Qriously Launches Beta Survey Design Tool
By Christopher KahlerJul 16
Published January 23rd 2019
How many people saw my campaign last week?
This is one of those questions that perfectly captures what data science at Brandwatch is all about.
It’s a simple request, right? Absolutely. And it makes perfect sense why you would ask it. You’ve spent all that marketing budget on last week’s campaign, of course you want to know how many (real) people (really) saw it. This simple question should have a simple enough solution.
But start thinking about how you’d answer this question and you quickly realise it is far more complicated than you first thought.
Mentions from some sites will come with lots of data to help us work this answer out but others very little. But even when there is a lot of information it is possible to get this wildly wrong. For example, on Twitter it’s neither true that everyone that sees your post follows you, nor that everyone that follows you sees every one of your posts.
Yet you’ll still find a lot of reach or ‘impressions’ metrics based on calculations like adding up the number of followers of the author of every tweet or retweet. This is the kind of math that can end up with reach estimates larger than the human population of planet earth if you’re not careful, and we did see this in our research of the media monitoring tool landscape.
So it’s not so simple after all. But in marketing reports across the industry, these kinds of numbers are being used to assess campaign performance and validate budgets. We wanted to do something more scientific and while it’s impossible to actually offer the ‘real’ numbers we wanted to give our users a metric they can trust to be as close as possible.
The Brandwatch reach metric solves the Twitter problem by creating an estimation about how many of your followers would likely see your post on average, on the basis of the likes, retweets and shares that content tends to get on Twitter per amount of followers.
How many people see a social post also needs to be calculated differently to those that saw a news story or forum post so we also translate the metrics attached to any given post in Brandwatch into values that are relative to each other. This makes them comparable across data sources; enabling you to compare reach of content from different platforms and track development over time.
To achieve this we use regression models to convert all available information about a post into a number representing “average number of likes for posts like this” and then we make assumptions about how this number correlates to reach for that source.
This method allows us to improve the reach metric if and when we get access to new metadata making sure it can be kept as accurate as possible in the future.
The new Reach score is now visible to all Brandwatch users. Request a demo today to see it in action