Twitter is a valuable data source for tapping into the opinions and behaviour of people in real time. As with most high volume unstructured data sources, pulling the signal out of the noise can often represent a significant challenge. Yesterday Twitter introduced new metadata for tweets; with the objective of helping developers pull out the most “valuable” tweets.
The problem is that “value” is highly subjective and will vary on the context within which it is being viewed. Ranking purely on the influence of the author or the sharing of an individual content item is potentially dangerous.
Truly valuing conversations is a much deeper challenge that must be based on many more factors.
Related Resources from B2C
» Free Webcast: Know Your Story, Understand Your Customer
Twitter is introducing new metadata for Tweets so that if you consume their stream – directly or via one of their data distributors – you will receive tweets tagged up with value levels; initially just none, low and medium. High will follow along later, and it has been mooted that advertisers will be able to pay to elevate the value level of their tweets. The aim is to make it easier for developers to surface what is arguably the better and more interesting content from otherwise noisy or high-volume tweet streams. If you have ever searched on Twitter you may have noticed that Twitter initially shows you “Top Tweets” for your selected topic. You have to deliberately switch to “All Tweets” view to see everything that everybody is saying on that topic.
The big question is what is Twitter using to classify tweets? Obvious criteria that they may be applying are author and engagement factors. If a tweet is authored by a user with a large number of followers, or receives a high volume of retweets or replies, it may be considered to be more valuable. However this approach may serve to simply reinforce a social network oligarchy where only those people who are already established get more extensively surfaced, which leads to more followers and so the cycle repeats. Serendipity and new opinions are the lifeblood of social networks, so it will be interesting to see how Twitter addresses this. It shouldn’t take cold hard cash for a new voice with good content to rapidly demonstrate high “value”. Factoring in active engagement also has its limitations. Viewing a content item may well affect your subsequent opinions or behaviour but Twitter has limited visibility of these levels of passive engagement.
The other major criticism of this approach is that value should really be judged by how useful the Tweet is to the people who read it, rather than an abstract assessment of the Tweet itself. As a brand, to truly value a conversation you need to take into account a vast number of factors:
- Participants – not just the influence of the original author, but of those you are engaging with and who are responding to the conversation. Are participants existing customers? Customers of a competitor? Do they just “look” like prospective customers? Are any of your employees participating in the conversation? What about the passive participants? Who is retweeting this conversation and who are their followers?
- Topics – is your brand or any of your products being discussed? What about a competitor or their products? A market trend or new development? Can you infer anything from the way the topic is being discussed that might lead to action? For example, is a product quality issue being discussed that may snowball into a bigger problem? Is a product being positively advocated that may directly influence new purchases?
- Place and Time – What is the visibility of this content item? Does the time that it is posted mean it will be front and centre on the stream of people who matter to you? What is the longevity of the conversation?
No automated system can currently evaluate all of these factors and identify the most valuable conversations for you. Twitter’s development is a baby step towards being able to filter out the noise but you need to look at the noise first, and determine whether their ranking algorithm is genuinely pulling out the valuable content for you. We would argue that a better approach is to work on pulling the signal out of the noise based on the factors that are important for you, which will almost certainly go well beyond those that Twitter is applying.