Most active LinkedIn users are used to posting short updates using the “share an update” feature of LinkedIn. The number of views on your update tells you how many people may have seen your update. Effectively, the number of views of your update is the effectiveness of your update. However, earlier this year, LinkedIn disabled the feature to view a number of opinions on your updates. If you post multiple updates in a day, at most, you can know the number of views only on 1 update. This was so disappointing. I know many people have raised this issue with LinkedIn.
But there is a workaround!
We can use data from past LinkedIn updates on the company page and accurately predict views for your updates.
If you post updates on the company’s LinkedIn page, you can see number of views (LinkedIn calls it ‘impressions’), number of likes, number of comments, etc., for all your updates. You can use this data to predict how your personal LinkedIn profile updates will be received.
Data Used for the Predictions
LinkedIn updates work like this: If I post an update and one of my contacts (say Mr. ‘X’) likes the update, the update is put in the feed for all my contacts and X’s contacts. Thus, the more the likes of my update, the more the update is propagated to more people. So, there must be a correlation between how many people like an update and how many people view the update.
I extracted the data of all updates from my company page.
There were 30 updates that we had posted in the last 2 months. For all updates, the number of likes, comments, and impressions (views) was tabulated. LinkedIn defines ‘Interactions’ as the number of times people have liked, commented on, or shared each update. This data was used to predict the number of views for an update based on number of interactions.
So here is how I did it
- The first step was to find Pearson’s coefficient between number of impressions and number of interactions. It came out to be 0.95, which signifies a significant correlation between the two. The value of the correlation coefficient between impressions and likes was 0.93, and that between impressions and comments is 0.86.
- Now that we found that there is a significant correlation between impressions (for example, views) and interactions (for example, comments + likes + shares), the next step was to model the relationship between the two using regression. I used Excel for a building regression model.
No. of Interactions = No. of likes + No. of shares + No. of comments - Using the regression model, the relationship between Interactions and Impressions came out as follows:
(No. of Impressions) = 92.1 + 83.2 x (No. of interactions) - R-Square for this model came out to be 0.906. This indicates the percentage of movement in a number of impressions that can be expressed by movement in a number of interactions. Thus, a fairly significant number of impressions can be due to changes in interactions on the update.
Validation
During the last few days, I have applied this equation to predict the number of impressions on my updates on personal as well as company pages, and it has come out to be accurate with +/- 5% error.
What next
There are many ways to improve this prediction:
- The equation means that each of my contacts who likes or comments on an update brings 83 additional views to the update. Now, for someone else who has ‘rich’ contacts (for example, contacts with large number of contact bases), each like may bring a significantly higher number of views. Thus, the average number of contacts of people who publish the update should be a factor to be considered in the analysis.
- The time of the day when an update gets published is very important. If I post an update at 3 a.m., surely very few of my contacts will see it. It has to be a factor to be considered in analyzing the effectiveness of updates.