Crystal is an Amazon-approved application that produces three separate forecasts to help you evaluate how much inventory to purchase. These three forecasts (P90, P70, & P50) create a range of possible demand values. This range accounts for uncertainty in demand.

Quantiles can provide an upper and lower bound for forecasts. For example, using the forecast types 0.5 (P50) and 0.9 (P90) provides a range of values called a confidence interval. The observed value is expected to be lower than the P50 value 50% of the time, and the P90 value is expected to be higher than the observed value 90% of the time.

P50:
A forecast at the 0.50 quantile means actual demand will be higher than the forecast half of the time. This is the lower bound of the forecast. Over time, buying inventory to this number likely causes you to fall out of stock as you will not hold enough safety stock.

P70:
A forecast at the 0.70 quantile means actual demand will be higher than the forecast 30% of the time. This is the recommended point forecast. Over time, buying inventory to this number will balance uncertainty with being overstocked.

P90:
A forecast at the 0.90 quantile means actual demand will be higher than the forecast only 10% of the time. This is the upper bound of the forecast. Buying inventory to this number likely causes you to become overstocked as you must hold much more safety stock to cover all demand probabilities.

Our rule of thumb is to treat the P70 forecast as “normal” demand in normal times. Remember, the P70 is the probability that actual demand will be lower than this number 70% of the time at the time the forecast was run. Consumer behavior changes frequently, so we recommend you refresh your forecast regularly to ensure ongoing accuracy.

At Forum, we buy digitally-native consumer brands and build them into category leaders. We partner with owners of the best everyday products selling through Amazon FBA and other 3rd party marketplaces. As long-time investors, operators, and industry experts, we represent the most trusted buyers in e-commerce.

We keep up to two years of sales data, and merchant identification data.