Identifying Signal vs Noise in Online Trends
1. The sorts of things we call trends.
Smart brands have become remarkably good at monitoring the social landscape for anything that looks like a trend: from hashtags within inner-circles on Twitter to links passed between colleagues on LinkedIn and Facebook, to watching for emerging Google Trends search data. . . monitoring hasn’t been our problem for a long time. What most lack, however, is a sensible approach to making sense of what’s a true signal among the noise.
What constitutes a trend online, in the most basic sense? If you’re Google Trends, you care about search-term usage over time. This makes sense: for Google, search input is the thing that moves the most. A user typically only cares about a search engine for its output; their input is, for them, a means to an end.
Although Google has plenty of meta-information about the searcher, the trend itself is composed of that user’s input, when compared in aggregate with many others. Often, some of this metadata is presented as a filter of the identified trend itself. This can be very valuable analysis: an uptick in searches for “flu symptoms” is relevant to the CDC, and a downtick in brand asset searches speaks to potential marketing concerns.
Still, social networks provide a different kind of online trend that I think is worth spending more time exploring. Let’s consider Twitter.
There are 2 principal people who use Twitter:
- Like Google, Output matters. The content you read on Twitter is typically what you use the platform for. Your stream is your means to connect to the platform and, through it, the larger social world.
- Then again, Input matters here, as well. Twitter, unlike a search engine, is also a tool for self-expression. This builds the above output for other users.
(It is worth noting only about 10% of Twitter users post more than every other day)
With platforms like Affinio, you can also aggregate these metrics over a content-related filter. We’ve observed, for example, that in 2017 fewer Americans self-identify their location by their country (“USA” or “United States”), and more are opting to be noted for their city or state, instead. Metadata trends in this sense can be useful to make determinations about regionalism or national identity.
2. The Attention Economy and the properties of Interest Trends
At Affinio, we think a lot about the properties that make a trend interesting. Principally, we look at the trend’s underlying metric (like hashtags or content, generally; or for Google Trends, user search input). There are a lot of axes on which social metrics differ, but the ones we’ve found to support the most interesting trends include:
1. Ease in which the platform supports changing the trend’s metric
Hashtags, keywords, and following new accounts rank high here. Twitter supports these things above all else: publishing new tweets with hypermedia like hashtags that can be searched and found leads to deep user exploration, and following new accounts leads to wider exploration.
By the same principle, unfollowing accounts ranks low. You’d never see a trend map worth paying attention to the “Most Unfollowed Accounts” simply because the action is so counter to Twitter’s business model that the activity is buried multiple clicks deep, the software equivalent of being severely frowned upon.
Likelihood that a trend achieves permanence
With respect to Google Trends’ search input, permanence is totally dependent upon circumstances outside the search box. We only care to search about a celebrity or a presidential candidate to the extent that, elsewhere, we become curious about them. A search engine does not inherently create that drive.
Twitter, on the other hand, is built to be a feedback loop. Retweets are first-class citizens; following an influencer based on seeing their content second-hand is a very common occurrence, and leads to seeing that influencer’s content all the more regularly.
Content metrics, like hashtags and mentions, can be fleeting or seasonal: presenters at the academy awards see a 2-day bump of mention activity, but a very long tail of new followers, for example.
I consider the act of following someone on Twitter to be an essentially interesting activity – it implies not just a temporary resonance of that person’s content and their immediate relevance, but the belief that their relevance will continue into the future. As such, watching influence patterns develop on a macro scale is one of the best ways to understand a given audience on Twitter.
3. Reacting to Trends as an Organization
This is a great way to handle customer service, but at a glance, it doesn’t speak to a much bigger picture.
My suggestion here would be to pay attention to trends of attention, like new accounts being followed: those trends imply permanence, speak to a belief in influencers’ relevance, and affirm the basic function of the Twitter platform: to find new and interesting things on which to spend your time.