Skip to Content

A New Model for Predicting Social-Media Impact

An economist at a digital ad agency devises a way to use Twitter and Facebook to forecast sales of everything from cars to tampons.
October 20, 2010

What’s social media good for? Marketers see it as a new way to engage with consumers. Economist turned advertising executive Jason Harper sees an additional function: as a real-time laboratory for measuring how multimillion-dollar ad campaigns are succeeding or failing to drive product sales.

Social soothsayer: Jason Harper of the digital ad agency Organic explains how his “Velocity and Acceleration” model helped predict sales results for a Kotex campaign.

Armed with a master’s degree in applied economics, Harper was hired three years ago, at age 30, by the Detroit office of Organic, a leading digital ad agency, to crunch campaign data for car companies–with the goal of seeing whether digital marketing efforts were helping move vehicles off lots. Assigned to Chrysler’s Jeep and Dodge Ram truck accounts, Harper had to figure out how to gauge whether TV commercials were increasing website visits, Twitter conversation, and activity on Facebook brand pages. He also had to calculate whether that online activity was leading to an increase in test drives at dealerships. “The biggest question with social media is ‘What’s the value?’” he says.

Instead of waiting to measure the value of social media after the marketing campaign was over–akin to looking in a rear-view mirror–he saw a way of using social media to foresee turns in the road ahead, to predict whether a campaign was on target to meet sales objectives. The approach caught the attention of Harper’s new boss. “We needed some predictive tools,” says Steve Kerho, senior vice president of analytics. “To hit [Chrysler’s] sales objectives, we needed to see this many site visitors, this many key activities, this many scheduled test drives.”

To gauge the predictive power of tweets and Facebook sign-ups, Harper borrowed the concepts of velocity and acceleration from the world of physics. To come up with those numbers, he had to collect data during three phases of a campaign: the baseline, or the number of tweets or Facebook fans before an ad campaign starts; the “hot zone,” or the main surge of activity during the campaign; and the “fallout,” the inevitable decline when the campaign is finished.

Under Harper’s model, which he calls Velocity and Acceleration, the idea is to constantly measure the number of related tweets, blog mentions, and Facebook fan sign-ups during the campaign. By using calculus to compute the velocity, or rate of change, of the tweets and sign-ups, Harper can easily calculate any acceleration–the rate of increase in velocity over time. Using these two metrics, Harper says, he can predict whether a mass marketing campaign will reach its overall goals within the first few days it begins running. The resulting curve typically takes a steep upward slope before leveling off, a pattern known in the industry as “the kick-ass curve.” Says Harper: “The idea is to predict the height of the plateau.”

The model came about during his work for Chrysler. Harper homed in on how Jeep’s TV commercials were driving traffic to the “Jeep Experience” website, as well as the rate at which the website was triggering sign-ups to Jeep’s fan page on Facebook. Then he tried to see if the social-media activity had any effect on the number of test drives. Using statistical-analysis software from SAS Institute, Harper came up with a correlation: consumers who engaged with one of Jeep’s online touch points were about twice as likely to schedule a test drive at a dealership. Since the auto industry is so focused on increasing test drives as way to reliably boost car sales, this was a promising start.

He put the idea into action during the first week of the Ram truck campaign, when the predictions looked dire. “We weren’t getting the numbers,” Harper says. “Truck sales weren’t going to be where we needed them.” He suggested that Chrysler simply change its TV commercials to increase the amount of time that the website address was displayed on screen. It worked: “They were able to increase that call to action, and we started tracking to our goals within a couple of weeks.”

In early 2010, Chrysler completed a previously planned switch to a different agency, ending Harper’s work for the client. But he was already running his experiments using an even richer set of social-media data from an entirely different client. Kimberly-Clark, the Dallas-based maker of personal-care products, was launching a major campaign for U by Kotex, a new product line meant to bring feminine hygiene into the social-media age. The goal of the campaign was to get five million young women to request tampon samples.

The TV commercials used humor both to provoke people into talking about periods and tampons in a more honest way and to poke fun at the absurd way that Kotex itself had advertised its products in the past. An edgy series of video vignettes on its site and on YouTube went even further. One Candid Camera-style video in a drugstore featured a clueless guy asking a series of random shoppers for advice on which product he should get for his girlfriend. Product placements on the TV shows of Tyra Banks, Chelsea Handler, and the Kardashian sisters reinforced this more open way of talking about previously taboo subjects. As some of these video segments went viral, Twitter posts and brand-page sign-ups on Facebook soared.

The numerous “hot zones” of social data resulting from all these media events were put into immediate effect to forecast future sales. “Organic’s Velocity and Acceleration model helped us project the plateau level of tweets following the U by Kotex launch,” says Aida Flick, the Kotex brand director at Kimberly-Clark. “From there, we were able to tie in the relationship between the tweets and the sample requests.”

The model helped Kimberly-Clark optimize its media spending, product placement, and website features in real time, in an effort to reach its ambitious goal. It also helped guide on-the-fly changes to the creative content. Experimentation revealed, for instance, that a green-and-black color scheme for Web pages drove the most sample requests. So did a stronger call to action for visitors to locate the nearest store.

These efforts yielded sales forecasts that turned out to be correct. “Thanks to the model,” says Flick, “we knew within weeks that we were on track to exceed our sales goals and deliver a 10 percent incremental market-share gain to the Kotex brand.”

In this case, Harper admits, the results were never in doubt. The creative for the Kotex campaign was so strong that “we were forecasting ahead of target right at launch.” Now Organic’s VP of marketing intelligence, Harper has moved on to apply his model to the launch of a line of prepared meals for a global packaged foods company. He hopes the campaign will provoke enough social-network conversation to keep his model well fed.

Keep Reading

Most Popular

Scientists are finding signals of long covid in blood. They could lead to new treatments.

Faults in a certain part of the immune system might be at the root of some long covid cases, new research suggests.

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.