---
title: "How to track a peptide stack so the log means something | Peptyn"
description: "Record design for multi-compound protocols. What to write down, when, and why a log without dates or a defined outcome attributes nothing to anything."
canonical: "https://peptyn.orlyn.ai/articles/how-to-track-a-peptide-stack"
last-updated: "2026-08-16"
---
# How to track a peptide stack so the log means something
*Record design for multi-compound protocols. What to write down, when, and why a log without dates or a defined outcome attributes nothing to anything.*
Educational · Not medical advice · 18+
Some version of this question turns up wherever people compare protocols: *I am on several things at once. Which one is doing something?*
Published 15 August 2026 · 11 min read · 6 sources
The honest answer is that the record cannot say. Not because the compounds are inert and not because the person asking is careless, but because a log built one way can support attribution and a log built another way cannot, and the difference is decided before the first entry, not after.
This article is about that difference. It is record design, not protocol design. Knowing how to track a peptide stack is a separate skill from choosing one, and only the first is entirely yours. What you take, when you start it, and whether you take one thing or six is settled between you and your prescriber. What is entirely yours is the quality of the record you keep about it, and that is worth more attention than it usually gets.
## Why a record of simultaneous changes cannot attribute
Attribution is not an intuition problem you can solve by trying harder. It is a structural property of the record.
The relevant term is confounding. StatPearls defines it as what happens "when the effect of an exposure on the outcome intermingles with the effect of another factor (confounder) not under study, leading to an inaccurate assessment of the true causal effect of interest."[4]
When two things begin on the same date, each one is the other's confounder. Say your protocol includes something that changes appetite and something that changes sleep, both starting the same Monday. Three weeks later your training sessions feel better. Your log can prove that both began on that Monday and that the change appeared later. It cannot separate the two, because nothing in the record ever varied independently. No amount of re-reading fixes that. The information was never captured.
This is also why waiting does not help. A longer stretch of the same undifferentiated record holds more entries but no more separation between the things that started together. Confidence that grows as the file grows is tracking the size of the file.
## What clinical research does about this, and what it does not
There is a formal design for the problem of learning something about one specific person: the N-of-1 trial. Clinical investigators at McMaster University named it in a 1986 *New England Journal of Medicine* paper, describing single-patient trials in which one patient runs through a series of paired treatment periods, one active and one comparator, with the order set by random allocation.[1][2]
It matters what that design actually contains. N-of-1 trials are "multiple crossover trials done over time within a single person," and running one properly means specifying the periods, the sequences, the randomization, and the run-in and washout intervals between them.[3]
A personal log has none of that. It is not randomized, nothing is blinded, there is no comparator period, and there is no crossover. Anyone who tells you that writing things in a notebook makes you an N-of-1 trial is selling something.
What a personal log can borrow is the part of the design that survives without the apparatus:
1. **A prespecified outcome.** Decide what you are watching for, in words, before there is anything to observe.
2. **A stretch of record before the change.** Whatever you are already doing, logged, so that later entries have something to be different from.
3. **Dated changes.** Every change to the protocol carries the date it happened.
4. **Everything else held steady and written down.** Sleep, training, travel, illness, other medications, stress.
5. **Specific observations, recorded close to when they happen.**
That is a reasonable record. It is not a trial, and the gap between the two is exactly the set of claims you are not entitled to make from it.
## What goes in the record
The difference between a useful entry and a useless one is specificity. "Feeling better" is not data. A log that can support attribution has three properties.
**A defined outcome.** Written down before there is a change to evaluate. Not "feel better" but the specific thing: sleep onset, how the third set feels, energy at 3 PM, nausea after eating, a mood rating on a scale you define and then reuse. Writing it down is not decoration. It constrains what you attend to, and it stops you from quietly relocating the goalposts to whatever happened to improve.
**Dates.** When did the protocol change? When did you first notice something? How many days apart were those? "A few weeks in" is a story. A date is a record, and dates are the only thing that lets one stretch of the log be compared against another.
**Specificity.** "Was sick" is not comparable to anything. "Woke 5:30 with a headache, gone by 10, first time this month" is. The more specific the entry, the more of it survives contact with normal week-to-week variation.
