Knowledge of a person’s whereabouts in the home is key to context-aware applications, but many people do not want to carry or wear a tag or mobile device in the home. Therefore, many tracking systems are now using so-called weak biometrics such as height, weight, and width. In this paper, we propose to use body shape as a weak biometric, differentiating people based on features such as head size, shoulder size, or torso size. The basic idea is to scan the body with a radar sensor and to compute the reflection profile: the amount of energy that reflects back from each part of the body. Many people have different body shapes even if they have the same height, weight, or width, which makes body shape a stronger biometric. We built a proof-of-concept system called FormaTrack to test this approach, and evaluate using eight participants of varying height and weight. We collected over 2800 observations while capturing a wide range of factors such as clothing, hats, shoes, and backpacks. Results show that FormaTrack can achieve a precision, recall, direction and identity accuracy (over all possible groups of 2 people) of 100%, 99.86%, 99.7% and 95.3% respectively. Results indicate that FormaTrack can achieve over 99% tracking accuracy with 2 people in a home with 5 or more rooms.