top, or a more tailored experience where stuff you’re interested in is prioritized for you?

The right answer, in case you have any doubt, is a mix of both², regardless of whatever starry-eyed social gurus claim.

Remember that when you build your own news aggregator – I know mine will take that into account.