Background: There is very limited data available on how most breast cancer recurrences, either distant metastases or locoregional recurrences (LRR), are actually discovered in routine clinical practice. Patients and methods: From a prospective cohort of 621 women diagnosed and treated for early invasive breast cancer between 2003 and 2013, we analysed the patients who were later diagnosed with distant metastases (n = 61) and the patients who had locoregional recurrences (LRR; n = 34). The patients had routine control visits for up to 10 years from initial diagnosis, with annual clinical visits, mammography, blood count, plasma creatinine and liver function tests. Results: Most distant metastases (n = 38, 62%) were found when a patient contacted health care services because of a symptom; only ten (16%) were detected at pre-planned control visits. The most common first sign or symptom of metastasis was pain (n = 23, 38%). Pain as the first indicator of metastasis indicated a lower survival in metastatic disease (hazard ratio 4.40; 95% confidence interval 1.77–10.94; p = 0.001). How relapse was detected or whether patient was symptomatic did not affect overall survival (OS) of patients with distant metastases. LRRs were mostly found at pre-planned control visits (n = 14, 41%). Abnormalities in routine laboratory tests did not lead to any detection of recurrence. Discussion: In this prospective, contemporary, real-world study, the vast majority of both distant metastases and LRRs were detected outside the pre-planned control visits. These results highlight the importance of finding ways to lower the threshold for contacting the surveillance unit, rather than frequent routine controls.