HER2+ early breast cancer is a heterogeneous disease, comprising all the intrinsic breast cancer subtypes. The only biomarker available nowadays for anti-HER2 treatment selection is HER2 status itself, but estrogen receptor (ER) status is emerging as a robust predictive marker within HER2+ disease. In this Perspective, we discuss the biological and clinical differences between patients with HER2+/ER positive (ER+) disease versus those with HER2+/ER negative (ER-neg) tumors, namely short- and long-term (>5 years after diagnosis) prognosis, response to neoadjuvant treatment and benefit from adjuvant anti-HER2 targeted therapies. We also address other possible biomarkers to be used for patient selection in future clinical trials, like gene signatures, PAM50 subtypes, tumor-infiltrating lymphocytes, PIK3CA mutations, and changes in Ki67 score during treatment and discuss their limitations. Finally, we suggest new clinical trial designs that can have an impact on clinical practice, aiming to test treatment de-escalation separately for patients with HER2+/ER+ and HER2+/ER-neg tumors. We also propose an integrated classification of HER2+ disease, comprising DNA, RNA, protein expression and microenvironment characteristics, in order to identify those tumors that are truly "HER2-addicted" and may benefit the most from anti-HER2 treatment.