By Decree of Gabe Newbot: Measuring the Cost of a Review Prompt

Let's use another hypothesis: Steam uses some feedback mechanism, so part of the growth can be explained just from knowing the previous dynamics. In our internal tests across all games, this hypothesis holds true with R^2=0.86. Basically, there is a strong correlation between how many reviews you get and how many you already have. Simply put - if we remove those "extra" reviews from the spikes, it would not only reduce the total by the subtracted value, but also lead to less growth later, resulting in a much more dramatic change. The model isn't perfect and we are still fine-tuning it, but it's valuable for getting an idea of the magnitude of the effect. Also, most publicly available sales/revenue estimates have a strong correlation between reviews and sales, sometimes even embedded into the estimator, but the effect this specific anomaly had on it remains unclear. Conclusion : the estimated 'clean' review count, without the anomaly playing a part, is in the range between 17k and 26k, compared to the real 33k the game eventually got. That's roughly 50% to 80% of reality.