Unmetric’s proprietary machine learning algorithm estimates Reach & Impression values for any Facebook Post.
We used a training dataset based on thousands of Posts from multiple industries and brand pages of sizes ranging from thousands to millions of Fans. Our dataset contains information about Likes, Comments, Shares for Posts at various points in time after the Post is created along with the total number of Fans of the brand page when the Post was created, sector and the brand's geographical region. Using this data, we created velocity metrics based on different types of engagement metrics. These metrics along with absolute metrics were used as predictor variables to build the machine learning algorithm.
Our dataset has known values for both predictor variables and outcome metrics (Reach and Impressions). Since we know this, and because the outcome metrics are continuous measurements, we used Random Forest Regression Trees to build our algorithm.
Why Our Methodology is Superior to Others
Our approach to estimating Reach & Impression metrics is far superior to other approaches because of following reasons:
- Our approach is based on a machine learning algorithm which automatically captures complex patterns in engagement metrics, number of Fans, sector and region of brand that impact Reach & Impressions of a Post.
- We not only look at the absolute number of audience interactions but also the velocity with which a Post is accumulating audience interactions. A Post with higher velocity of engagement is likely to have higher Reach & Impressions compared to a Post with same number of audience interactions but with low velocity of engagement.
- Our approach assumes that Reach & Impressions is directly proportional to the number of Fans of the brand page unlike other approaches that assume that a Facebook Post reaches all Fans, which we know is not true.
- We capture the differences in patterns based on sector and region of the brand instead of applying same factors to all brands from different sectors and regions while estimating Reach & Impressions.
Get in touch with us at email@example.com for more information.
Please note: Our Reach & Impressions feature went live the first week of July 2016. Data for this metric will only be available after this time.