Simulating a social media campaign before launch: what OranSim promises and what it does not

What if you could test a social media campaign before spending the budget? That is the idea behind OranSim, a system presented on 23 September by a team of six researchers, with the code published. It is a preprint: it has not been peer reviewed yet, so it is best read as a serious proposal and not as a settled result.

What the simulator does

OranSim links four decisions every social media marketing team makes to what happens next:

• The creative: which piece is published.
• The creator: who publishes it.
• The targeting: who sees it first.
• The budget: how much is invested.

From there it simulates who sees the campaign, how they respond and how that response spreads across 60 population segments. To compare two scenarios it always starts from the same initial population, so the difference comes from the change you made and not from chance.

The figures

The models that predict engagement were trained on 39,000 posts from RedNote, the Chinese social network, and tested on a further 12,154. They explain between 56% and 62% of the variation in engagement (R² of 0.56 to 0.62, on a log scale).

The most striking experiment is the budget one. When it is doubled, reach nearly doubles and the cumulative 14-day response rises to 1.96 times the baseline. But the average fit between the content and the people who see it drops, and so does the probability that each person engages. The extra budget reaches people who care less about the piece.

Why we think it is a trend worth following

Much of the work with simulated audiences consists of asking what a type of person would think. This is a different step: simulating the whole journey of a social media campaign, from first exposure to spread.

How we read it at Trendia

With interest, and with three cautions.

It is for comparing, not for predicting. A model that explains 60% leaves 40% unexplained. Saying "scenario A performs better than B" is reasonable. Promising a reach figure is not.

The data comes from a single network. Nothing guarantees that what was learned on RedNote carries over as is to Instagram or TikTok.

The creative goes in as just another input. And it is the hardest part to turn into data. What format a video has, what emotion it conveys and what story it tells is exactly what we analyse at Trendia, on real videos and with human review. A simulator will only be as good as what it knows about the piece it simulates.

If you want to know what is working in your category today before the next campaign, tell us what you need to know.

Source: OranSim: Simulating Social Media Marketing, arXiv, 23 September 2026.