This post is part of a series entitled The Four Pillars of Social Media.  This week’s topics revolve around the first pillar, Research.

If you’re aiming to tackle social media monitoring, then finding a suitable tool is simple. A stable of popular products such as Alterian SM2, Radian 6, and Sentiment Metrics offer users the ability to conduct complex searches through millions of historical records. The process generally entails inputting a list of important keywords–some of which are bound together by Boolean operators–and examining the output of records in order to answer questions such as, “which web communities are the most popular for a given keyword?”, and “who are the key influencers in these communities?”.

In theory, the social media monitoring tools provide these answers along with a variety of pretty charts and graphs that further illustrate the point. In practice, however, those of us that are tasked with reporting on this output are occasionally encountered by a sea of unintelligible records in which the charts and graphs are at best, misleading, and at worst, completely erroneous. Every client is different, and the variance between each keyword set can make one project a breeze while another project a burden. If you aren’t experienced in finding coherence within a large data set, you’re likely to be left with a disproportionate number of burden projects. But have no fear, adhering to these five key principles can drastically improve your efficiency and help you fully navigate through a sea of records.

1. Start with the most recent data – Your end goal may be to provide six months of historical data about a product, brand, or a specific subject area, but there may not be a need to immediately query all six months at once. If you’re critical, you’re operating under the assumption that the first set of search terms will need to be tweaked several times before returning acceptable results. By starting with the most recent results, your social media monitoring tool spends less time searching and returns fewer total records. This allows you more time to find negative terms, and more time to find recent relevant records that may warrant new keywords. It has the added bonus of ensuring that the top domains returned are likely to still be active and the sentiment will reflect the most current market perceptions.

2. Branded terms may be your bread and butter – When prioritizing which questions you want to answer through social media monitoring, you should consider that the most specific keywords will often return the most relevant results. Branded terms often offer the kind of keyword specificity that can perform the heavy lifting required to filter irrelevant results. For example, imagine how much more targeted and relevant search results for “Sausage McMuffin” would be instead of results for “Sausage Biscuit”. By leveraging specific branded terms, you can dramatically decrease the amount of time you spend manually filtering spam.

3. Be aware of ambiguous terms – Sometimes, branded terms are not the silver bullet for relevancy.  For example, for every specific car brand such as “Volvo”, “Mercedes-Benz”, or “Lamborghini”, there may be a “Dodge”, “Saturn”, or “Smart”. The latter terms also offer little in the way of negatives that could refine these searches; attempting to create a comprehensive list of negatives can potentially open a can of worms that sucks time away from other valuable areas of investigation. Therefore, if you are trying to gauge the relative volumes of conversation about Volvo versus Dodge, you’re better off comparing specific models such as the “Volvo S40” versus the “Dodge Avenger”.

4. Do not always equate high volume with high influence – When reporting on key influencers, it may be tempting to choose the domain in which a keyword appears the most often. Make sure you consider factors such as multi-channel reach, unique monthly visitors, PageRank, and other factors that may enhance the authority of one domain over another.

5. Show no mercy for Twitter – When searching for short-tail keywords, spammy Twitter feeds can often overcrowd the result pool. The solution to reducing the volume of these results can be to use more complex search terms that reduce the likelihood a result will be returned given then 140 character limit. You may miss some important domains, but you can be confident that the ones returned are more likely to be targeted and relevant. A better way to find important twitter feeds may be to find high value domains that also use twitter to release content and engage their audience.

Social media monitoring is a work in progress, but hopefully these five key principles will help you decide on how to proceed. Surely, semantic analytics will advance by both process and technology, but in the meantime, there is a large enough space for ingenuity in social media monitoring to drive a truck through.