> For the complete documentation index, see [llms.txt](https://docs.growthfactor.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.growthfactor.ai/data-and-analytics/model-explainability.md). # Model Explainability Model explainability shows how a sales forecast got to its number. Alongside the forecast, you see a plain-language description of the model and a benchmark for each variable it ran on. Each benchmark places this site against your existing stores and shows whether that variable pushed the forecast up or down. Explainability appears in two places: * The **Sales Forecast** card when you search a site. See [Sales Forecast](/data-and-analytics/sales-forecast.md). * The **Forecast** tool on a study pin. See [Pins](/study/pins.md). What you see depends on the model. Some models return benchmarks for every variable, some for a few, and some for none. If a model returns nothing, the forecast displays without these sections. *** ### About This Model **About This Model** sits under the model picker. It's a short description of what the selected model forecasts and what drives it, written for your workspace's model. Click **About This Model** to collapse or expand the description. It only appears for models that have one. To add or change the description for one of your models, reach out to your GrowthFactor team at . *** ### Variable Benchmarks **Variable Benchmarks** appears below the forecast result. It lists the variables the model used, one row each. Click the **Variable Benchmarks** heading to collapse or expand the list. Each row shows: * **The variable name**, such as a demographic or a site property you entered. * **This site's value** for that variable. * **A direction mark** showing how the variable moved the forecast. * **A distribution plot** of the same variable across your existing stores, with this site marked on it. * **A rank** placing this site among your existing stores. A variable only gets a row when the model has a value for this site. #### Direction The mark beside the value shows which way the variable moved this site's forecast, compared with a typical site. Hover the mark to see its label. | Mark | Meaning | | -------------- | ------------------------------------- | | Green up arrow | Raises the forecast vs a typical site | | Red down arrow | Lowers the forecast vs a typical site | | Grey dash | In line with a typical site | Some models only report direction for certain variables. A row without a mark still shows the distribution and rank. #### Numeric Variables Numeric variables, such as population or median income, display as a violin plot. The grey shape is the spread of that variable across your existing stores. It's widest where most of your stores fall and thins out toward the extremes. The dot and line mark this site. It takes the color of the direction mark, so a green marker means this value is helping the forecast. A hollow dot means this site falls beyond the edge of the plot, either lower or higher than nearly all your existing stores. The rank on the right is this site's percentile among your existing stores. **24th pct** means this site's value is higher than 24% of your stores. #### Categorical Variables Categorical variables, such as store format or drive-through availability, display as a bar chart. Each bar is one category, sized by how many of your existing stores fall in it. This site's category is the filled bar in the direction color. The percentage on the right is the share of your existing stores in this site's category. Hover a bar to see its category and share. When there are more categories than fit, scroll the chart sideways. A fade on either edge shows there are more bars in that direction. #### Row Details Click a row to expand a one-line summary. It combines the direction with the rank, for example: > Raises the forecast vs a typical site. Ranks in the 72nd percentile of your existing stores — above average. For a categorical variable, the summary gives the share of your stores in that category instead. *** ### Reading the Benchmarks Together The benchmarks explain the forecast one variable at a time. A few patterns to look for: **Variables at the edges.** When a forecast looks unusually high or low, start with the rows where this site ranks near the top or bottom of your stores. Those are the likeliest explanation. **Mixed directions.** A forecast close to your typical store can hide variables pulling in opposite directions. Check which rows are red before assuming the site is average. **Values beyond your portfolio.** A hollow marker means this site sits past nearly all of your existing stores on that variable. Your stores give the model little to compare against there, so treat the forecast with more caution. It's the same as when there are [few close analogs](/data-and-analytics/sales-forecast.md#interpreting-analog-matches). *** ### In PDF Exports When the sales forecast is included in a PDF export, **About This Model** and **Variable Benchmarks** come with it. This applies to both [Site Analysis exports](/site-analysis-legacy/sharing-and-exporting.md) and [study exports](/study/sharing-and-exporting-studies.md). In the PDF, each benchmark prints as a line with the variable name, this site's value, its rank or share, and the direction arrow. The distribution plots don't print. --- # 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.growthfactor.ai/data-and-analytics/model-explainability.md?ask=&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.