Should we really be optimizing for saves on social media? 

One notion I have heard frequently is that saves is a good indicator of measuring how well social media content performs. 

Many social media consultants and experts mention saves and shares as the main metrics Instagram and Tiktok use in tracking engagement and deciding which posts to push out. While I agree with shares being an accurate engagement representation of engagement, let us explore why saved might not be the best metric for this.  

What Makes A Good Metric?

A good metric is representative of the expected behavior. There is a saying that goes, “you cannot improve what you don’t measure”. So the first step to improving anything you care about is to track it. That is what metrics are for.

While it is so easy to want to create metrics for everything (this is a mistake many people actually make), you actually should carefully create metrics that mirror ideal behaviors you would like to optimize.

The most common metrics for products can be grouped under usage metrics, engagement metrics, and retention metrics. These are the foundational metric groups. 

Usage metrics track the volume, frequency, and overall duration of interactions and customers on a product e.g. daily users, while engagement metrics track the depth and quality of said interactions. Retention metrics, on the other hand, track the return of users to either the entire product or on a specific feature. 

What Is engagement?

Typically, engagement is measured by how much time people spend on the product, number of interactions with features/areas on the product, feedback given, or the value derived from interacting with the product. For a social media platform, engagement can be measured by either time spent on the app, count of posts (posted and viewed), and count of interactions that incentivize users to spend more time on the app and view more posts.

Engagement differs from usage metrics because it not only tracks how many people came to the product. It measures the depth of the users’ interactions and how involved they were on the product. Engagement metrics are best suited to track loyalty and feature adoption.

Examples of some common engagement metrics for social media platforms are:

  1. Active time spent
  2. Views
  3. Likes
  4. Dislikes
  5. Comments
  6. Shares
  7. Saves (newer)

For this article, we’ll be exploring the last one, Saves, and I’ll be discussing why I don’t think it’s the best metric for signaling engagement on social media platforms. 

Why Is Saved Not The Best Engagement Metric?

Save in itself might not be the right engagement metric because how likely is it that people save a post and actually return to it ever? 

Using Instagram as an example, how many times have you been scrolling distractedly and then absentmindedly saved a post to come back to it later when you can view it properly? And then you never came back to that post. 

Is this a good signal for engagement? In this case, can we still count you as an engaged user?

If you’re looking to increase engagement, you should be choosing metrics that track people interacting intentionally with features on the app, in this case, the content posted. Saving a post and never checking the saves does not actually represent that. 

How Could The Save Metric Be Improved To Track Engagement Correctly?

Instead, I propose a new metric called revisited saves. 

What do I mean by revisited saves? 

This is the calculation of posts that were saved and returned to in a given time period. The period for returning to the saved post can be set to 7 days, 30 days or even infinity. This way, we’re not just measuring posts that got saved (and possibly never returned to), but specifically posts that got saved and caused the user to return to it for some reason. 

Although this might be a more complex metric to calculate, I believe it is a more efficient representation of the behavior we want to model. 

The challenge with implementing this metric is the added post-computation time and complexity needed to derive the metric. When building your data pipeline, on a user-level, you can track 

  • the number of posts viewed, 
  • the number of posts saved, 
  • the number of revisited saved posts, and, 
  • revisited saves/unique users in a given period.

However, on a post-level, this metric needs to be computed offline and added to the downstream, or updated periodically for each saved post that was returned to. 

What Is The Takeaway Here?

  1. When defining metrics, it is important to align it as closely as possible to the best case scenario and ideal behavior. 
  2. Instead of implementing saves blindly as an engagement metric, consider adding a caveat specifically for saved posts users returned to. This will help filter out unintentional saves and reduce noisy signals. 
  3. Copying off-the-shelf metrics without mapping it to your specific product use-case could give misleading insights. 

If you enjoyed reading this piece, subscribe to my Substack to receive a notification when I publish more articles like this, unpacking different metrics for businesses and products, sharing guidance for early data teams to setup metrics for their products and derive actionable insights from them, as well as, wearing the DS hat in the AI and agentic world today. 

I look forward to reading your thoughts, comments, suggestions, and questions below. 

Thank you for reading. 

Aniekan

To not miss out on any new articles, consider subscribing.

Subscribe