Caffeine, measured honestly

TrackCaffeine exists to answer two questions accurately: how much caffeine is in a drink, and how long it stays with you. Most pages on the web answer only the first, and often with numbers copied around without a source. We pair a curated dataset with a transparent decay model so the second question gets a real answer too.

Our data

Caffeine values come from public manufacturer disclosures, USDA FoodData Central, and published measurements. They are estimates — caffeine content varies with brewing, batch and serving size — so we label them as such and never present them as exact or medical figures. The full set of 87 drinks is published as an open dataset you can download as JSON or CSV and check or reuse.

Our method

Caffeine is eliminated through first-order pharmacokinetics. We model each dose as decaying with a half-life — 5.7 hours by default, the commonly cited population average — and sum the doses to estimate how much is in your system at any moment. This is the same engine that powers our iOS app, CoffeeLog. It is a model, not a measurement of your blood, and individual metabolism varies widely.

Who makes this

TrackCaffeine is built and maintained by Merey Tleugazin, published by VAST FLOW. CoffeeLog is our caffeine-tracking app for iPhone and Apple Watch. If you spot a value that looks wrong, tell us — accuracy is the whole point.

Not medical advice

Everything here is for general information. It is not medical advice, diagnosis or treatment. If you have questions about caffeine and your health — pregnancy, heart conditions, medication interactions — talk to a qualified professional.

Who writes this site and what does that person do and not do?

Merey Tleugazin is the primary author and maintainer of TrackCaffeine and is responsible for assembling the public dataset and implementing the decay model used across the site. Merey Tleugazin is not a clinician, dietitian or pharmacologist, which means the site does not offer individualized medical advice, diagnosis, or treatment and should be used only as a public information resource.

Merey Tleugazin is the founder of VAST FLOW, the publisher of TrackCaffeine; this is the official publisher name and contact route for inquiries and legal requests. VAST FLOW is the organizational publisher for TrackCaffeine, and you can contact VAST FLOW via our contact page for corrections, permissions, and press inquiries.

The role of the author is to curate and cite caffeine measurements, to implement and document a transparent pharmacokinetic model, and to maintain the open dataset; those are the core, repeatable tasks that produce the site content. The limitation—being explicitly not a clinician, dietitian or pharmacologist—means complex clinical questions (for example drug interactions, pregnancy guidance, or individualized dosing) should be directed to qualified professionals and not resolved by this site.

What is our editorial process from a value entering the dataset to a page being published and revised?

A proposed caffeine value enters TrackCaffeine via a sourced submission, is validated against primary public sources, is added to the open dataset with source metadata, and then a derived page is published after a review cycle. Subsequent revisions follow a tracked-change workflow: source re-check, dataset version bump, page update, and public changelog entry.

1. Submission and sourcing: each new value must cite a primary source (manufacturer disclosure, USDA FoodData Central, or published measurement) and include a URL or citation; entries without primary sourcing are rejected.
2. Validation: an editor compares the submitted value to the cited source and to comparable entries in the dataset; if the source conflicts with multiple manufacturer disclosures or with USDA FoodData Central, the entry is flagged for secondary review.
3. Dataset entry: validated values are added to the open dataset as a new row with a timestamp and source field; the dataset is licensed CC BY 4.0 (Creative Commons Attribution 4.0), which allows reuse with attribution.
4. Model derivation and page rendering: the new dataset value is rendered into the site pages that use it and is used by the public calculators in Tools, including the caffeine half-life calculator.
5. Publication and transparency: published pages include the source citation, a timestamp of the latest dataset version, and a link to the dataset row so readers can inspect the raw source; corrections are annotated in a changelog entry on the page.

All dataset changes are versioned so that every published page can point to the exact dataset snapshot used; this preserves reproducibility and allows readers to verify the arithmetic on a given publication date. The dataset’s license and the requirement for primary sources are enforced at the dataset entry step to maintain provenance and reuse clarity.

Why did we choose a first-order half-life model and where is it deliberately simpler than the pharmacology literature?

We use a first-order elimination model because it is the simplest pharmacokinetic model that produces a transparent, reproducible estimate of how much caffeine remains over time; the model is implemented as remaining = dose × 0.5^(hours ÷ 5.7) with a population-average half-life of 5.7 hours (342 minutes). First-order elimination is a model in which the rate of elimination is proportional to the current concentration, and half-life is the time it takes for a quantity to fall to half its value; both definitions are intentionally concise and used here to keep the model auditable and easy to reproduce.

