A few weeks ago, I wrote a piece on the marketing lessons we can all learn from the 2016 U.S. general election, knowing that there were many more to come.
One of the biggest lessons since learned concerns social listening, its current state and what we expect of it in late 2016. In the wake of the election results, social media listening outfits were quick to claim credit for seeing the Trump victory before it happened. But in reality, sentiment analysis is telling a much less definitive story than one might be led to believe.
Sentiment analysis and engagement proved to be softer measures of success than other indicators during the run-up to November. The traditional criticism of sentiment analysis is that it tends to get things wrong, failing to pick up basics like sarcasm and potentially misinterpreting reactions to various posts. As for engagement, it’s hard for analysts to attribute much to it, other than increased awareness or perhaps the volatility of a particular issue or news item.
While engagement seemed to be more of a core KPI than sentiment analysis, using those two indicators in predictive analysis still managed to do better than the polls.
But what digital signals managed to show a more compelling analysis story than social listening over the last election cycle?
An AI algorithm seems to have run up an impressive streak of correct predictions. MogIA, developed in India and named for Mowgli, the main character in Rudyard Kipling’s The Jungle Book, made the call loud and clear, well before the election. Its approach was different from those taken by pure-play social listening platforms, but if you look at the data used to feed the algorithm, it shouldn’t escape notice that it also leveraged data from search engines like Google in addition to data from Facebook, Twitter and other social networks.
Could it be that the AI simply had a better set of signals than the listening platforms?
One such signal might be search intent. There’s little that can be discerned from someone searching “How to vote” on Google, but if a follow-on search looks to find out “how to vote for Donald Trump,” that can be considered more of a hard signal than mere reactions to social media posts.
What’s the lesson to be learned? When we try to understand perceptions of our brand using signals from digital channels, not all KPIs carry equal weight. Looking to harder measures of success, rather than trying to tease a measure out of surrogates, should be the approach that succeeds more often.
The key is to map your various signals to the metrics senior management considers to be drivers of success. They could be in-store sales or measures of lift in brand awareness, as opposed to vote totals in swing states or the like. But whatever they are, it’s clear that your digital KPIs should be battle-tested and correlated with the metrics your boss will use to gauge business success.
Dashboards and on-the-fly analytics reporting, visualized through engines like Tableau or Domo, can be valuable assets in clarifying these correlations. And when your KPIs start to deviate significantly from your business goals, it should prompt a self-examination of what predictors are most useful. Correlation does not necessarily equate to causality.