Your migraine tracker starts its clock at the wrong zero.
Prodrome to postdrome, one attack can run for days. Most trackers open a form at head pain and close it when the pain stops, which records only the middle third.
The attack that started nineteen hours before the headache
At 3:40 on a Tuesday afternoon, the yawning starts. Not the tired kind. Eleven yawns in twenty minutes, in a meeting, after eight hours of sleep.
By six the neck is stiff on one side. By seven there's a specific, insistent want for something salty, and the crackers go fast. By nine the words come out slightly rearranged and the kitchen light sits a half-stop too bright. Bed at eleven, sleep at one.
The head pain lands at 10:20 the next morning. That's when the app gets opened. Attack start: Wednesday, 10:20am. Suspected trigger: the salty snack, or the bad sleep.
Both of those entries are wrong. The craving was part of the attack. The disrupted sleep was part of the attack. The attack didn't begin at 10:20 on Wednesday, and the only surviving record of its first nineteen hours is a memory that a day of head pain is about to blur.
When does a migraine attack actually start?
Migraine has four described phases, and only one of them is the headache. Prodrome (the premonitory phase) comes first, then aura for some people, then head pain, then postdrome. This isn't a fringe model. It's the structure the International Classification of Headache Disorders uses, and the one the American Migraine Foundation puts in front of patients.
So here's the gap worth naming: nearly every migraine diary on the market makes its first required field the start time of head pain.
Call it the wrong zero. The diary's clock and the attack's clock disagree by anywhere from a few hours to roughly two days, and every number computed downstream inherits that offset. Duration is short. Frequency is undercounted. Time-to-treatment is measured from the wrong anchor. And the trigger list gets contaminated with things the attack itself caused.
Phase
Timing relative to head pain
What people report
What a standard diary records
Prodrome (premonitory)
Begins up to about 2 days before
Yawning, fatigue, neck stiffness, food cravings, mood shift, thirst and frequent urination, trouble concentrating, early light sensitivity
Nothing, or a retrospective 'trigger' tag added later
Aura
5 to 60 minutes, just before or overlapping pain; occurs in a minority of people with migraine
Visual disturbance, sensory changes, speech or language difficulty
An optional checkbox
Headache
4 to 72 hours if untreated (ICHD-3)
Pain, nausea, photophobia, phonophobia
The full record: start time, severity, medication, end time
Phase definitions and timing per ICHD-3 (International Classification of Headache Disorders, 3rd edition) and American Migraine Foundation patient materials. The fourth column describes the common required-field set across mainstream migraine diaries, not any single product.
Why "headache started at 8am" is the wrong first field
The pain-onset field didn't come from the disease. It came from the paperwork.
Headache days per month is the unit clinical trials use for endpoints and the unit a neurologist has to extract in a seven-minute appointment. It's a scoring convention, not a description of the illness. Diaries copied the convention and inherited its blind spots, because a timestamp on pain is easy to collect and a mood shift at 4pm on Tuesday is not.
The cost of that shortcut lands on the person doing the logging. Premonitory symptoms are highly convincing as causes: a craving arrives, you eat the chocolate, the headache follows, and chocolate goes on the trigger list. Same with neck stiffness, which patients frequently log as a posture trigger when the headache literature has described it as one of the most common premonitory symptoms. A diary that starts at pain can't tell those apart. It hands you a list of suspects assembled from the attack's own opening symptoms.
What the research says about premonitory symptoms
Two lines of evidence matter here.
The first is behavioral. Electronic-diary work going back to Giffin and colleagues in 2003 asked people with migraine to log non-headache symptoms in real time and then flag when they expected an attack. People who reported premonitory symptoms were able to predict oncoming headache at a rate well above chance Giffin et al. (Neurology, 2003) found patients correctly predicted migraine headaches from 72% of diary entries with premonitory symptoms. ([source](https://pubmed.ncbi.nlm.nih.gov/12654956/)). Patients could see it coming. The diary just wasn't asking.
The second is physiological. Schulte and May's 2016 imaging work in Brain scanned a single migraine patient daily across 30 days and three spontaneous attacks, and found hypothalamic activation rising in the roughly 24 hours before pain onset. Whatever the prodrome feels like from the inside, something measurable is already happening.
