What to track for chronic fatigue — and what's just noise

You're already exhausted. The last thing you need is a homework assignment. "Track your food, sleep, stress, exercise, mood, supplements, caffeine, water intake, and bowel movements for six weeks" is a recipe for doing it perfectly for three days, then never again.

Fatigue tracking needs to be minimal enough to sustain and specific enough to be useful. The goal isn't capturing everything — it's capturing the right things consistently so AI has enough data to map your system.

The energy curve — your most important metric

Before tracking causes, track the thing itself. Three times a day, rate your energy on a 1-10 scale: morning (within an hour of waking), afternoon (around 2-3 PM), and evening (around 7-8 PM).

That's it. Three numbers. Takes 10 seconds.

After two weeks, those three daily ratings reveal your energy shape — and the shape is diagnostic. A flat low line all day suggests systemic causes (thyroid, depression, chronic inflammation). A morning low that improves suggests sleep quality issues. An afternoon crash suggests blood sugar or caffeine patterns. A morning peak that collapses suggests energy depletion from unsustainable output or poor recovery.

Research in Psychosomatic Medicine found that the pattern of energy fluctuation across the day is more informative than average energy level for identifying fatigue drivers. The shape tells you where to look.

The six things worth tracking alongside energy

Once your energy curve is established, these parallel variables let AI test what's driving the pattern:

Sleep — quality, not just quantity. Two things: hours slept and how rested you feel on waking (1-10). If you use a wearable, that data is useful too — but the subjective "how rested do I feel?" score is the single most predictive metric. Research in Psychosomatic Medicine found it outperforms any objective sleep measurement in predicting daytime function. Also note: what time you went to bed and woke up. Schedule consistency matters as much as duration.

Food timing and composition. Not a food diary — just what you ate and roughly when. "Cereal at 8, sandwich at 1, pasta at 7" is enough. The Tracking Agent structures it from there. What matters for fatigue investigation is the timing (did you skip meals? were gaps too long?), the composition (high-carb without protein or fat?), and the relationship to your energy crashes. If your 2 PM crash comes two hours after a carb-heavy lunch, that's a testable hypothesis.

Caffeine. Type, amount, and — critically — timing. Caffeine has a half-life of 5-6 hours, but research in the Journal of Clinical Sleep Medicine found it disrupts sleep architecture even when consumed 6 hours before bedtime. The caffeine-sleep-fatigue loop (poor sleep → more caffeine → worse sleep → more caffeine) is one of the most common and most modifiable fatigue patterns. Track it so AI can tell you whether it's running.

Stress. A daily score (1-10) or high/medium/low. If you can note what the stress is about, even briefly, that helps distinguish types of stress that drain you from those that don't. Work stress and relationship stress may have different effects on your sleep and energy.

Medications and supplements. Everything you take, every change. Many medications cause fatigue as a side effect — beta-blockers, antihistamines, SSRIs, statins, blood pressure medications, and others. Supplements can also affect energy: iron and B12 supplementation can help if you're deficient, but some supplements (melatonin, magnesium, certain herbs) can cause daytime drowsiness. If you're taking supplements, ask Iris to review them — some might be contributing to your fatigue rather than helping.

Activity. What you did physically and when. Not detailed exercise logs — just "walked 30 min morning" or "gym evening" or "barely moved today." This matters because both too much and too little activity affect fatigue. Morning exercise often improves daytime energy; evening intense exercise can disrupt sleep. And the inactivity spiral — too tired to move, which makes you more tired — is one of the most important patterns to identify.

What you can skip

You don't need to track water intake (unless you genuinely suspect dehydration). You don't need to measure precise portions or count calories. You don't need to track mood separately from energy unless you suspect the fatigue-mood loop is active (in which case a simple daily mood score helps).

The goal is six variables tracked consistently alongside your energy curve. Not more.

Making it sustainable when you're already exhausted

This is the tension: the people who most need to track fatigue are the people with the least energy to do it.

The end-of-day voice note. At bedtime, just talk for two minutes. "Energy was maybe a 4 this morning, perked up to a 6 around 11, crashed hard at 2, maybe a 3, recovered a bit by evening, 5. Slept about seven hours last night, felt like a 4 on waking. Had coffee at 8 and 1. Skipped breakfast, had a big lunch around noon, dinner was light. Pretty stressed today, maybe a 7. Didn't exercise." The Tracking Agent structures all of that — energy curve, sleep, caffeine, meals, stress, activity. You don't fill out forms. You ramble. It works.

Conversational logging with Iris. If voice notes aren't your style, Iris can run the sprint as a back-and-forth in chat — she reaches out at the times you've agreed on and asks the few things the investigation needs (energy now, sleep last night, caffeine, what you ate, stress). You answer in a sentence or two and the Tracking Agent structures it from there.

Three numbers minimum. If you can't do anything else, just log your three daily energy ratings. That alone, over two weeks, gives AI something meaningful to work with. Everything else is refinement. Don't let perfect tracking prevent any tracking.

When wearable data helps (and when it doesn't)

If you have a fitness tracker or smartwatch, the sleep data is useful — duration, consistency, and broad disruption patterns. But don't mistake the numbers for the full picture. Consumer wearables are reasonably accurate at total sleep time but poor at sleep stage classification, and they can't measure the subjective quality that matters most.

Research in the Journal of Clinical Sleep Medicine found significant discrepancies between consumer wearable sleep staging and clinical polysomnography. Use wearable data as one input alongside your subjective ratings, not as a replacement for them.

The most useful wearable insight for fatigue is the objective-subjective mismatch: nights where the wearable says you slept well but you feel terrible, or vice versa. That mismatch itself is diagnostically informative — it points toward sleep quality problems that the wearable's algorithms can't detect.

When a tracking sprint makes sense

You don't need to track your energy forever. Most of the time, you and Iris are just talking — about how you're doing, what your labs showed, what you've tried, what you're worried about. Tracking enters the picture when there's a question worth answering: is the afternoon crash actually a blood-sugar pattern? Is caffeine after noon costing you sleep? Is the new thyroid dose moving the morning baseline?

That's when Iris may suggest a tracking sprint — a focused stretch of consistent logging for two or three weeks, aimed at the question on the table. When the sprint ends, the daily logging ends with it. Outside a sprint, the relationship is conversational.

How long a sprint needs to run

Two weeks of consistent daily tracking inside a sprint is the minimum for meaningful analysis. Three weeks is better. The energy curve stabilizes, the parallel variables accumulate enough repetitions, and AI can start testing whether your 2 PM crash consistently follows certain meal compositions, or whether your worst mornings consistently follow high-stress days.

Don't analyze too early. One week almost always shows "patterns" that are coincidences. Wait for the data to accumulate, then ask Iris: "Look at my energy ratings alongside my sleep, caffeine, food, and stress data and find the strongest correlations." The Data Analyst runs the multi-factor comparison and identifies which variables most strongly predict your energy levels.

References

Sleep quality versus sleep quantity — Psychosomatic Medicine, 2006. Subjective quality as strongest predictor of daytime function.
Caffeine effects on sleep taken 0, 3, or 6 hours before bedtime — Journal of Clinical Sleep Medicine, 2013. Caffeine timing and sleep disruption.
Diurnal variation of fatigue and its clinical significance — Psychosomatic Medicine, 2006. Energy curve shape as diagnostic tool.
Consumer sleep technology accuracy — Journal of Clinical Sleep Medicine, 2019. Wearable limitations and appropriate use.
Medications causing fatigue — American Family Physician, 2018. Common medication side effects including fatigue.