What is the Foresight Engine? Predictive Creative Performance Explained

Overview

From the sibyls of Ancient Greece to the horse racing bookies of today, humanity has long sought to unveil the future before its time. Now, advancements in technology have given us a sliver of this awesome power: we can accurately predict engagement metrics on social media posts. With our state of the art prediction model, Foresight Engine, next-gen computing is leveraged to provide you with accurate engagement predictions for your brand’s next social media post.

Takeaways

• Predictive creative performance uses AI to measure efficacy of your creative assets.
• Foresight Engine trains on huge social media data sets in order to provide predictions based on genuine engagement trends.
• Inside the Foresight Engine, several advanced computational models work in tandem to generate accurate predictive figures.

What is Predictive Creative Performance?

Predictive creative performance is a branch of marketing technology that, usually using AI, seeks to be able to quantify or qualify your creative assets before launch. In effect, this practice helps diminish guesswork surrounding your campaigns before they’re taking up your customers’ valuable screen real estate.

Foresight Engine is Soulmates.ai’s predictive creative performance model. It takes your drafted social media posts and analyzes them to synthesize statistic-based predictions in the form of projected engagement metrics.

But how does it work?

Big Data

Just as Big TV is conspiring to keep me addicted to Traitors (Season 4 out now), Big Data conspires behind the scenes of Foresight Engine.

Foresight Engine’s insights are powered by a detailed analysis of over 600,000 posts from over 6,000 brand channels across Youtube, Tiktok, and Instagram — and those numbers continue to grow. A larger dataset, among other mathematical factors, results in more accurate predictions from Foresight Engine.

As the data is analyzed, our models are able to not just how a post performed, but are able to synthesize the why that is used to determine the efficacy of your own post.

Post Examination

Foresight Engine’s frontend kicks into gear when you feed it a sample post and caption. At this stage, an image to text model gets to work deciphering the visual component of your post. Images are analyzed and synthesized into plain text descriptions, which are then converted into computer-speak for the models to understand.

At the same time, your caption is analyzed by a deep language model, which defines your copy against an expansive series of benchmarks (tone, word count, emoji usage, etc.).

Foresight Engine’s comprehensive examination of post copy and visual give it a detailed, nuanced understanding of your post as a whole. Foresight Engine’s draft post analysis is not a cursory glance – but a deep and thorough reading.

Enter the Neural Network

After initial post analysis, Foresight Engine’s Neural Network comes into play. Just as your draft-post was analyzed, the 600,000 posts in the data training set have been analyzed by a neural network, and algorithms surrounding post efficacy have been developed.

The neural network is split by two towers: one which analyzes your post, and one which analyzes your account. Your post might be the best post Instagram has ever seen (unsure what this could be, maybe a photo of Lady Gaga in a weird dress at some gala or a video of a puppy who learned sign language), but that doesn’t matter if you have limited reach or your followers aren’t Gaga-heads.

The two towers of the neural network combine their findings into one unified prediction, which the Engine then delivers to you.

Inside Foresight Engine

Perhaps it's disingenuous to refer to Foresight Engine’s findings as one unified prediction, as actually, it’s several. The Foresight Engine analyzes and reports on different metrics depending on the platform your post was designed for: likes, comments, saves, etc.

Beyond that, each prediction metric that Foresight Engine gives you is actually two predictions. The Engine performs a prediction for lowest number of likes the post could get (for example), then a prediction for highest number of likes your post could get. A range of expected likes (or comments, etc.) is prepared for you based on these predictions.

With these predictions of post engagement metrics, Foresight Engine also calculates a “score” for your post’s engagement, based on your estimated number of engagements measured against average engagements for your channel. This score feature provides a simple way to check your post’s engagement metrics against channel benchmarks.

Beyond quantitative data, Foresight Engine also provides insights about why your post performed the way that it did. From its analyses of your channel, Foresight Engine synthesizes analyses about what causes one of your posts to be successful (this is what allows it to predict engagement metrics) as opposed to simply providing predictive data, Foresight Engine also provides actionable insights as to how your post could better maximize engagement.

Foresight Engine as Predictive Creative Performance

Sometimes, as exciting as AI is, it can be a scary buzzword. In some eras, in some cultures, Foresight Engine’s divinatory powers might’ve gotten it into serious trouble. In order to alleviate your potential AI related fears, it’s perhaps helpful to just think of AI as a way to sort through (and synthesize around) large data sets.

Foresight Engine is the perfect example of this; it’s Foresight Engine’s detailed analysis of huge numbers of social media posts that allow it to accurately synthesize predictions on your posts.

In effect, putting aside the complex mathematics and computation, all that predictive creative performance models do is analyze your posts against historic data sets. What makes Foresight Engine unique in this space is its use of large scale social media data.

Many predictive creative performance models predict the efficacy of your assets based on nebulous scores. Sure, the models may have been trained on “good” vs “bad” creative assets, but how does one determine what makes a creative asset “good” in the first place, without hard data to back it up? Social media engagement metrics are the perfect, hard data with which to measure performance, allowing you to receive quantitative scores that are not based in the qualitative realm of “good” or “bad”.

If your assets are designed with social media in mind, why not compare them against social media data?

And if you’re going to put in the effort and money to draft social media posts for your brand, why not go the extra mile and ensure they’ll actually resonate with your audience?

FAQ

Why does the Foresight Engine provide its engagement predictions in a range?

Social media engagement metrics are, by definition, nebulous. Specific platform algorithms stabilize engagement at different time lengths, and even though a post may have reached that time-stability threshold, there’s nothing to prevent someone from engaging with a post. Short answer: it’s impossible to provide predictions in anything other than a range.

How accurate is the Foresight Engine?

Foresight engine’s prediction range has a 20% accuracy buffer on either side. What this means is there’s a 20% chance that the engagement will be lower than predicted, and a 20% chance that the engagement will be higher than predicted.

Is the Foresight Engine training on my data?

No, the Foresight Engine trains on social media data from brand channels, not personal accounts.