# Site rotation: why it's a data problem By The Optimized Team (Product). Published: 2026-04-14T00:00:00+00:00. Updated: 2026-10-27T21:58:25.330749+00:00. Canonical: https://optimizedtracker.com/blog/injection-site-rotation-is-a-data-problem Rotation only works if you remember where you've been. Memory is bad at that — which is exactly why it belongs in a log, not in your head. Rotation is one of those things everyone knows they should do and almost nobody tracks properly. The intention is fine; the memory isn't. Ask yourself which site you used four entries ago and you'll get a shrug — and a shrug is how a plan to spread things out quietly turns into the same two spots over and over. The framing here is deliberately narrow: this is about *recording* where you've been, so a pattern is visible. It isn't guidance on technique — that's a conversation for a licensed professional. ## Why memory fails at rotation Rotation is a sequence problem, and human memory is famously bad at sequences of similar, low-salience events. Each entry feels identical to the last, so nothing sticks. Without a record you're not really rotating — you're guessing, and guesses drift toward habit. The spot that felt fine last time gets picked again, not by decision but by default. The value isn't the picture — it's that every past entry is written down and positioned. A [site rotation map](/features/site-rotation) turns "I think I used that side last time" into something you can actually look at. ## What a site log actually captures Treat the site as just another field on the entry, the same way you'd record a timestamp or a vial ID. Once it's structured data rather than a vague memory, a few useful things fall out for free: - **Recency** — when a given area was last used, at a glance. - **Frequency** — whether one region is quietly carrying more than its share. - **Coverage** — which areas you've been avoiding without meaning to. None of that requires you to remember anything. It requires you to record the site once, at log time, and let the history do the accounting. ### The pattern only shows up in aggregate A single entry tells you nothing about rotation — of course it doesn't. The signal is in the distribution across dozens of entries, and a distribution is precisely the kind of thing a person can't hold in their head but a map can show instantly. This is the same principle behind [turning check-ins into signal](/blog/turning-check-ins-into-signal): one data point is trivia, the shape across many is information. ## Structured beats freeform There's a real difference between jotting "left side again" in a note and selecting a site from a fixed set of positions. The freeform version can't be counted, sorted, or mapped — it's prose, and prose doesn't aggregate. Structured selection turns the same information into something a tool can actually reason about: it can tally each region, sort by recency, and render the distribution without you doing any arithmetic. The constraint of picking from defined positions isn't a limitation; it's what makes the history queryable instead of just readable. That structure also enforces consistency you'd otherwise have to sustain by hand. "Left delt," "left shoulder," and "l. arm" are three labels for one place in a freeform note, and they'll never line up in a count. A fixed set of positions collapses that ambiguity automatically, so every past entry compares cleanly against every other — the same consistency principle that makes any [protocol log auditable](/blog/building-a-protocol-system-you-maintain). ### It's a record, not a rule Worth being clear about the boundary. A site map records where you've been; it doesn't tell you where to go next or how anything should be done. Those are questions of technique and judgment that belong with a licensed professional. The map's entire job is to make your own history visible so that whatever pattern you and a professional decide on, you can actually see whether you're following it. ## Make the record the easy path The reason rotation lapses isn't that people don't care — it's that recording the site is one more small friction, and friction always wins over time. The fix is to make logging the site take a tap, not a decision, so the record simply accumulates as a side effect of normal logging. Do that, and rotation stops being a thing you have to *remember* and becomes a thing you can *see*. The map holds the sequence your memory won't, and the pattern — whichever way it's leaning — is finally visible instead of guessed at.