What to track so AI can actually find patterns

Most people think tracking for AI means journaling — write down what happened, AI reads it, patterns emerge. That's not quite how it works.

Iris does handle the structuring for you — when you say "had a bad headache after lunch, maybe a 6/10," the Tracking Agent transforms that into a structured entry with fields for severity, timing, and possible triggers. You don't need to fill out forms. But the quality of what you say still determines the quality of what gets extracted. "Felt bad" gives the system almost nothing to structure. "Headache started around 2 PM, severity 6, happened after skipping lunch" gives it specific, queryable data points.

Research on symptom tracking published in the Journal of Medical Internet Research found that structured daily logging produces significantly more actionable clinical insights than unstructured journaling. Iris bridges this gap — you log naturally, the system structures — but the more specific your input, the better the structured output.

Four principles separate tracking that generates real insights from tracking that wastes your time.

Timing precision

When you log a symptom, the exact timing matters more than the description. "Headache" is nearly useless. "Headache started around 10:30 AM, peaked at noon, resolved by 2 PM" gives AI a window to search for triggers. Research validating the Food and Symptom Times (FAST) method found that precise timing data reveals correlations invisible to recall-based methods.

The same applies to everything you track. If you suspect coffee is an issue, log when you drank it, how much, and when symptoms appeared. That 2-3 hour window becomes searchable.

Specificity

"Gut felt off" versus "bloating started after lunch, lasted 4 hours, happened again Wednesday same time." The second gives AI actual variables to test — it can cross-reference what you ate, sleep the night before, stress levels, and start building a picture.

Instead of "energy low," try a 1-10 scale with time of day. Instead of "slept badly," track hours and how many times you woke. Measurable specificity is what turns feelings into data.

Co-tracking

This is where patterns hide. Log related factors at the same time, even if they seem unconnected. Got a rash? Also log what you ate, stress that day, sleep quality, and menstrual cycle if applicable. A rash in isolation looks random. A rash that coincides with high stress, poor sleep, and certain foods becomes investigable.

AI finds correlations across multiple variables. The more context you capture in each logging moment, the more correlations it can test.

Baseline first

Before you start eliminating foods or trying supplements, establish a baseline. Log normally for 1-2 weeks without making changes. AI needs to know what "normal" looks like for you — your baseline symptom levels, sleep patterns, energy rhythms.

Then when you introduce a change, AI can compare against your personal normal rather than some population average. This is the principle behind every well-designed clinical trial, scaled down to one person.

The consistency rule

Consistent logging beats detailed logging. Research on ecological momentary assessment published in Psychological Assessment found that brief, frequent measurements produce more reliable data than detailed but sporadic entries. If you log once a day for 30 days, AI can work with that. Three days of intense detail followed by two weeks of silence breaks the pattern-finding.

When you're in a tracking sprint — a focused stretch of consistent logging for a week or two, run because there's a specific question worth answering — pick 5-7 key data points and log them consistently for the duration. You can do this in conversation with Iris (she'll reach out at the times you've agreed on and ask the few things the investigation needs) or by creating entries directly. The system structures everything the same way regardless of how you log it. What matters is that, while the sprint is running, you log consistently. Outside a sprint, this rule doesn't apply — daily logging isn't the baseline.

References

1. Measuring Diet Intake and Gastrointestinal Symptoms in IBS — PubMed Central, 2020. Timing precision in symptom tracking.
2. Mobile health for symptom monitoring — Journal of Medical Internet Research, 2017. Structured versus unstructured symptom tracking.
3. Ecological momentary assessment in behavioral research — Psychological Assessment, 2007. Brief frequent measures outperforming sporadic detailed entries.