Measuring Twitter sentiment the smart way
In the last year, Twitter has become the go to place for information about your consumers and potential consumers. But as a heaving hotchpotch of meaningless hashtags, cat memes and employees-gone-rogue commandeering their brand’s feed, it’s easy to dismiss Twitter at first glance and conclude you can make more meaningful assessments elsewhere.
That’s where tools like Crimson Hexagon’s eponymous algorithm change the script. It was put together in Harvard no less, so you know it’s probably really clever. The US company’s algorithm was described by Laterall founder Dan Naylor as a “heavy lifting tool”, before he shot a sheepish look at the Crimson Hexagon guys and adroitly added “but a clever one.” I think his description nails it.
The algorithm enables marketers to measure sentiment for a brand, ad campaign, product or just about any phenomenon using a fairly sophisticated form of natural language programming. They have a wealth of tools in their arsenal, including the backing of Twitter and access to their “firehose” streaming API; the full, unbridled back-catalogue of tweets unfiltered and uncensored. Their product is also language agnostic, so you can fine tune it in English and gather sentiment from many world languages, provided you have someone who can read and interpret that language.
Why counting keywords and engagement doesn’t cut it
Wayne St. Amand, Crimson Hexagon’s vice president of marketing, first dismissed outdated modes of measuring Twitter such as keyword counting or looking at your baseline stats of retweets, replies and use of your hashtag. A tweet may get a high number of retweets for all the wrong reasons, for example.
Consider what happened when HMV’s administrators hit the marketing team with p45’s. Within minutes, the Twittersphere was ablaze, as HMV’s feed delivered a blow by blow account of the company’s demise. In raw terms, these tweets garnered unprecedented engagement for the company, but it’s not the type of conversation they wanted and frankly a PR disaster. Anyway, it’s not for me to dance on their grave, so we move on.
What makes Crimson Hexagon different?
Amand informed that Crimson Hexagon uses a sophisticated keyword algorithm to account for people’s true sentiment on Twitter. And with an error margin +/- 3% and firehose access, you’ve got the whole of Twitter covered almost all of the time. The fun and slightly crude example he gave was along the lines of someone tweeting:
“My new Samsung* phone is the sh*t”
*A complete random example, not a comment on Samsung
Most sentiment tools would give this tweet a black mark and register it as a negative comment about a Samsung product, where really we know this is quite the opposite. Crimson Hexagon allows for colloquialisms like this and has an ever-increasing bank of language data to work from to constantly redefine what counts as positive, indifferent or negative sentiments.
There’s no question it’s a very neat bit of kit, but I think it’s the firehose access and ability to trawl vast mines of data that sets it apart, which is brings Dan Naylor’s comments back to mind – Crimson Hexagon is a very heavy lifting tool, and quite a clever one at that.