First-order elimination is a standard pharmacokinetic simplification used when a single, exponential decay fits aggregate population data; half-life is the single parameter that summarizes that decay. The model choice prioritizes transparency over individualized accuracy: it gives a clear arithmetic rule that anyone can apply to any listed dose from the dataset, but it does not model multi-compartment kinetics, genetic variation in CYP1A2, or drug–drug interactions that appear in specialist pharmacology literature.

To illustrate the arithmetic we show two direct examples using values from our dataset: an Espresso (single) at 63 mg after one half-life (5.7 hours) is calculated as 63 mg × 0.5^(5.7 ÷ 5.7) = 63 mg × 0.5^1 = 31.5 mg remaining. A Latte (large, 2 shots) at 126 mg after two half-lives (11.4 hours) is calculated as 126 mg × 0.5^(11.4 ÷ 5.7) = 126 mg × 0.5^2 = 126 mg × 0.25 = 31.5 mg remaining.

The deliberate simplifications are: we do not stratify the half-life by genotype, age, pregnancy status, liver disease, or concurrent medication, and we do not attempt to convert to blood plasma concentration or predict receptor occupancy. For statements about sleep effects we point to a published sleep study that measured sleep impacts directly: Drake et al., 2013 demonstrated that 400 mg taken six hours before bedtime reduced total sleep time by more than one hour (Drake et al., 2013, Journal of Clinical Sleep Medicine).

How do you report an error and what is our correction path if we are wrong?

If you believe a value or a calculation is wrong, report it using the contact form on our contact page with the page URL, the value you believe is incorrect, and the source you prefer us to consult. We will acknowledge receipt, investigate the source material, and—if justified—update the dataset row, publish the corrected page, and record the change in the page changelog.

A correction path is a defined series of actions for fixing erroneous content: receipt, verification against primary source(s), dataset update with a timestamp and source field, page re-render, and changelog entry; this path preserves provenance and visibility. We keep the original source citations on every dataset row and retain previous dataset versions so that any change can be audited; if an update alters a previously published number we include a note on the page stating what changed and why.

Minor factual errors that conflict with the cited primary source are corrected within our normal editorial cycle; material disputes that involve ambiguous or contradictory primary sources trigger a secondary review and, when appropriate, an explanatory note on the page describing the disagreement. If we cannot resolve a disputed value to consensus we will present the competing sources side-by-side on the page instead of removing provenance.

How is TrackCaffeine funded and what is its relationship to the app mentioned earlier on this site?

TrackCaffeine is published by VAST FLOW and is funded primarily by the publisher with supplemental revenue derived from the developer’s iOS app and voluntary support from users; the app uses the same dataset and model as this site. The funding model is transparent: publisher-supported infrastructure plus user-facing product revenue and optional reader support; we do not publish sponsored data that changes the dataset values.

Being publisher-funded means editorial control and dataset stewardship remain with VAST FLOW and are not determined by third-party advertisers or manufacturers. The relationship to the app is operational and technical: the same open dataset and the same first-order decay code power both the site calculators (for example the last-cup-before-bed tool) and the app’s tracking features, which aligns product behavior with the published pages and dataset.

For legal and administrative requests you can reach the publisher through /contacts/, and our operating terms and privacy practices are available at /terms/ and /privacy/.

What will TrackCaffeine never publish?

We will never publish dose recommendations, individualized medical advice, or any claim that cannot be linked to a primary source; those categories are excluded from our editorial remit. Dose recommendations are medical interventions and require a clinician or pharmacist; TrackCaffeine is an informational reference that provides sourced quantities and an auditable decay model, not treatment guidance.

Specifically, we will not publish: personalized dosing guidance, claims about drug interactions that are not supported by primary pharmacology literature, or promotional content that changes dataset values for commercial reasons. If a manufacturer requests a data change we will treat it as a source submission and require primary documentation; we will not accept payments in exchange for changing a published caffeine value.

If you need guidance about safe daily limits we link to authoritative sources in our guides, for example our Safe Caffeine Intake guide and the pregnancy-specific guidance in Caffeine During Pregnancy: How Much Is Safe?; for clinical decisions consult a qualified professional. For public health limits, note that EFSA summarized relevant risk assessments in 2015 and Drake et al. measured sleep effects in 2013 (EFSA, 2015; Drake et al., 2013), and our dataset is published under CC BY 4.0 (CC BY 4.0), which lets others reuse the rows with attribution.