I'm sure there are models and they have low and high estimates. But how to do they decide on the percentage growth? A bit of art + science?
Interesting question - Definitely one that is a blend of art and science and doesn't lend itself to a tidy formula however, I think I can give you an idea of a few broad methods and concepts. Whether a firm's sales lend themselves to more quantitative vs qualitative type of models largely depends on their individual characteristics and the particular stage they are in their industry's life cycle as well as unique factors like M&A recent/future activity:
From an analysts perspective and wanting to work from a more macro point of view - a firm in a growth oriented industry a not a ton in the way of M&A activity, one approach is to forecast an expectation of total industry sales and form an expectation of market share to apply
For a firm that sells products that operate on a regular-ish life cycle (think new versions of Ipads, new Windows OS ext...) they might try and forecast organic unit growth based off of customer back log orders, try to get a sense of new product launch demand and how expected cannibalization will impact existing product demand, divide by (1 - anticipated gross margin) and apply an expectation for price inflation) Basically, organic unit growth x pricing power.
For larger, stable firms their growth might be well approximated with regression based models - for instance an AR(1) log linear model might give reliable predictions.
There are some companies that don't require that much "art" when forecasting their earnings.
One example is regulated utilities - they earn the amount that the regulator allows them to earn with not much volatility.
Another more volatile example is pure play mining companies (like Antofagasta) - they have two main moving parts - mining volume and commodity prices. Mining volume is relatively easy to forecast and is usually guided by the company. Commodity prices - if you don't have a view you can just take current forward prices in the market. Everything else in the forecast is mostly simple arithmetics.