Bad Days vs. Real Changes: How to Tell the Difference

If you track cognitive performance every day, you will have bad days. Not because anything is wrong — but because cognitive performance is variable by nature, and roughly 15-20% of sessions will fall meaningfully below your personal baseline just from normal statistical variation. This is inherent to any measurement of a variable system.

This means: if you take Keel daily for a year, you should expect around 55-75 sessions that look bad by the numbers, without anything actually being wrong. The question is not 'did I score below baseline?' — that will happen regularly. The question is 'how does today's score fit into my longer pattern?'

The most common causes of a genuine but reversible dip in cognitive performance — and the ones worth ruling out before worrying about anything progressive — are: poor sleep (by far the largest single factor; even one night of less than six hours measurably suppresses processing speed, reaction time, and working memory), acute illness (a cold, flu, or any systemic infection produces cognitive slowing that resolves with recovery), alcohol from the previous day (even moderate drinking the night before measurably suppresses cognitive performance the following morning, largely through REM sleep disruption), a new medication (many common medications — antihistamines, sedatives, some blood pressure drugs, some antidepressants — have cognitive side effects), high psychological stress (sustained elevated cortisol suppresses working memory and executive function), and significant time zone disruption or irregular sleep schedule.

Before interpreting a downward trend as concerning, ask: did anything change in my life around when this started? New medication, new job, major life stress, sleep disruption, illness? A trend that maps onto a clear life event and reverses when that event resolves is almost certainly not progressive cognitive decline.

A pattern worth taking seriously has several characteristics: it persists over 4-6 weeks without a clear reversible explanation; it appears across multiple cognitive domains rather than just one; it does not recover when obvious causes are resolved; and it is accompanied by functional changes — things you notice in daily life, not just in your Keel scores.

The last point is important. Keel catches subtle data patterns that may precede noticeable symptoms. But a decline that shows up only in Keel data with no corresponding change in your daily function is less concerning than one where both converge. If your scores are declining and you are also noticing word-finding difficulty in conversations, slower driving reactions, or increased difficulty following complex tasks, those two data streams reinforce each other.

Keel lets you log contextual notes after each session: sleep quality, stress level, whether you were ill, significant events. These notes become invaluable when you are trying to interpret a trend retrospectively. A two-month dip that maps perfectly onto a period of severe work stress and poor sleep — documented in your session log — tells a very different story than an unexplained two-month dip with no context.

The discipline of logging context is small — 10 seconds after each session — and the payoff when you or a doctor needs to interpret your trend is significant. It transforms your cognitive data from a graph into a narrative.

If you have a genuine, persistent, unexplained trend over 6-8 weeks, the appropriate next step is to see your primary care physician. Bring your Keel data — the trend charts, the session log, the domain-specific breakdowns. Ask for a review of your medications (many cognitive effects are medication-related and entirely reversible), basic labs (thyroid function, B12, CBC, lipids, blood glucose), and if indicated, a referral for formal neuropsychological evaluation.

Keel is not a diagnostic tool. It cannot tell you whether a trend reflects early neurodegeneration or a reversible cause. It can tell you that a trend exists and give you data to bring to the professional who can answer that question. That is the right division of labor.