There is a fourth property that is really a habit: write it near the time it happens. Recall bias is a recognized source of error in health research, defined as "any error due to differences in an individual's recollections and what truly transpired." StatPearls gives two ways to blunt it: shorten the interval being recalled, since accuracy is better over shorter periods, and corroborate subjective recollection against records or other objective measures wherever possible.[5] A contemporaneous log is both of those at once, which is most of the argument for keeping one. A log filled in on Sunday for the whole week has already given up the first.
## The stretch of record before a change
If your protocol is about to change, the days before the change are not dead time in your log. They are the reference. Without them, any later observation is being compared against a memory, and memory is the thing you started keeping a log to avoid.
This is the least glamorous part of record-keeping and the part most often skipped, because logging a week in which nothing new happens feels like logging nothing. It is the opposite. It is the only entry set that establishes what your normal actually looks like in writing, including how much it varies on its own. It is easy to underestimate how much sleep or energy moves week to week with nothing changed at all, which is precisely why any change that follows a new compound looks meaningful.
A reference stretch is doing its job once it shows a range rather than a single number. One quiet week gives you a point. Several give you the spread, and the spread is what later entries have to clear before they mean anything.
The asymmetry is hard to miss: people put months into a protocol and almost nothing into the design of the record that is supposed to tell them about it.
### If you are already in the middle of one
If you did not start logging on day one, you are in the ordinary case, and the advice to record a reference period before a change arrives too late to be useful. You are somewhere inside a peptide stack you have been running for months with nothing behind you but recollection.
Two things are still available, and both are worth more than they sound.
The first is that today is the reference period for the next change. Whatever you are on now, logged consistently from now, becomes the stretch that any later adjustment gets compared against. A reference period is not a one-time opportunity that closed behind you. It is whatever you are recording while nothing is changing, which is most of the time.
The second is a decision about the months already behind you: leave them empty. The temptation is to sit down and reconstruct twelve weeks from memory so the log looks complete from the start. That reconstruction is not a record. It is a recollection in a record's handwriting, and it carries exactly the distortion the log exists to avoid, with none of the warning labels. If you want the earlier period represented, write down only the dates you are genuinely confident about, mark the entry as reconstructed, and keep it visibly separate from things written at the time. A log with an honest gap in it is worth more than one with an invented middle.
## What a good record can tell you
Over a few months, a well-kept log can show:
- Whether the outcome you defined in advance looks different after a dated change than before it.
- Roughly when any difference appeared, relative to that date.
- Whether it was consistent or came and went.
- Whether it persisted after the protocol changed again, which is why the log continues past the change.
- What else was going on at the time, because you wrote that down too.
That is genuinely useful. It is not proof in a statistical sense, and it is honest about being one person's observations in sequence.
## What a good record cannot tell you
This list is the more important one, and most content in this space omits it entirely.
- **Why something happened.** Mechanism lives in the research literature and in the conversation with your prescriber, not in your log.
- **Whether anyone else would observe the same thing.** One person, unblinded, unrandomized, sample size one.
- **Whether you observed something because of the compound or because you expected to.** Expectation effects are real effects and a personal log cannot separate them from anything else. It records the total.
- **Anything about dose.** Your log stores what you entered. It does not work backward from an outcome to a number, and it should not: a record that started inferring doses from how you felt would be inventing them.
- **What should happen next.** That is a clinical decision. The log is an input to that conversation and never the output of it.
It is worth sitting with how much the published evidence itself leaves open, because it sets a ceiling on what a notebook can settle. A 2023 review in *Endocrine Reviews* covering NAD+ in aging biology concluded that the "clinical pharmacology, metabolism, and therapeutic mechanisms of NAD+ precursors remain incompletely understood," and called for adequately powered randomized trials.[6] If controlled research describes its own conclusions that cautiously, a personal log is not going to resolve the question. It can only tell you what you observed, and when.
## What an entry actually looks like
Structure beats prose here. An entry that can be used later carries four things: the date, whether the protocol changed that day and how, the outcome you committed to tracking, and the context that could explain a change on its own.
