How to Plan Your Social Integration Strategy

By Brian Sullivan at Mashable: 
No longer does a business wonder whether its site should contain social components and functionality. The utility of social plugins on websites has been established. Now the question is how to implement social technologies correctly — merely adding technologies like social login, sharing, commenting and game mechanics is not enough.
Unfortunately, the web is ripe with sites that offer social implementations that miss the mark. And often, businesses suffer from a “set it and forget” mentality, in which a product manager will slap a Facebook Like button on his sites and declare the business’s website now social. Quite frankly, that is not enough.
Instead, for marketers and IT professionals, it’s critical to ask “why do we have this technology?” and “how do we define success?” For a vast majority of sites, some combination of social plugins is worth the effort. The best way to figure out what works for your site is to test, test and test again. Be it A/B or multivariate testing, here are my thoughts on why testing is so important, and how to get the most from your site’s social technology.

Which Social Login Provider Should I Choose?

Testing your site’s social functionality starts with testing your social login providers. You can go to any number of sites across the web that only offer Facebook Connect for social login. (My company’s CEO, Patrick Salyer, explains why multiple online personas make it imperative that companies offer social logins beyond Facebook Connect.)
When I’m helping clients conduct these tests, I almost always find that a great way to start is by offering user-sets different login options to determine which social network APIs should be implemented. Sure, Facebook will almost always be the dominant player in share-of-logins, but we’ve seen the other players account for up to a combined 60% share of social logins. Are you willing to throw away those users and the insight that comes with them?
There’s huge value in adding a mix of other login options such as Twitter, LinkedIn, Google and Yahoo. We know from reviewing our client base that one set of social login providers that works for one site doesn’t necessarily work for another. Sure, you don’t need to run a test to realize LinkedIn probably doesn’t work for a site targeting preteens, but do you have a higher than expected number of visitors that prefer to use Google? You won’t know until you test.
Additionally, if your site has a user base in other markets like Europe, South America or Asia, you may also want to offer login through various international social networks like Vkontakte, Orkut, RenRen or Mixi to ensure your site appeals to global users. The only way to accurately find out which of these networks and identity providers should be offered is through testing.

What Is the Best Testing Strategy?

A/B testing is a great place to start finding out what does and doesn’t work for your site. A/B testing means having one control group and one variant group (sites often overlook the need for a second control group, which we’ll call “C”) to establish that the control group has a large enough sample size.
Rather than use a formula used in hypothesis testing, I prefer to use convergence testing. Convergence testing effectively means “A/B/C” testing, where A and C offer the same login options, and B offers a different set of login options. The “convergence” part means that A and C will eventually yield similar conversion rates (conversion in this case being whether users logged in or not). If A and C have no variable other than the being presented to different user sets, they should quickly see the same results. If the result varies it means you haven’t seen enough traffic to reach convergence (the point at which both controls yield the same result), and you should keep the test running until you do. B poses a different set of login options placed in a different order from A and C. That’s the delta you’re looking for.
Keep in mind that a successful A/B test is any test that yields insight into visitor behavior. If the change you introduce from testing yields a 2% decrease in conversion rates, the test was still successful – you learned the lesson quickly and now have confidence that your control is still the best way to go.
In today’s business climate, it’s crucial for sites to integrate with social networks and, with focused testing, to capitalize on that integration to realize a number of benefits. For example, by discovering 30% of your site’s users prefer to share via Twitter versus Facebook, you would likely offer Twitter as a prominent authentication option. While Facebook almost always leads in number of logins, Twitter tends to drive the most referral traffic per share of any social network (data my company has collected shows that shares from Twitter drive five click-backs, on average). Chances are that you probably would want to take advantage of that valuable referral traffic, and the only way to effectively optimize your site for Twitter sharing is through testing.
The above just scratches the surface of ways to better leverage social via testing. The key concept to remember is that in order to ensure successful testing, you need to let the process run until you see convergence. Keep that in mind and you’ll be well on your way toward making the best use of your site’s social technology.


Brian Sullivan is vice president of client services at Gigya, where he advises clients on implementing social technology. Gigya offers websites a suite of social technology like social login, comments, game mechanics and a social identity management platform.