Professional Lactate Analysis Without a Lab: The Gap List
Which parts of a lab lactate test transfer to a home setup and which don't: an 11-point parity audit, a step-grid precision rule and a sampling checklist.
A lab does three separable things to a lactate test: it standardises the conditions, it runs the sampling, and it converts the numbers into thresholds and training zones. Only the third one is software, and software moves. The arithmetic transfers completely. The sampling transfers if you rehearse it. The standardisation is the part nobody hands you, and it is where home datasets usually come apart.
So the useful question is not whether a calculator can match a lab's analysis package. It is which of the remaining capabilities you are willing to rebuild yourself, and which you have decided to live without on purpose rather than by accident.
The eleven capabilities, and which survive the move
| # | Capability | Where it lives | Survives the move? |
|---|---|---|---|
| 1 | Curve fitting and threshold detection | Software | Yes |
| 2 | Choice of detection method and its parameters | Software | Yes |
| 3 | Zone boundaries derived from the thresholds | Software | Yes |
| 4 | Stored sessions and test-to-test comparison | Software | Yes |
| 5 | A chart and table you can hand to a coach | Software | Yes |
| 6 | A written stage protocol you can repeat | Paper | Yes |
| 7 | Load you can trust (ergometer, treadmill belt) | Hardware | Only if you verify it or freeze the machine |
| 8 | Samples taken at fixed points without altering the effort | Operator | Only if rehearsed, or with a second person |
| 9 | A pre-test state you can reproduce | Planning | Only if you record it and repeat it |
| 10 | A second person watching the final stages | Person | No |
| 11 | Medical screening and cover | Clinic | No |
Six transfer as they are. Three transfer only if you do something deliberate about them, and doing nothing quietly counts as a "no". Two do not transfer, and software does not substitute for either.
The two that don't transfer
Rows 10 and 11 are the honest boundary of any home setup. The final stages of a step test are the ones where sampling is hardest and where the decision to stop gets made, and both of those are worse when you are the athlete, the operator and the timekeeper at once.
Screening is not a protocol setting. If you have not been cleared for maximal-intensity exercise, or you get symptoms during a test — chest pain, unusual breathlessness, dizziness — that belongs in a conversation with a doctor, not in a spreadsheet column. Nothing on this page substitutes for that conversation, and no analysis method changes it.
Your step size caps the precision before any calculator opens
You measure lactate at the end of each stage. Between two stages you have no measurement at all, only interpolation, and a fit will happily place a threshold anywhere inside that gap and print it to the watt.
A workable convention: report a threshold no finer than half your step increment, and put that granularity in the output next to the number.
| Step increment | Half a step | Sensible reporting granularity |
|---|---|---|
| 10 W | 5 W | nearest 5 W |
| 20 W | 10 W | nearest 10 W |
| 25 W | 12.5 W | nearest 10 W |
| 30 W | 15 W | nearest 15 W |
| Running, 0.5 km/h | 0.25 km/h | nearest 0.25 km/h |
Worked example on the bike. You want to cover 150–270 W. At 20 W increments that is 7 stages; at three minutes per stage, 21 minutes of test and 7 lactate samples. Halve the increment to 10 W and the same range costs 13 stages, 39 minutes and 13 samples. You bought 10 W of reporting resolution with 18 extra minutes at rising intensity and six extra strips.
A tool that prints 247 W from a 20 W grid is not being precise. It is interpolating across twenty watts you never measured and rounding the guess to the unit. That distinction matters most when you line up two tests from different months — see How to Compare Lactate Test Results Over Time Without False Precision for the comparison side of it.
The twenty seconds at the end of each stage
Everything a lab operator does at a sampling point, you now do while pedalling or holding pace. Count the actions:
- Catch the stage clock at the sampling point.
- Wipe and dry the sampling site.
- Lance.
- Handle the first drop the way your meter's manual specifies.
- Fill the strip to the mark in one contact.
- Read the value when it appears.
- Write it against the correct stage number.
Seven actions, hands occupied, breathing rate high, usually inside twenty seconds.
Here is the decision rule. If you cannot get through steps 2 to 7 without your cadence dropping more than about 5 rpm or your stride breaking, you have two options, and improvising each time is not one of them: bring a second person, or build a fixed 20–30 second sampling break between stages and then repeat that identical break in every future test. A break that appears in one test and not the next does more damage than no break at all.
