# # Know your capacity

**BestDay's data-driven capacity estimation** gives you an **accurate understanding of the current maximum achievable production**. Use this as a capacity baseline to **compare actual production** and **identify optimization potentials**. BestDay evaluates current constraints to ensure a valid capacity baseline.

The BestDay maximum production capacity is calculated using the input of **historical actual production**, **experienced deferments**, and **production constraints**, such as regulatory limits for the concentration of oil in disposed water or the power consumption per produced barrel.

The historical **evaluation window span**, **production constraints**, and **outlier threshold** is defined per client in the BestDay configuration.

**In this article:**

## # View BestDay capacity

You'll find the estimated BestDay capacity on the **Deep dive** and **Well summary** pages. On the **BestDay Inspector** side panel, you'll find the capacity calculation details.

## # The BestDay data science model

The BestDay capacity estimation is based on the BestDay data science model (BDM). The purpose of the BDM is to estimate a Best Day capacity target using linear extrapolation of **recent historical production** and **deferment data** for any given asset.

### # Tuning parameters

These are the tuning parameters with the strongest influence on the BestDay model performance.

- The
**Local Evaluation window parameter**defines the rolling time range where the regression analysis is carried out. The default setting is 90 days but can be changed to a different time range depending on the desired fit between the BestDay capacity curve and historical production data. A shorter time range results in a conservative BestDay capacity curve that fits the historical data more, while a long time range yields a smoother, less conservative curve.

The

**Statistical Outlier Detection parameter**discards days violating defined statistical metrics by considering these as outliers. The default threshold is three standard deviations from the distributed average (median), but you can also use percentiles of highest/lowest data points. Values above/below this threshold are not allowed to influence the prediction of tomorrow’s Best Day. Prior to detecting outliers, the capacity time series are detrended linearly to minimize bias towards older data points.**Production constraints**are defined to ensure the BestDay capacity is not violating safe, reliable, and high-quality production. You can set these constraints according to your production hub. Dates where production constraints are violated are disqualified from the regression calculations.

### # Linear regression model

- The trained
**Linear regression model/analysis**runs once per day for a selected time range according to the**Local evaluation window**. By default,**Weighted Least Squares**(WLS) is used to fit the linear regression model such that recent production has higher importance. Optionally,**Ordinary Least Squares**(OLS) can be used to give equal weighting across the Local evaluation window. Then, the BestDay model re-baselines the regression curve by shifting the curve to intercept the two days with the highest production in the time range. This re-baselining ensures that the predicted BestDay capacity follows the expected decline in production from the standpoint of the highest performing production days.