One Playbook? Not Here.

There isn’t one “right” lever. In the same region, the same lever can help one store and hurt another.

• Transactions vs. retention flips sign. Store 004 shows a slight positive link, while 001 shows a strong negative one.
• Sales volume often hurts retention. Five stores trend negative; 001 is the exception and even there it’s weak.
• Team performance always helps retention — but the strength of that help varies a lot (2× spread).
• UPT and retention fight each other: higher UPT correlates with lower retention across all six — balance matters.

Same region, different realities: scale, retention, and team dynamics don’t line up the same way everywhere.

• Diagnose first, decide second. Run store‑level models before setting targets.
• Protect loyalty while chasing sales. Track retention when pushing UPT & promos.
• Coach the team like a product. Team score moves retention everywhere — invest here.
• Match the cadence to φ. Sticky stores (φ≈0.85) need fewer, bigger moves; volatile ones (φ≈0.70) need lighter, more frequent nudges.

We used a Bayesian regression per store to see how each factor connects to retention. Instead of forcing one “regional truth,” we measured each store’s own pattern and compared the directions and sizes. Where patterns matched across stores (e.g., team score → better retention), we propose standardizing. Where they diverged (e.g., transactions ↔ retention), we argue for local control.

TL;DR: Use a hybrid strategy — standardize what’s consistently helpful, localize what varies or flips direction.