Alisa Leonard put out an interesting post today on the heels of getting demos from both Collective Intellect and Radian6, two competitors in the social media measurement space, and probably some others as well. (Disclaimer: I used to work for Collective Intellect). As someone who used to work in the measurement space, as well as a marketer, I find it interesting that the measurement piece wanted most by brands and PR folks alike is post-level tonality. Specifically, the ability to decipher whether a post, video, tweet, etc. is positive or negative towards the brand or whatever subject at hand.

Just to be perfectly clear: I also think it would be great, however, measuring tonal sentiment is difficult - for humans and machines. So, even if you are currently using a manual approach (ie, you've got a dark room full of low-paid interns reading posts and viewing videos, then categorizing for tonality) â€" you're getting very inaccurate information.

Why? Plenty of reasons, here are a few:

• All people do not score a post in the same way: I might think a post is really negative, you might think its neutral
• Sarcasm: some people are sarcastic, others are not, but again, depending on the reader, sarcasm could be inferred when there is none

Basically, what I am getting at here is that particularly for written forms of social media, humans are not going to agree on tonality because how they read something is reflective of many things: comprehension, state of mind, attention level, etc.

So, is automated tonality better? Yes, but with a major caveat: Its better only in the aggregate. Anyone who tells you that they can measure sentiment at the post level with high accuracy is really pulling a number on you. Here are some reasons why:

Associations: determining tonality is complicated by the language used (how exuberantly negative or positive) and also how many different subjects are discussed in either a positive or negative way. For example, if a blogger writes about Coke, Pepsi and Diet Rite, and is positive about Coke and Diet Rite but negative about Pepsi, how is the post ranked? Even if you are only interested in posts about Pepsi, the automated tonality engine is looking for text that indicates tonality, so if it sees more positive than negative language, guess what? The post is scored as positive.

BUT, again, if you are looking for an aggregate view of all social media postings, then some of those misses become less important (unless you are measuring such a small number of posts that there is no real aggregate).

Social media measurement tools are very helpful in measuring certain things at a granular level, but tonality isn't on that list. For anyone out there investigating these tools, you will be greatly helped by having realistic expectations of the accuracy of tonality from any of these vendors.