The failure I would bet on in a solo dataset is not a wrong reading. It is a correct reading written against the wrong stage — one row out, which turns a clean curve into a dogleg and drags whatever the fit calls the upper threshold along with it. Number your log rows and pre-fill the stage loads before you start, not while you are gasping. The mechanics of the sampling routine itself are covered in Self test lactate analysis online: measure and interpret results.
Measure your own noise band before you trust a "first rise"
"First rise above baseline" is a detection rule with a hidden dependency: it assumes you can tell a rise from scatter. Your meter and your technique together produce that scatter, and until you have measured yours, you have no threshold for what counts as a rise.
Do this once, during any test:
- At three different stages, fill two strips from the same drop, back to back.
- Record both values as a pair.
- Take the absolute difference of each pair, then average the three differences.
That average is your working band. It costs three extra strips and roughly a minute of test time.
Illustrative numbers: the pairs come out 1.2/1.3, 2.1/2.4 and 3.6/3.8. The differences are 0.1, 0.3 and 0.2, so the band is 0.2 mmol/L. Now look at three baseline stages reading 1.1, 1.2, 1.3. The whole sequence sits inside the band, so it is not a rise — it is flat data with noise on it. The first movement you can defend is one larger than 0.2. Apply the same band when you compare a threshold from this season against one from last season.
What an LT1/LT2 calculator is deciding for you
A calculator is not reading your physiology. It is making four decisions on your behalf, and each one moves the output:
- Which detection method it applies. Methods can return different numbers from the same dataset.
- Whether it fits a smooth curve through your points or works on the raw points as measured.
- What it does with a point that does not fit — drop it, keep it, or flag it for you.
- What it assumes about your protocol: stage length, units, and whether the final stage was completed.
Three questions to put to any tool before you accept its number:
- Does the output name the method it used? If it does not, the number cannot be lined up against a number from a different tool, or against its own output next season once a default has changed.
- Can you re-run the same dataset under a second method and see the spread? That spread is information: it tells you how much of the threshold is your data and how much is the method.
- Does the saved record carry the protocol with it — stage length, increment, meter, date, pre-test notes?
A tool that answers yes to all three is doing the job the lab's analysis package did. One that returns a single confident number with no provenance has swapped the analysis for a claim. The selection criteria are worked through in Lactate test analysis software: how to choose the right tool.
Protocol questions that decide whether the data is usable
Are three-minute stages enough for LT1 detection?
Methods differ in what they assume about stage length, so treat the tool's documentation as part of your protocol rather than as optional reading. The part fully under your control is consistency: change your stage length and you have started a new series. The old numbers are not wrong, they simply stop being comparable.
How many stages before a fit means anything?
This one is arithmetic, not physiology. A curve with three free parameters fitted to four points is not a fit, it is a connection. Give a curve-based method at least two stages that sit low and flat, and at least two above the point where the rise clearly starts. Below that coverage the curve has nothing to bend around — and it will still print a number.
Can one test give both LT1 and LT2?
Both are read off the same curve, so in principle yes. Whether your particular dataset supports both is the coverage question above: if the test started too high, the flat lower section that a baseline-type rule needs was never measured, and no method recovers data you did not collect.
Parity checklist before you call a home test comparable
- [ ] The protocol is written down before the test: start load, increment, stage length, sampling point, cadence or gradient, and what ends the test.
- [ ] Same machine as last time, or a machine whose load you have checked against a second source.
- [ ] Log rows numbered and pre-filled with stage loads before stage one.
- [ ] Duplicate pairs planned at three stages for the noise band.
- [ ] Pre-test notes recorded: last hard session, sleep, last meal and its time, room temperature, time of day.
- [ ] Meter manual read for sampling and strip-handling instructions, and those instructions copied into the protocol.
- [ ] Analysis method named and fixed before you look at the result.
- [ ] A stopping plan, plus a second person if the test goes to maximal stages.
When to book the supervised session anyway
Screening you have not had. Clearance for maximal exercise is a medical question. Take it to a clinician rather than adjusting the protocol until the test feels manageable.
A report someone else has to accept. If a federation, an employer or a clinician is going to act on the result, ask that recipient first which formats and providers they accept. That is a question for them, not for a calculator.
Load you cannot verify. If the ergometer or treadmill cannot be checked and cannot be kept constant between tests, your watt or pace axis is drifting underneath the lactate values, and every threshold you calculate inherits that drift.
No second person and no workable sampling window. If steps 2 to 7 cost you your form at every stage, a supervised session buys the one thing your setup cannot: someone whose only job during those twenty seconds is the sample.