Most TRT tracking efforts fail within two weeks. Not because tracking is hard, but because the structure people use doesn't hold up.
A notes app full of dated entries that say "felt tired today" is not a log. It's a diary. You can't analyze it. You can't find patterns in it. You can't compare this week's energy to four weeks ago with any precision. It records that something happened. It doesn't build data you can use.
The consistency problem
Logs become useless at the first significant gap. Log every day for two weeks, skip four days, come back and fill in from memory, skip another week during travel - and you haven't built a longitudinal dataset. You've built a partial record with gaps and unreliable fill-ins scattered through it.
The gaps are invisible when you're analyzing the data later. You see a stretch of low scores and a stretch of better scores and assume the pattern is real. But some of those entries were made on the day. Some were reconstructed from memory days later. You can't tell which, and it matters, because retrospective entries reflect your mood at the time of filling, not your mood on the day in question.
For a pattern to be meaningful, the data has to be consistent. Missing 30% of days and filling them in retrospectively is not consistent.
The note-taking format
Free text doesn't scale. "Felt decent, a bit tired after lunch, sleep was okay but woke up at 3am" tells you something about that day. It tells you nothing you can use when you're looking back at 60 days of entries trying to understand why your energy declined in month two.
You need a number. Not because the number captures everything, but because numbers can be compared, trended, and correlated. A 1-10 energy score logged daily is worth far more for analysis than 60 days of prose that can't be aggregated. Both take roughly the same amount of time to enter. One is analyzable.
Logging the dose but not the state
Some people track their injections carefully - dose, day, sometimes lot number - and consider that a protocol log. It's half of one.
The dose log tells you what the inputs were. Without daily state scores alongside it, you have no way to connect the inputs to the outputs. You know what you injected on which days. You don't know what happened to your energy, mood, and recovery in the weeks after a dose change. The log is a calendar, not a dataset.
No separation between events and state
This is the structural failure that makes logs hard to analyze even when they're otherwise consistent. When a hard training session, a stressful day, a poor night's sleep, and your baseline daily state all get collapsed into a single entry, the data loses analytical value.
You end up with low scores you can't attribute to anything, because the context that would explain them is baked into the score rather than logged separately. Keeping events and daily state distinct is what makes correlation possible.
The too-detailed trap
The opposite failure is logging so much that it becomes unsustainable. Hour-by-hour symptom journals, detailed food logs, sleep stage analysis from a wearable, cortisol tracking - this is the approach of someone who wants to do it right and ends up doing it for 10 days before quitting.
More variables are not more signal. Usually they're more noise and more friction. The core daily state metrics - energy, mood, libido, sleep, recovery - logged consistently every day, is the foundation. Everything else is secondary and optional. Build the habit with the minimum viable structure first.
What makes a log actually useful
The requirements are simpler than most people expect. Consistent scale: same questions, same 1-10, every day. Logged at the time, not reconstructed later. Daily cadence without significant gaps. Events and state kept separate. Protocol changes logged as dated events so you have the sequence later.
That's it. Nothing complex. The value comes from doing those things consistently over months, not from doing something sophisticated for a week.
The logs that fail are usually the ones that tried to do too much at once, got abandoned during the first disruption, and never recovered consistency. The ones that work are boring by design. Same questions, every day, for a long time.