Background: Repetitive transcranial magnetic stimulation (rTMS) is an effective treatment for depression, yet few studies have mapped the trajectories of symptom change over treatment. Tracking clinical response during early treatment may be helpful to predict outcome, particularly non-response. Methods: We used naturalistic data (N = 117) to examine changes in the Daily Symptom Index (DI-5) scores of adult patients with unipolar or bipolar depression who underwent ≥16 treatment sessions of left dorsolateral prefrontal cortex rTMS at a private psychiatric facility in Western Australia, between 2016 and 2019. Results: Two response trajectories were charted: non-response (N = 71, 61%) and response (N = 46, 39%). Both trajectories diverged at 99% confidence interval at session 10, which was used as the point to predict treatment response at session 20. The response group showed a reduction of 4.21 in the mean DI-5 score from baseline at session 10. On this basis, a 4-point reduction in the DI-5 score at session 10 was defined as predictor of responder status at session 20. If the improvement is < 4 points at session 10, the probability of non-response at session 20 is 75%. If the improvement is ≥ 4 points, the probability of response at session 20 is 66%. Limitations: The DI-5 scores were not examined beyond 20 treatment sessions, which may have shown delayed responders in the non-response group. Conclusions: In this study of depression response trajectories with rTMS treatment, prediction of response at session 20 can be made at session 10 of treatment. Further research is required to generalise the current findings.