How to get in Google’s real time search results?

I’ve been experimenting with how Google’s real time search works, particularly as it pertains to Tweets.

To start off, I’m making one major assumption: That tweets are not just chosen at random (that would just encourage a barrage of twitter spam affiliate links — send enough out and some will get through). That leads me to my second assumption: there has to be an algorithm at work here to determine which tweets are chosen and which never see the light of day.

To test things out, I chose the search term ‘walmart.’ I chose this for three reasons. 1. People always have something to say about WalMart. 2. It’s a popular target term right now for people promoting affiliate coupons/discounts. 3. I thought it might be interesting to get a sample of real time results for a well-known store/brand.

Analysis

After parsing and saving thousands of real-time search results for ‘walmart,’ my preliminary analysis shows no correlation between profile elements (thematic tweets, bio link vs no bio link, etc) and the likelihood of having your tweet included in the real time results. It was worth exploring, but right off the bat, I’m going to go ahead and toss that hypothesis in the trash for the time being.

With the profile/thematic hypothesis out of the way, the next step I wanted to take was to compare Twitter search results to Google real time search results for the same term.

Comparing Twitter search with Google search

The first thing I noticed when comparing Twitter search with Google search is that Google’s real time results are updated with FAR less frequency than they are  on Twitter. Part of this is likely a usability decision. There is just far too much noise on a popular Twitter term to digest anything at all.

However, the decreased frequency is likely also a side effect of a more important process — namely, the inevitable filtering process that all tweets must go through before they are widely distributed on Google. Now, given Google’s overall strategy against fighting spam, it would make sense to me that the first step in the filtering process would be checking tweets for links.

In comparing Twitter results with Google results, I noticed that tweets with links in them are generally filtered out, unless the tweet links to a site that Google trusts (links to sites like the New York Times, The Wall Street Journal, twitpic, etc, seem to always get through).

However, tweets with affiliate links rarely get through, and when they do, the affilate landing pages are *almost* always on trusted domains like blogspot.com. In fact, out of a few dozen links that got through, none were to the actual product page on the site of purchase — they were all to 3rd party landing pages.

Manual Review Hypothesis

I have no proof of this, as I am not a Google employee and have no insider knowledge into their development of real time search. But knowing how they have operated in the past, I would be totally shocked if there was NOT some form of a human review process at play here. After all, Google has already resorted to manual reviews on their traditional SERPS, having seemingly resigned themselves to the reality that no algorithm can prevent spam from breaking through without human assistance.

The problem here of course is that Google can’t manually review every tweet that it spits out in real time, as it would be prohibitively costly. So what could they do to solve this? Why, use US as the manual reviewers, of course.

Here’s an idea of how it might work:

Step 1: Display a given tweet to 1000 users in real time search.

Step 2: If 0 of the first 1000 users click a tweet, then trash it. If >x of the first 1000 users click a tweet, then show it to 10,000 more users. If >y of those 10,000 click a tweet, show it to 1 million more. And so on.

Now, big question here is whether Google would start to trust the tweeter over time. For instance, if 99 out of 100 of my tweets became ‘popular,’ would my tweets start to carry more weight and require fewer clicks to get greater distribution? If so, they would be opening a whole can of worms Digg style — I’d be surprised if they do it, but you never know.

Next Steps

As I stated, having a link in your tweet seems to decrease the likelihood of your tweet getting widespread distrubtion. But you know what doesn’t? Responding to someone else. In its current implementation, if your tweet is in response to someone else, then their name is clickable in the real time SERPs.

So first off, I’m going to try creating linkless posts in response to a dummy account. The most recent tweet at the top of the dummy account will be a link to one of my sites (no affiliate stuff, just a straight link), so that there is a decent probability of someone clicking straight to my site from the real time SERPs.

It’s worth testing out — as SEOs, we have to find out how to make this new development work in our favor. The possibility of appearing on the front page for any term we want in real time is just too good to pass up.