Social Media Posting Lifecycle
It is expected for a post to have a lifecycle (in a similar sense to a half-life) starting at the point of publication until it becomes irrelevant or stale and is no longer served to an audience by the network. This lifecycle can be mapped by looking at a cumulative performance metric over time, such as impressions or engagement, where this metric will increase until it reaches a maximum after a sufficiently long period of time. By considering an aggregation of posts, the average or expected lifecycle of a post can then be defined in terms of the time since publication.
Analysis Methodology
As a first step, it is necessary to isolate the change in the performance metric as the feature for direct investigation, as the change in value is more important than the actual values. This will also account for controlling other features, such as the audience size, industry, growth, and other idiosyncratic factors related to the array of accounts. To do this, the performance metric of a post at each recorded point in time is divided by the maximum performance metric for that post (at the last point in time). Any outliers are not a concern, as they are expected to be minimal relative to the amount of data available, while the data discussed also spans over 3 years to mitigate any transient or algorithmic complications, such that the results should be objective regardless of momentary conditions.
LinkedIn Accumulated Results
Using unique impressions as the performance metric, the cumulative results show the progression over time and can be plotted together for all of the posts in the data. As mentioned, the average at each point in time can then be calculated, as the expected value at that point in time. A curve can then be fitted to these values for a smoothed relationship - after various iterations, a sigmoidal function was chosen to best represent the shape (specifically, a curve following Michaelis-Menten Kinetics, which is commonly used when saturation effects occur in the data being modelled - characteristic of a response which approaches a maximum as the independent variable increases). It should be noted that, in this case, the curve is forced through the origin and, since an artificial limit was used as the cut-off to define the maximum (more accurately described as the reference baseline up until that point), it is technically possible for the values to increase above 1 with additional time (nothing actually restricting impressions from increasing further), so this is not forcibly set as the asymptote to which the function approaches.
Interestingly, this shows that, for the average post, the majority of impressions are received within 5 days of being published, while the post reaches at least 90% of impressions at around 10 days after which it becomes stale. In a sense, this can be viewed as a form of half-life, where the progression in the days since publication shows an exponential decay in the rate at which impressions increase (although it may be more accurate to plot unity minus the cumulative impressions fraction).
Post Types Breakdown
It is possible to perform the same analysis with the data segmented by post type. The considered post types include video, image, article, document, poll, repost, and other (including events, jobs, etc). Although the differences are minimal, the results indicate that polls have the shortest half-life (which is understandable if the poll has a set deadline), while videos, images, and articles have a similar lifecycle (possibly grouped together as media) with documents, reposts, and other having the most extended lifecycles.
Other Network Results
Expanding the investigation, a similar analysis can be performed for Facebook, Instagram, and Twitter. However, in most cases, the amount of data available is more limited, so it is understandable to expect a greater dispersion in results. Similar results are seen, where each of the networks appear to have a shorter lifecycle than LinkedIn in a decreasing order of Facebook, Instagram, then Twitter. This is somewhat understandable if the chronological nature of the networks are considered, as it is common to see older posts from connections on LinkedIn and, to a degree, Facebook, but Instagram and Twitter tend to be a lot more fast-paced unless a post is an outlier with prolonged sharing.
Summarized Conclusions
...