A week of that reads something like this:
```
Mar 3 protocol unchanged sleep onset ~25 min normal week, trained Tue/Thu
Mar 4 protocol unchanged sleep onset ~20 min normal
Mar 5 protocol unchanged sleep onset ~45 min late meeting, ate at 22:00
Mar 6 protocol unchanged sleep onset ~25 min normal
Mar 7 CHANGED: added [compound], first administration, evening
sleep onset ~30 min normal otherwise, no travel
Mar 8 protocol unchanged sleep onset ~20 min normal
Mar 9 protocol unchanged sleep onset ~35 min normal
```
Nothing in that is clever. What makes it usable a month later is that the change is on a dated line of its own, the outcome is the same measured thing every day, and the context column is doing real work. That March 5 entry is why you will not later attribute a bad night to a compound you had not started yet.
Note what this week does not show. The three entries after the change all sit inside the range the four entries before it already covered. That is the ordinary result, and it is the reason the reference stretch matters: without those first four lines, there is no way to know whether a 30 minute night is unremarkable or not. A log that began on March 7 would have contained the same three numbers with none of the context, and whoever read it later would have been free to invent a story about them.
Notice also what the record does when the protocol changes several things at once. It records all of them, faithfully, on the same date. It just does not pretend afterwards that it can tell them apart. A record that is honest about a confounded stretch is more valuable than one that quietly assigns credit.
## The log as a shared document
There is a good reason to keep this well beyond your own curiosity: a dated contemporaneous record is the version of your history that has not been edited by how things turned out. Recall is systematically shaped by outcome, which is why recall bias is a named problem in study design rather than a personality flaw.[5] Your memory of the last three months is being rewritten in small ways every time you feel differently about them. The entry you made on March 7 is not.
If you see a prescriber, a pharmacist, or a coach, that record is the most honest thing you can bring to the appointment. It answers the questions that are otherwise answered with "a few weeks ago, I think."
Peptyn is built to hold the dated half of exactly this: what you entered, when each thing started, and a history calendar for reading weeks at a time. It has no free-text field, so the surrounding notes stay wherever you already keep them. It does not calculate doses. Making what happened visible is the whole job.
## Start where you are
You do not need to change anything about your protocol to start keeping a record worth having. Today's entry can be the first one. Write down what you are already doing, the date, one outcome you actually care about, and what else was going on.
Then keep going through the next change, whenever it comes and whatever it is. The difference between a log that answers questions and a log that just accumulates is decided in the first week, and it costs about five minutes a day.
## References
1. Guyatt G, Sackett D, Taylor DW, Chong J, Roberts R, Pugsley S. Determining optimal therapy: randomized trials in individual patients. *N Engl J Med*. 1986;314(14):889-892.
2. Mirza RD, Punja S, Vohra S, Guyatt G. The history and development of N of 1 trials. *The James Lind Library*. 2017.
3. Porcino A, Vohra S. N-of-1 Trials, Their Reporting Guidelines, and the Advancement of Open Science Principles. *Harvard Data Science Review*. 2022. PMCID PMC10686313.
4. Dhawan R, Shay D. Principles of Causation. StatPearls, NCBI Bookshelf.
5. Popovic A, Huecker MR. Study Bias. StatPearls, NCBI Bookshelf.
6. Bhasin S, Seals D, Migaud M, Musi N, Baur JA. Nicotinamide Adenine Dinucleotide in Aging Biology: Potential Applications and Many Unknowns. *Endocr Rev*. 2023;44(6):1047-1073.
## Keep reading
- [What to write down when something feels off](https://peptyn.orlyn.ai/articles/side-effects-tracking)
Severity, duration, and the timing axis almost everyone records at the wrong resolution.
- [What a six-month injection site log shows](https://peptyn.orlyn.ai/articles/glp1-injection-site-rotation-log)
What the rotation research really found, which drug it was found in, and what your own site history adds that a chart cannot.
- [The glossary that will not do your math](https://peptyn.orlyn.ai/articles/peptide-glossary)
Plain definitions from lyophilized to 503A, with sources. It defines every term and stops where the arithmetic starts.
This article is educational. It does not recommend doses, schedules, or products, and it is not medical advice. Every factual claim above is linked to its source. Questions about your own protocol belong with the prescriber who wrote it.
---
Source: