In industrial environments with a high density of end devices, a flexible and low-latency transmission scheme is necessary. NOMA is one of the key candidates to serve multiple end devices with a limited amount of resources. To enable successive user decoding in NOMA, careful power allocation is required. Typically, either user-fairness or good SNR performance for a single user is guaranteed. In this paper, we combine a NOMA scheme with constellation shaping to relax the SNR requirements and therefore ease the requirements on power allocation schemes.