What does “harmony” mean?
Does it mean we sing the same words and notes, or does it mean the words and notes each of us sings go well together?
As anyone who sang in the shower knows, the answer is the latter; leaving the metaphor, our message must not change to uniform itself with the rest of the choir and sing from the same score.
Message originality must be preserved, but it must be tuned to the fact that there is a conversation already going on; when joining a group conversation at a party, any socially not inept person uses the first minutes to listen to what the conversation is about before chipping in. Of course if you’re already drunk by then you may not do this, and this is usually a sure sign you had a few too many.
The first step in my methodology calls for some reconnaissance, i.e. using the fantastic privilege that the vast majority of conversations in the digital world happen in writing for a tuning exercise: how are they talking about the topic we’re interested into? And who are “they”? Where to they go to find conversation partners? Are we (maybe) already in those conversation, despite our silence? Are our clients, employees, advocates and detractors already doing something to our Brand?
The end result of this phase is a so-called Ontology (in fact, looking at the chart representing it, my good friend Rod re-christened it “The Onion Chart”) where the topic is divided in sub-topics, their relative importance is measured and a hierarchy is established.
Experience shows Ontologies are relatively stable over time (of course, I am sure there will be counter examples) which means that a yearly refresh is usually enough.
For all the ink that was poured over automatic sentiment analysis software, while I am sure we will eventually get there (maybe using Cognitive Computing) I still think the human mind is the unsurpassed tool to perform this phase, its only problem being its scalability: going beyond the dimensions that can be analyzed by a single analyst in a reasonable time, requires the addition of an organizative layer in the form of someone who keeps the individual analysts in sync and makes sure the end result is a close approximation of reality.
That approximation however balances out with the fact that we do not rely on statistical techniques to ensure that the sample is representative (as for example they do with opinion polls) because we analyze the whole universe and also the fact that we observe (public !!!!) conversations without intervention, therefore avoiding the influence that sometimes biases polls.