## DocumentationSkuBrain 101 |

# Replenishment policies

Replenishment policies specify how SkuBrain should calculate the amount of stock that you need to order. You can configure these policies in the settings for your forecasting project. In the simplest case you’ll have a single (universal) replenishment policy that is used by all of the products that you sell, but you can optionally enable ABC Analysis, which lets you specify different policies for different products.

In any case, a policy consists of 2 settings:

- A desired service level (the likelihood that you
*won’t*run out of stock) - An overstock warning threshold

## Service Levels

A service level, in inventory planning, is the **probability that you won’t run out of stock**. OK great – so how do you get these calculated?

To understand service intervals, take a look at one of your demand forecasts and turn on the advanced settings in the top right of your screen. In advanced settings, you’ll be able to see prediction intervals for your forecast.

These prediction intervals get generated when SkuBrain generates the quantitative forecasts for your products, and reflect the statistical likelihood that
demand for your products will fall within a certain range. So there is a 95% probability that sales will be below the the `High 95`

prediction interval
in any given period and an 80% chance that they will be below the `High 80`

forecast.

If you hold enough stock to cover demand forecasted by the Hi95 prediction interval then, essentially you’re holding enough stock to avoid 95% of stock-outs, so prediction intervals can be used to define service levels for the products you sell.

The **Final forecast** in SkuBrain is a mean forecast – so it represents the 50% prediction interval. If you specify a service level of 50% then, SkuBrain will
simply use the final forecasts when calculating how much stock you need to reorder. If you specify a service level of 95% then SkuBrain would take the final forecast, adjusted by the variance represented by the 95% prediction interval.

For example, if the mean forecast in a particular period was 5 units, the Hi-95 forecast was 9 units, the final forecast was 7 units and you specified a service
level of 95%, then SkuBrain would use the forecast `95% service level = final + hi95 - mean = 7 + 9 - 5 = 11 units`

.

Other than that, the rest of the calculation for how much stock you need remains exactly as described in the Order Cycle section above.

## Overstock Threshold

The final piece of the puzzle won’t actually affect how many units SkuBrain recommends you order. In fact, it’s quite the opposite – SkuBrain uses it to let you know when it thinks you’re carrying too much stock.

The overstock threshold is a percentage that is used to calculate a maximum stock level as follows:

```
Maximum = lead_time_demand + cycle_demand * overstock
```

If your stock on hand goes above this threshold then SkuBrain considers the product overstocked.

For example, if your lead time demand was 15 units, your cycle demand was 45 units and you had configured an overstock threshold of 200% of requirements, the stock maximum would be:

```
Maximum = 15 + 45 * 200 / 100
= 15 + 90
= 105
```

If you were holding more than 105 units when you ran the replenishment report then, SkuBrain would list this product in the overstock tab of the report, along with details on how many units you were holding in excess of requirements.