Twitter Analysis: Making Measurement Meaningful

@Tim | Measurement & Analytics

Here’s another great guest post from our friend Mike Layton, Co-Founder of the hot social monitoring startup, Symscio, Inc.  It is important to remember that measurement programs will vary, but hopefully this post will assist you in asking the right questions towards setting up a meaningful measurement program. If you have any comments or questions, we’d love to hear from you. If you would like to learn more, Symscio can be found at symscio.com and on Twitter (@symscio).

One of the aspects of Twitter that I find most fascinating is the variety of approaches that companies take to advance their business objectives. Strategies range from branding and promotion to competitive intelligence and market research to customer service and recruitment. The possibilities are vast to say the least and many organizations employ multiple approaches.

As with any strategic initiative, measurement is vital for gaining intelligence and maximizing efforts. A few areas where we can lean on measurement include:

– Organizing data so that it is readily available to assist in decision making

– Evaluating the impact of efforts to assist in the development of future strategies

– Identifying potential threats that demand immediate attention

– And everybody’s favorite – demonstrating results to upper management

So how do you measure Twitter?
What do you measure? What metrics are right for your organization? Given the variety of uses for Twitter the short answer is – it depends. Measurement must align with your objectives, reflect the nuances of your industry and deliver actionable data. To make this happen, measurement must possess two primary characteristics to be meaningful: accuracy and relevance.

Accuracy matters. Can you trust the data?
Accuracy matters and, fortunately, is the easy part. Only trust human analysis for your metrics. Automation simply will not deliver meaning. It relies too heavily on the use of keywords, offers limited metrics and is wildly inaccurate. Automation reminds me of sweeping the dirt underneath the rug. It’s fast and looks nice at first glance, but it didn’t accomplish anything as your room is still dirty.

To be actionable, it must be meaningful.
Making measurement relevant requires a little more attention up front. The first step is to address the collection of tweets to be included in the analysis. The best advice I can give is to not be restrictive in the search criteria. You can always remove unqualified tweets but if your search failed to bring in the qualified ones (most likely due to the overuse of keywords) then you may never know what has been omitted from your research.

Next, you will want to identify the qualitative metrics that tie back to your initial objective(s). Here are a few metrics that offer value to a variety to different measurement programs:

Audience: The beauty of Twitter is that we have access to profile information and past tweets to help us better understand “who” is tweeting (analysts, competitors, customers, etc.).Subject: Categorize tweets into different “buckets”, such as market segment or purpose (i.e. product feedback), to help recognize “what” is driving discussion and segment reporting for internal audiences.Sentiment (tone): Alone sentiment is meaningless, but when combined with metrics, such as audience or subject, it points us in the right direction towards pulling insight from the analysis.


Categories