What-If Tool review
A free, open-source tool for visual model explanations, counterfactuals, and fairness checks.
Reviewed by iTechGuides Editors · Editorial team · Updated Oct 2026
What-If Tool (WIT) is an open-source visual interface for investigating trained machine-learning models with minimal coding. It is suited to notebook users who want to probe classification or regression behavior, compare models, and examine how results vary across datasets and subgroups. Users can edit datapoints to test hypothetical changes, view partial-dependence and datapoint visualizations, inspect feature attributions, search for counterfactual examples, and review fairness metrics. Its scope also covers tabular, image, and text data, making it useful for exploratory analysis across different kinds of inputs.
WIT’s main appeal is the range of questions it supports in one visual workflow. Attribution views include SHAP, LIME, integrated gradients, and SmoothGrad through custom functions; counterfactual search and subgroup performance analysis complement those views. Model comparison can be performed across datasets and subsets. The tool connects with Jupyter, Google Colaboratory, Google Cloud AI Platform or Vertex AI notebooks, and TensorBoard, with TFMA Fairness Indicators also among its integrations. This makes it a natural fit for teams already working in those environments, rather than a standalone explanation service.
The project is distributed under the Apache License 2.0, with no-cost access and no paid tiers described. Model support includes TensorFlow Estimators, TensorFlow Serving models, Cloud AI Platform models, and arbitrary Python models supplied through custom prediction functions. That flexibility has a boundary: users outside those model paths may need to adapt their prediction workflow. WIT also has no export feature for audit-report generation, so teams that need packaged audit reports should choose a different tool. Pick WIT for visual investigation and fairness exploration in notebooks; look elsewhere when report export or broader out-of-the-box model support is central.
What-If Tool pros and cons
- Where it wins
- Edit datapoints and test hypothetical changes interactively
- Explore attributions, counterfactuals, and subgroup fairness metrics
- Integrates with Jupyter, Colab, Cloud AI notebooks, and TensorBoard
- Where it doesn't
- Listed model support centers on TensorFlow and custom prediction functions
- No export feature for generating audit reports
- Requires a notebook, TensorBoard, or supported model environment
What-If Tool fact sheet, pricing and score →
Advertiser disclosure: iTechGuides is reader-supported. Vendors can pay for top positions in our rankings and for a place on other products' pages, and we may earn a commission when you click some links. How we rank.
Last updated · How we research and update