We are still, I would argue, in the infancy stage for social media monitoring, particularly as a research or strategic input. Monitoring for possible “tactical interactions,” such as opportunities for customer service or pre-sales questions, is immensely valuable, however. It strikes me that there are enormous opportunities even with tactical communications to model and segment responses, but the “modeling” one would have to do is likely unique to each brand, product or service.
Still, opportunities abound, if you’re willing to do a little work. One potentially powerful tool to make use of involves longitudinal data. Currently, much of the data thrown off from social media monitoring tools can be trended in aggregate (number of packaging problems over time, number of prospects per month, etc.) but a trended study of tweets is not exactly the same thing as a longitudinal study of tweeters twitterers people using Twitter.
Consider: trending the direction of sentiment over time (assuming this is done with sound methods you can trust) may be an excellent way to monitor the general zeitgeist of your brand – but what about the opinions of individuals? A simple measure of brand mentions, as I’ve discussed before, can be a random walk, but measuring an individual change in sentiment, or a movement from awareness to consideration to purchase, could be an immensely useful metric – albeit a thorny problem to solve.
The answer lies in modeling, and potentially in combining server data with survey data. We may not be able to accurately model an individual’s behavior, but if we can place an individual with some reasonable sense of certainly into a given bucket (“Aware of product but not looking to make a purchase in category,” “Aware of product, in the market for category but negatively predisposed to brand,” etc.) then we can make more sense of the data being thrown off by social media monitoring platforms.
This would be a significant project for a brand or even for one of the monitoring players out there, but it would involve taking an initial pass at some kind of natural language identification of the character of social media messages, placing users into tentative “buckets” based upon those messages, and then reaching out with a survey instrument to persons within each of those buckets to hone and clarify those buckets into actual behavioral clusters of people along the purchase continuum (or even the post-purchase continuum). What this vision would enable is the formulation of more accurate segment-based responses for brands, and the ability to measure the effectiveness of that messaging.
Imagine the results: a social media monitoring platform identifies, say, 1000 people who are in the market for a given product but are negatively predisposed towards your brand. A/B testing could send samples of those individuals differential messaging, and longitudinal tracking of those users could identify which messages were more effective at moving individuals from “Aware-but-reject” to “Aware-and-considering.” Again, we don’t want to get caught up in the actual individual, but if we can to some degree of certainty (through a combination of survey research and unstructured data) place them in one segment or another, a longitudinal view of the data could track the effectiveness of your monitoring and response efforts at moving people from one bucket to the next (ideally closer to purchase or complete satisfaction).
What do you think? Science fiction? Or already being done? What are some of your ideas for strategically tracking the messengers in addition to the messages?
Buffer
Pingback: What Is Leadership In Social Media? | Finding Answers