It’s inherently more difficult to predict the trajectory of keywords at the very lowest end of menu penetration (i.e. on the fewest number of menus). When an item moves from being on 2 menus to 4 menus, the absolute change is tiny, but the percentage “growth” looks enormous. With such a small sample size, it’s impossible to be perfectly accurate every time, and the data at that level often shows natural ups and downs that don’t necessarily signal a real trend shift.
This is exactly why we use multiple data sources to identify and validate trends. Social media, for example, is full of items that spike in conversation but never become true trends—think Dubai chocolate or cottage cheese. They have a viral moment, but they don’t materialize into sustained behavior, especially not on restaurant menus.
In contrast, arancine is a strong example of a real trend we’ve been watching. While it has seen occasional dips on menus over the last few years, the overall trajectory is upward. We see arancini as “one to watch” not just because of menu growth, but because it’s now appearing on influential menus—for instance, The Cheesecake Factory recently introduced it.
When we say 98% accuracy, we’re not claiming perfection—but we are saying we’re right most of the time. Our trend predictions don’t rely on Menucast alone. We layer in multiple proprietary datasets—Consumer Preferences, Social Listening, our historical menu data—alongside the expertise of our in-house analysts, who are trained to distinguish short-lived fads from durable trends. It’s this combination of data and expertise that allows Datassential to fine-tune and strengthen our trend forecasts.