Twitter has been promoting a new study it commissioned from Datalogix, which focuses on research tying online activity to the real world. According to the results, promotional tweets can significantly increase brick-and-mortar store sales.

Normally this might seem like a tenuous argument to make. After all, how do you follow people from social media to a physical store? I went through a series of email exchanges with Eric Roza, the CEO of Datalogix, and it sounds as though they have a way to make it work, although there are still potential shortcomings. But if you could start to tie social media to physical actions, it could go a long way to improving your marketing insight.

The Methodology

Let's look at how Datalogix works to connect behavior online and offline. First recognize that market research often uses pre-existing panels of users who receive incentives to respond to surveys. That is because recruiting people who are qualified for surveys and who will actually answer them is expensive and difficult. Panels at least let the researchers find enough people to respond, making the study possible. It's become a common technique. Ideal? No, not at all because you ultimately have a self-selecting sample, which can throw off reliability.

Here is Roza's short explanation of the Datalogix method:

Rather than relying on small incented panels, as in traditional market research, Datalogix analyzes purchasing behavior across tens of millions of households with hundreds of billions of dollars in spending. We match our database to Twitter's using anonymous hashed keys such as email or user name, and conduct a similar process with retailers where the advertised [CPG] products are sold. The retailers track product sales primarily via Loyalty Card purchases. The matches are completely fire walled from one another, with Datalogix acting as the trusted third-party, ensuring that no matching information is shared between Twitter and the retailers. In addition, Twitter engaged a Big 4 Auditing firm to confirm that no personally-identifiable information is shared between our companies, and that all reporting that we provide is done at the aggregate level.

Datalogix knows the common identities and can match data from Twitter--presumably including who receives what promoted tweets and who does not--with sales data from retailers. The retailers have the loyalty card information that is probably tied to an email or social media address.

The Results

Roza says that everything is controlled for, including having the same preponderance toward possessing loyalty cards. That sounds good, but if the retailers primarily track sales by loyalty cards, then they largely can't trace sales versus non-sales for customers that don't use the loyalty cards. So it still seems as though the results aren't representational for the buying public as a whole.

Still, knowing the behavior of those who have loyalty cards is important because are likely more or less your best customers. And the results from the study were interesting. Measuring across 35 CPG brands, users who "engaged with a brand's Promoted Tweets purchased more from that brand than a statistically identical control group" to the tune of a 12 percentage sales lift, as Twitter mentioned in a corporate blog post.

There's still a lot of information missing that you might want before taking the results as a given. Would results go up for people without loyalty cards? Are there greater degrees of matching for certain types of products? No way to say. But it looks as though testing social media marketing as an influencer on physical sales should be on any business owner's short list.