Data Science is threatened by a looming credibility crisis: too many scientific results are not reproducible. Unfortunately, data scientists have accidentally contributed to the problem. We made science look like math, implying that one can prove scientific results (p < 0.05) without reproducing them. We need to adopt a new standard of reproducibility, one that encompasses the data, code, and decisions that underly scientific work. This change will be a windfall to commercial data scientists because code-based reproducibility is repeatable, automatable, parameterizable, and schedulable.