> For the complete documentation index, see [llms.txt](https://docs.buildings.ability.abb/collection/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/review-a-target.md).

# Review a B\&T Config

Review the selected baseline's target preview and model metrics before activating it or changing its settings. You need a [-> saved baseline](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/create-a-baseline.md) with historical source data.

1. [-> Open Benchmark and Targets](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/open-benchmark-and-targets.md).
2. Click **Configure target** for the required B\&T config.

The preview does not activate a baseline or create its generated attributes. Date-range actions such as **Apply** and **Save target** do save the range; they are not preview-only controls.

## Read the charts

<figure><img src="https://2013058394-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FOqGnkz52o2zVUsEGEZGk%2Fuploads%2Fgit-blob-a28ca5a0ffa588f235aa1a1895bfff5aad962d92%2Fbenchmark-targets-target-line-overview.png?alt=media" alt="Target line overview with a selected baseline and four populated charts for predicted versus actual values, difference, percentage difference, and CUSUM"><figcaption><p>Fig. 7: Target line overview for the selected baseline</p></figcaption></figure>

1. Select a saved baseline **\[A]** in the left pane.
2. Select **Target line overview** **\[B]**.\
   ▷ The four charts display the selected baseline's preview. The shaded region identifies its reference period.

<table><thead><tr><th width="183.99993896484375">Chart</th><th>Interpretation</th></tr></thead><tbody><tr><td><strong>Predicted vs actual</strong></td><td>Compare the fitted target with measured values and the highlighted baseline period.</td></tr><tr><td><strong>Difference</strong></td><td>Actual value minus target value. Positive values are above target; negative values are below it.</td></tr><tr><td><strong>Difference %</strong></td><td>Percentage deviation between moving averages of measured and target values. This is not the percentage difference for each individual observation.</td></tr><tr><td><strong>CUSUM</strong></td><td>Cumulative sum of the actual-minus-target differences. A rising curve adds positive differences; a falling curve adds negative differences.</td></tr></tbody></table>

### Predicted vs actual: check that the model makes sense

Start with the shaded reference period. Does the model broadly follow the measured pattern, including ordinary peaks, quiet periods, and seasonal changes? A model that misses a regular operating pattern may need a more representative reference period, a different timeframe, or a relevant feature attribute such as outdoor temperature.

Then inspect the period after the baseline. A persistent gap may indicate a change in operation rather than an isolated unusual day. Compare it with known interventions and operating conditions. The chart shows whether the baseline provides a credible comparison; it does not establish the cause of a difference.

With a configured percentage offset, the target is intentionally lower than the fitted operating model once that offset applies. Check the [-> offset settings](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/configure-a-baseline.md#offset) before interpreting such a gap as poor model fit.

### Difference: locate departures in the source unit

Look for intervals where measurements are consistently above or below the target. A single large difference helps locate an exceptional interval; repeated differences in the same direction suggest a persistent mismatch. For energy consumption, positive values indicate excess against the selected allowance, while negative values indicate use below it.

Use the absolute difference when the size of an excess in the source unit matters, for example when choosing an alarm threshold. Check the calculation timeframe as well: the numerical difference depends on the values being compared.

### Difference %: compare the size and direction of drift

The percentage comparison makes departures easier to judge relative to the target's size. It is calculated from smoothed actual and target values, so it emphasizes sustained behaviour rather than each individual fluctuation.

Look both at the level and the trend. A rising line means measurements are becoming higher relative to the target; a falling line means they are becoming lower. A rise from -20% to -5% is still below target but leaves less margin. A fall from +20% to +5% is improvement towards the target, but remains above it.

### CUSUM: distinguish persistent bias from individual events

A roughly horizontal curve means positive and negative differences are balancing over that part of the history. A sustained upward slope means actual values are consistently above the target; a sustained downward slope means they are consistently below it. A steeper slope means a larger imbalance is accumulating. A bend can help locate a change in performance.

If an upward curve levels out after an intervention, the new excess has stopped accumulating; the earlier excess has not automatically been recovered. The absolute height of CUSUM carries past differences, while its slope describes what is happening over the current part of the chart.

### Check the baseline before judging later performance

For a representative reference without an intentional offset, aim for differences that are broadly balanced, a percentage difference near zero without a persistent trend, and a reasonably flat CUSUM within the highlighted baseline. Perfectly flat curves are not required, but a strong slope already inside the baseline makes later drift harder to assess.

