Social Signals and Search: What Google Actually Admitted

The year’s most useful piece of search journalism arrived this month with almost no fanfare. Danny Sullivan at Search Engine Land simply asked Google and Bing, in writing, whether they use social data in ranking. Tweets, shares, the identity of sharers. Both said yes.

Specifically: both engines look at the social sharing of links. Both attempt to weigh the authority of the sharer, since all tweets are not equal. And both apply this mostly to timely content and news-ish contexts, as a minor input among hundreds.

After years of speculation, that is confirmation, and confirmation is genuine news. It is also, predictably, being inflated as it circulates. ‘Likes are the new links’ decks are already in the wild. So this post does the two jobs the moment requires. Record precisely what was admitted, and calibrate what it changes, which is less than the decks claim and more than nothing.

What was actually said, kept precise

The interview’s substance, stripped of extrapolation. Social signals are used. Neither engine denied it, and both described it. They are minor: one input among the hundreds both engines invoke, nowhere near the weight of the link graph.

They are identity-weighted. The engines attempt to assess sharer authority, meaning a link shared by recognized, followed, credible accounts carries more signal than the same link in spam streams. Bing was notably more expansive about this than Google.

And they are context-concentrated. Most relevant to fresh content and timely queries, which squares with everything our velocity coverage traced this year. The real-time deals, Caffeine’s plumbing, the whole apparatus reorganizing around recency.

What was not said matters equally. Nothing about aggregate like-counts as a ranking lever. Nothing about social presence substituting for the authority the link graph measures. Nothing that turns a follower count into a rankings strategy. The precise version is interesting. The inflated version is a sales instrument.

Why the engines went here, and where it plausibly goes

The strategic logic is worth a paragraph because it predicts the trajectory. The link graph, search’s foundational authority signal, has two growing problems. It is slow: links accumulate over months while the real-time web moves in minutes. And it is professionally gamed, as a decade of link-building industry attests.

Social sharing offers the engines a faster, human-scale signal for the timely content the link graph cannot yet see. With its own spam problems, hence the identity-weighting, which is the engines telling you they know bought retweets are coming.

Our expectation, put plainly: social signals grow as an input for freshness and discovery contexts while the link graph keeps owning durable authority. The engines keep investing in identity assessment, because signal quality there decides everything.

What we would bet against: any near-term world where social metrics rival links for evergreen rankings. The decks claiming otherwise this month are, as usual, selling the trend rather than reading it.

The advice that survives, unchanged and now better-armed

• Create genuinely shareable content: the substantive, specific, useful material our content advice has always described. Sharing amplifies reach and accelerates discovery, now with confirmed search assists in timely contexts. And it builds the audience that produces links, the signal that still governs. The ranking assist is the rounding error on returns that were already sufficient.

• Maintain credible, active social presences, the moments-insurance and audience-fit logic of our platform coverage. Identity-weighted signals reward exactly the account that is real, followed, and institutional. The verified, alive account our crisis-planning piece prescribed now carries a small additional benefit.

• Decline the arms race. Follower-buying, retweet schemes, and the ‘social SEO packages’ arriving in inboxes now are bets against engines that have already announced they are weighting identity. That arms race ends where the link-scheme one is ending. Skip the cycle this time.

What to measure, if you want to see the signal yourself

For the analytically inclined, here is the observation kit. ‘Minor and context-concentrated’ is checkable in your own data, and checking beats believing.

Watch the timeliness cases. When your organization publishes something that earns real sharing (the report that travels, the announcement that catches), annotate the date. Then watch the query-and-landing-page data for that content in the following days. The pattern consistent with the engines’ description is faster discovery and search arrivals for fresh, shared content, versus the slower burn of unshared equivalents.

Compare deliberately if you can. Two comparable pieces, one promoted through your social channels and one not, with indexed-and-arriving speed observed for each.

What you should not expect to see, and can therefore use as a fraud detector for vendor claims: durable ranking movement on evergreen terms attributable to social activity. The confirmed mechanism does not describe that, and neither will your data.

An afternoon of this per quarter keeps your social-and-search model calibrated to observation rather than to whichever deck visited most recently. The empirical habit this year of announcements kept rewarding, applied one more time before the year closes.

Where 2010 leaves us

This confirmation makes a fitting close for 2010, because it completes the year’s story in miniature. Search spent the year becoming faster, with Caffeine. More personal, with the personalization we covered last December, deepening since. More immediate, with real-time results and Instant’s streaming. And now, confirmed, modestly more social.

Every one of those changes was greeted with an industry panic. Every one, calibrated, resolved to the same advice, under different headlines. Build the genuinely good site. Publish the genuinely useful content. Maintain the genuinely credible presence. Measure your own data. And let the algorithm keep discovering reasons to agree with you.

Five years in, that advice has survived every announcement it has met. We will carry it into 2011. Between the responsive-design work ahead, the platform battles brewing, and whatever the engines announce next, it will presumably earn its keep again. See you in January. Do the thing for its own value. The rest keeps being rounding.

The blog turns five in March. Thanks for reading.