So…subscribe, set-up, run report, repeat. Easy as that? Um…not so much. You’re going to want to invest in a human due to the following…

2). Setting up search parameters and filters is labor-intensive.

Even the most robust tools are still just that – machines gathering data based on human direction. Not to disappoint any Sci-Fi fans, but we are a far cry from artificial intelligence. Unless you have an incredibly unique brand name that is not used in any other context anywhere in the universe (notice I don’t have any examples), you’re going to get false positives in your search results that will need to be filtered out.

Consider a keyword, “Silverado.” You’re likely thinking of trucks or a western movie from the 80s. But, what if I’m looking for senior living facilities by the name of “Silverado”? Sure, I can search specifically for the keyword phrase, “Silverado Senior Living Community,” but, we all know that’s not really how people talk, and in social media, that’s what people are doing…talking. To receive accurate and thorough results, filters such as, “not Chevrolet” or “not Western” (and dozens of others) are necessary. (Additional features like proximity searches can help with this.) This part of the process requires testing, modifying, reading results and a lot of trial and error. It’s a labor-intensive process necessary to achieve accurate results. I’ll paraphrase a quote from Tom Webster in his session at the Radian6 Social User Conference: You should be spending more time filtering out your results than setting up what you want to see.

3). Results need to be read for context and accuracy.

Recently, when sitting in a client presentation of a social media audit that we had just completed the client asked, “I know our brand name is somewhat ambiguous…How do you know that all of these mentions were actually referring to our company?” My response? “Because I read them.” <Cue look of bewilderment.>

First of all, you don’t know what to filter if you don’t read the results. Second, even the tightest filters won’t catch everything. And, the tighter your search parameters and filters, the more you rely on the computers to make “assumptions” on what you are looking for. Third, computers absolutely don’t understand context, syntax or sentiment. An automated report that we ran for one of our events showed negative sentiment was off the charts. Startling, for a moment, until we read them. Turns out, the computer assumed that “kicks ass” was a bad thing. And, a LOT of people thought it was a pretty kick ass event.

You can’t predict all contextual uses of a keyword. Best bet is to read results for accuracy before having to defend the data at the boardroom table.

4). Results need to be reviewed for statistical relevance.

One thing computers can do pretty well is crunch numbers. But, ask yourself — are those numbers relevant? If your search results return upwards of 10,000 mentions per month, is it necessary to analyze all of them? With a volume that large, a random sampling of the whole is probably accurate and more manageable. Or, will you take steps to filter by source, and measure what matters? For instance, maybe you’ll filter out things like Foursquare Check-ins or aggregated news mentions and instead focus analysis on posts from sources that are likely to be key prospects, influencers or advocates of your brand or your competition.

In that same vein, if you are analyzing too few results over a given period of time, you don’t have a large enough sampling to make relevant conclusions (especially with regard to sentiment).

5). Reports need to be customized and put into context for your business goals.

From widgets to word clouds and pie charts – oh my! What does it all mean? There’s no shortage to the number of ways to display results – by volume, influence, channel, comments, shares, and the list goes on. To truly make meaning of it all, however, it takes more than just running reports. A social media analyst starts with the goal, analyzes the information, and then interprets and reports it in ways that will contribute to that goal. Often times, default reporting mechanisms within monitoring tools can’t be customized to this extent. And, whether you use those charts or customize your own, they need to be explained, put into business context, and ideally, accompanied with some strategic direction or recommendations.

So sure…reports can run themselves. But, if you want meaningful results, you need to invest in the proper tools and human labor to get it done.