Learning your own diabetes

Running one change at a time

Why changing five things at once teaches you nothing, what a fair test of one change looks like, and how to keep learning when life interrupts it.

Written by Updated 5 June 2026 6 min read
The 60-second answer

When a pattern needs fixing, the instinct is to fix everything at once: breakfast, timing, the works. Change five things and something will improve, but you will never know what did it, so you cannot keep it, repeat it, or trust it. One change at a time is how you actually learn your own diabetes.

  • Glucose answers to many things at once, so a result only teaches you something when one variable moved and the rest held still.
  • Changes that touch insulin, background, ratios, corrections, are worked out with your diabetes team. The one-at-a-time discipline applies to those too: it is how the team can tell whether their change worked.
  • Give a change enough days to show its true shape. Glucose is noisy day to day, and judging a change on one day is judging noise.
  • Write down what changed and when. Two lines in your phone turns “I think it helped” into something you can actually see.
  • Life does not always cooperate, illness or a chaotic week resets the experiment, and that is fine. Pause it and pick it up when things settle.
What it means for you

Why five changes teach you less than one

Suppose mornings keep going wrong, so in one determined week you change your breakfast, move your insulin timing with your team’s blessing, start walking after eating and switch to earlier nights. Mornings improve. Which change did it? You cannot know. So now you are stuck doing all four forever, or dropping one at random and hoping, and if mornings drift again you have four suspects and no evidence.

Now run the same week with one change. If mornings improve, you know why, and you keep it with confidence. If they do not, you cross it off and try the next idea. Either way you learned something true about your own body, and truths like that accumulate into the thing this whole site is about: knowing your own diabetes better than any general rule does.

Slower feels worse in the moment and is dramatically faster over a year, because you never have to relearn anything.

Running it properly

What a fair test of a change looks like

Decide what you are changing and what you expect it to affect, before you start. “Porridge instead of toast, expecting a gentler mid-morning” is a test. “Eating better” is a hope. If the change involves insulin in any form, it is designed with your team; the discipline here is about how you run and read it, not about moving dials yourself.

Then hold it steady for a stretch of ordinary days, most changes need a week or so before the signal shows through the noise, and judge it on the days as a group, not on the best or worst one. A single rough day inside a good week does not sink the change; a single good day inside an unchanged pattern does not prove it. Keep the note honest: what changed, the date, and a line on what you are seeing.

And keep everything else boring while the test runs. The steadier the background, the faster the answer arrives.

When life interferes

When the experiment gets interrupted, and what counts as an answer

Illness, a holiday, exams, a chaotic week: any of these scrambles glucose enough to drown whatever you were testing. When that happens the experiment is not failed, it is paused. Note the interruption, ride out the chaos, and restart the clock when things settle. Data from a flu week tells you about flu, not about breakfast.

Know what you will accept as an answer too. Some changes will come back “no difference”, and that is a real result: it frees you to stop doing something fiddly that was not earning its keep. Take what you learn, either way, to your team. A tidy record of “we tried this, held it for two weeks, here is what happened” is the most useful thing you can put in front of them, and it makes their next suggestion sharper.

What varies

What changes how long a test needs

A quiet, regular fortnight can give a clean answer in a week. Shift work, monthly cycles and unpredictable weeks stretch the timeline, because the background keeps moving underneath the test. Change too much at once and you also lose the safety margin: a big cut can send you high, a big increase can drop you into a hypo (a low, below about 4 mmol/L), which is another reason to move one step at a time. A CGM shortens it, since you see whole days rather than snapshots. The size of the change matters too: something that shifts a whole meal’s shape shows up faster than a subtle tweak. When in doubt, run it longer than feels necessary.

how noisy your days areschool or shift patternshormones and monthly cyclesCGM vs finger-prickshow big the change iswhat else is going on in life
Real-life examples

The discipline in action

Examples, not instructions or doses.

The breakfast swap

Two weeks of porridge instead of toast, nothing else touched. The mid-morning climb softens on most days, so you keep it, and you know exactly what you are keeping and why.

The everything week

New breakfast, new bedtime, new walk, all at once, and mornings improve. A month later they drift again and you have no idea which lever mattered. The improvement taught you nothing you could reuse.

The honest null result

You test an earlier dinner for a fortnight and the overnight line looks identical. Not a failure: one suspect eliminated, one fiddly habit you no longer need, and a cleaner question to bring to your team.

What to notice

Worth paying attention to

Whether more than one thing moved at once, which quietly voids what the result seems to say
Judging on a group of ordinary days rather than the best or worst single day
Interruptions, illness, travel, chaos, which pause a test rather than fail it
Writing the change and date down at the start, because memory will swear it was different
Null results, which are real answers: something ruled out is something learned
What to ask your team

Questions that make an appointment useful

"We are considering a change, how many days would you want it held before we judge it together?"
"Which patterns in my data would you most like a clean test of first?"
"If a test involves my insulin, how do we set it up so I know exactly what to watch for while it runs?"
"Can I send you the notes from a finished test between appointments, and what format is most useful?"
Sources