Understanding social interactions and possible data issues
Parse.ly provides social shares data in its dashboard and API to help understand social engagement readers have with content. The way that we’re able to capture this data is by taking the Parse.ly canonical URL for an article and then pinging each social network’s API with that URL. The result is that we have the total number of shares returned from each social network.
Parse.ly does additional work in storing that total each time it pings the social network APIs. We store the total in order to understand how many shares occurred since the last time we pinged the API. The delta in this case is the number of shares that happened over that period. As an example, Post A was published yesterday. We pinged the Facebook API and it returned 300 shares for that article. Today, we ping the API and it returns 375 shares. We know that 75 new shares happened between the time that we pinged the social network API (since yesterday). We then store this data so you can see how many shares occurred on each individual day in the history of an article.
How often do we ping the social network API endpoints? The answer is it depends. It is mostly dependent on traffic to a post, particularly from social networks (aka social referrers). This could be every minute for high-traffic articles or it could be only once a day for articles that receive less traffic.
Unfortunately, the social network APIs do not always provide consistent data via their APIs. Due to how many URLs and shares they are tracking, there is some fluctuation in the data that we receive. This is only noticed if Parse.ly displays a negative share count for a particular day. You may be thinking “did that many people actually unlike this article?” and most of the time the answer is “no.”
So how is it that negative shares are possible? Let’s use Post A as an example again. Post A was published on July 1, 2015. Below are the first five days of share data that we receive from the Facebook API as well as the deltas from one day to the next.
| Date | Shares | Delta |
|---|---|---|
| 07/01 | 300 | – |
| 07/02 | 375 | 75 |
| 07/03 | 345 | -30 |
| 07/04 | 390 | 45 |
| 07/05 | 410 | 20 |
If we were to query Post A from 07/01 – 07/05, the Parse.ly dashboard would show 410 shares from Facebook as this is the total (to date). However, if we were to query an individual day, Parse.ly will return the delta. This could result in days like 07/03 where there are -30 shares. While you may assume this isn’t accurate, this data isn’t necessarily incorrect as Facebook is the only source of truth in the matter. Facebook is simply updating the share count on a regular basis and these variances can cause negative deltas like in this example. While this example was from one day to the next, this same could apply from one hour to the next or even from one minute to the next (if you see negative shares on the first day this is more likely to be the case).
If you feel that the negative share number is an error, please feel free to reach out to support@parsely.com as we’re happy to take a deeper look.
Last updated: August 16, 2023