Then there's treatment. The PRODROME trial tested a gepant taken during the premonitory phase, before pain began, against placebo Dodick et al., The Lancet, 2023: PRODROME trial primary endpoint was absence of moderate/severe headache within 24h after ubrogepant 100 mg dose during prodrome ([source](https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(23)01683-5/abstract)). Treating that window is now a live clinical question, which means the window itself has become something worth being able to identify in your own data.
This is a measurement argument, not a dosing one
Premonitory symptoms are common but not universal, not diagnostic, and not the same as aura. Plenty of attacks arrive with no warning at all, and a yawning afternoon does not reliably mean an attack is coming. Any decision about moving an abortive earlier belongs with the person who prescribes it: acute-medication frequency limits and medication-overuse headache are real constraints, and treating a prodrome that wasn't going to become an attack still counts against them. Knowing your own prodrome pattern is what makes that a useful conversation to have.
Three days that read as one bad Wednesday
Take the Tuesday above and run it through both clocks.
Under the pain clock, the record reads: one attack, Wednesday, 10:20am to 9pm. Ten and a half hours. Trigger: salty snack. One headache day. Thursday, when the brain fog made a two-paragraph email take forty minutes, is a blank.
Under the attack clock, the same event reads: prodrome onset Tuesday roughly 3:40pm (yawning, then neck stiffness, then craving, then word-finding trouble), pain Wednesday 10:20am to 9pm, postdrome through Thursday afternoon. About two and a half days of impairment, of which the head pain is the middle third.
Now scale that to a month with four attacks. One version tells a neurologist four headache days. The other shows ten to twelve days where something was demonstrably off, with the shape of each one visible. Those two records point at different treatment conversations, different accommodation requests, and different answers to "is the preventive working?"
What changes if the clock starts earlier
Four things move, and they're not subtle.
Trigger lists get shorter and more honest. Once cravings, thirst, neck stiffness, and early light sensitivity are logged as phase-one symptoms with their own timestamps, they stop masquerading as causes, and whatever remains on the list is worth actually testing.
Frequency numbers change. Preventive thresholds and insurance authorizations are counted in headache days, so systematic undercounting of attack burden is not a rounding error, it's the difference between qualifying and not.
The treatment window becomes visible. You can't discuss acting during prodrome if you have no record of your own prodrome length or its usual first symptom.
And the days you can't explain to your employer stop being unexplained. Tuesday afternoon and Thursday morning were part of the same event as Wednesday.
Pain onset is the loudest moment of a migraine attack. It has never been the first one.
Building from both ends of the attack
We built Postdrome from the back end of this problem. The recovery days went unrecorded everywhere we looked, so we started there: a continuous symptom timeline that keeps running after the pain stops, because for a lot of people that's where the lost day actually is.
The same argument runs in the other direction, and it has to, or the timeline is still missing a third of the event. Prodrome symptoms log as their own phase with their own start time, not as a trigger tag bolted onto a headache entry. Aura Mode exists because logging with a scintillating scotoma across half your field is a real problem that shouldn't require reading small grey text. Everything stays on your device, and your data exports whenever you want it, in a form you can hand to a neurologist or take somewhere else entirely.
What we don't do is tell you an attack is coming. Nobody can do that reliably from a phone diary, and claiming otherwise would be the kind of thing this piece is arguing against. You record what you notice. The timeline shows you the shape it makes.
What we're watching next
Two open threads. Whether prodrome-window treatment moves from trial result into routine prescribing practice, which would make an accurate personal prodrome record clinically load-bearing rather than merely interesting. And whether the headache-day counting convention gets revisited by the guideline bodies, since it's the measurement layer sitting under preventive thresholds, trial endpoints, and prior-auth decisions all at once.
We'll write both up when there's something specific to report, not before.
Start logging the whole attack, not the middle of it. Postdrome records prodrome, pain, and the recovery days on one timeline, keeps it on your device, and costs one payment for good.
We built Postdrome because of migraines in our lives, and because every tracker we tried treated the headache as the whole event. It logs the full cycle: premonitory symptoms with their own timestamps, the pain, and the recovery days after. One continuous timeline instead of disconnected headache entries. On-device by default, exportable on demand, and one purchase rather than a subscription that quietly renews while you're too foggy to cancel it.