This does not mean that the original consumption or target line should be flat. Seasonal and operating variation can be legitimate. It is the unexplained difference that should not already show a strong systematic trend. If it does, investigate the reference period and [-> slope mode](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/configure-a-baseline.md#slope-mode), rather than selecting settings only to make the charts look better. An offset deliberately active during the reference period can also introduce a positive difference.

Hover over a chart to compare values at the same timestamp across the charts. Tooltips and crosshairs are synchronized. Drag horizontally across a chart to zoom into a shorter period; the other charts follow the same time range.

## Review the correlation period

<figure><img src="https://2013058394-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FOqGnkz52o2zVUsEGEZGk%2Fuploads%2Fgit-blob-18753fb6deb4cec399dc141068c9239060c5538c%2Fbenchmark-targets-correlation-period.png?alt=media" alt="Correlation period chart showing source history, colored baseline segments, and the selected fixed period&#x27;s boundary handles"><figcaption><p>Fig. 8: Source history and selectable correlation periods</p></figcaption></figure>

1. Select **Correlation period** **\[A]**.\
   ▷ The source attribute's history appears with colored bands for the saved reference periods. The selected period is emphasized.
2. Click a segment label, such as **Seg 1** **\[B]**.\
   ▷ The corresponding baseline is selected, and its date values appear in the left pane.

Drag horizontally across the plot area to zoom into the history without changing the reference dates. To change a fixed reference period:

1. Select **Set Date**.
2. Drag either boundary handle of the selected range **\[C]**.\
   ▷ **From** and **To** update to the draft range. Handles snap to days within the available history and keep the range at least one day long.
3. Click **Save target** **\[D]**.\
   ▷ The range is saved and date-range editing ends.

**Apply** also saves the dates but keeps range editing open. **Reset** restores the most recently saved dates, including dates already saved with **Apply**. Changing a handle alone does not save the range. See [-> Create a baseline](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/create-a-baseline.md) for fixed and moving baseline procedures.

## Read the metrics

<figure><img src="https://2013058394-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FOqGnkz52o2zVUsEGEZGk%2Fuploads%2Fgit-blob-93b286d9a8a8365bfa82c18534c605808c585b6b%2Fbenchmark-targets-metrics.png?alt=media" alt="Metrics tab showing the selected time period, R squared, training count, and error metrics"><figcaption><p>Fig. 5: Model metrics for the selected baseline</p></figcaption></figure>

1. Select the baseline to evaluate in the left pane.
2. Select **Metrics**.\
   ▷ The available model metrics are displayed for the selected reference period.

Check **Time period** **\[A]** before comparing metrics. **R²** is marked **\[B]**; the error metrics are grouped at **\[C]**.

<table><thead><tr><th width="121.5999755859375">Metric</th><th>Meaning</th></tr></thead><tbody><tr><td><strong>R²</strong></td><td>Coefficient of determination. Values closer to 1 indicate stronger agreement with the measured observations. Do not assess model quality from this value alone.</td></tr><tr><td><strong>N (train)</strong></td><td>Number of observations used for model training.</td></tr><tr><td><strong>NMBE</strong></td><td>Normalized mean bias error, displayed as a percentage. Values closer to zero indicate less average bias.</td></tr><tr><td><strong>CVRMSE</strong></td><td>Coefficient of variation of the root mean square error, displayed as a percentage. Smaller values indicate less error relative to the mean measured value.</td></tr><tr><td><strong>RMSE</strong></td><td>Root mean square error, in the units of the evaluated attribute values. Larger errors receive more weight.</td></tr><tr><td><strong>MAE</strong></td><td>Mean absolute error, in the units of the evaluated attribute values.</td></tr></tbody></table>

{% hint style="info" %}
Model evaluation withholds a random 5% of the baseline observations, rounded up with at least one observation, from training. Metrics describe agreement on these withheld observations within the reference period, not accuracy on a future period. Undefined or unavailable metrics are omitted.
{% endhint %}

If the preview or metrics are unavailable, confirm that a baseline is selected and its source and feature history is available. Review any displayed loading or calculation error before changing the [-> baseline settings](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/configure-a-baseline.md).

When the preview is suitable, [-> activate the baseline](https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/activate-or-pause-a-baseline.md).


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.buildings.ability.abb/collection/platform/platform-core/baseline-and-factors/portfolio-benchmark-and-targets/review-a-target.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.
