Build a calibration workset

Use this article when you want deeper guidance than “include representative files.”

A calibration workset is simply the collection of FIT files you choose to let TRIPS learn from for one calibration pass.

There is no single perfect workset.

Different worksets can be valid if they match different planning goals.

The important thing is to understand what each kind of workset is likely to teach TRIPS well, and what it is likely to teach poorly.

What the workset is trying to support

Your workset is doing more than one job.

It needs to provide enough evidence for:

movement-shape review, such as how speed changes across grade
heart-rate review, such as whether strain behaves consistently across terrain
effort-intent interpretation, so easier and harder outings separate clearly

That means a workset can be strong for one job and weak for another.

For example:

a set with good terrain variety but weak heart-rate coverage may still help shape review, but it will be weaker for effort-intent interpretation
a set with good heart-rate coverage but very little terrain spread may help some effort review, but it will be weaker for grade-shape learning

What a strong workset means

A strong workset does not have to be perfect.

It means the files are relevant, internally consistent, and useful for the type of trips you want to plan.

A strong workset often includes:

the activities are relevant to the trip style you want to plan
the recorded movement is consistent with the activity
the elevation record is consistent with the route
heart-rate coverage is present when you expect to rely on calibration
most files reflect normal outings rather than rare oddities
the workset is organized enough that you know why each file is there

What a weak workset means

A weak workset is one where too much of the evidence points in different directions, or too much of the evidence is incomplete.

A weak workset often includes:

unrelated sports are mixed in heavily
many files are corrupted, incomplete, or strangely recorded
long stop-heavy sessions dominate the folder
the data is technically valid but not representative of your real planning use
converted or exported files preserve the track but lose useful effort signal
one unusual trip type overwhelms the rest of the set

Messy does not always mean unusable.

It means you should expect weaker or less stable calibration behavior and be more cautious about saving the result as a reusable profile.

The main dimensions of workset quality

When deciding what belongs in the folder, it helps to judge each file or subgroup on a few separate dimensions instead of thinking only in terms of “good” or “bad.”

Relevance

Ask whether the activity belongs to the same world as the trips you want TRIPS to help plan.

More relevant:

hiking
backpacking
mountain travel
loaded uphill and downhill travel
outings in terrain and altitude regimes similar to your planned trips

Less relevant:

road running
cycling
gym sessions
neighborhood walks that look nothing like your target use
highly specialized efforts you would never use as a planning reference

Signal completeness

Ask whether the file contains enough usable information to support calibration.

Higher-value files often have:

usable timing and movement data
usable elevation change
heart-rate coverage for much of the moving portion

Lower-value files often have:

missing or sparse heart rate
partial recordings
odd timestamp behavior
synthetic-looking fields created during conversion

Representativeness

Ask whether the activity reflects how you move on the kinds of trips you want to plan.

A file can be real and clean but still not be representative.

Examples:

a very fast day hike may be a real file, but it may not represent your loaded backpacking pace
a high-altitude mountain outing may be relevant for one planning goal but not for lowland training routes

Coverage

Ask whether the full workset spans enough variation to teach something useful.

Useful coverage often includes:

a range of grades
a mix of easier and harder outings
some variation in duration, load, terrain, and altitude

Too little coverage can make the workset look tidy while still leaving TRIPS with too little evidence to generalize from.

Consistency

Ask whether the files mostly reflect one consistent activity pattern.

A workset is more consistent when most files share:

similar activity type
similar device behavior
similar definitions of “normal effort”
similar planning intent

Some variation is helpful.

Too much variation without a clear reason can blur both the shape signal and the effort-intent signal.

Common workset styles and their tradeoffs

No single style is always right.

These are common workset patterns, with the main upside and downside of each.

Small, clean starter set

This is a modest set of obviously relevant hikes you trust.

Strengths:

easy to reason about
lower risk of unrelated noise
good first pass when you are unsure what belongs

Limitations:

may be too narrow to support strong shape review
may not include enough intensity range for effort-intent interpretation
can over-reflect a small slice of your hiking life

Broad but still representative hiking set

This is the default target for most users.

It includes many relevant hikes and backpacking outings, but still excludes obvious junk.

Strengths:

better terrain and intensity coverage
better odds that both shape and effort-intent review are informative
often the best balance between breadth and cleanliness

Limitations:

still needs deliberate pruning
can get diluted if it quietly accumulates too many borderline files

Large mixed archive

This is the “export everything and see what happens” approach.

Strengths:

often contains a lot of raw data
may still be useful if most files are from the right activity world

Limitations:

noise can overwhelm relevance
unrelated sports can distort the story
easier and harder effort bands may become less interpretable
technical ingest success can hide a poor calibration foundation

Very narrow trip-specific set

This is a set built around one trip style, season, or use case.

Examples:

loaded backpacking only
alpine day travel only
desert hiking only

Strengths:

may be very relevant for one narrow planning use
can create a consistent result within that niche

Limitations:

may generalize poorly outside that niche
may underrepresent easier or harder efforts outside that use
can be too narrow for a reusable all-purpose profile

Terrain-rich but physiology-light set

This set has good route and grade diversity, but weak heart-rate coverage.

Strengths:

may still help with some movement-shape review

Limitations:

weaker support for effort-intent interpretation
lower confidence when comparing effort bands
easier to mistake route geometry for complete calibration quality

Heart-rate-rich but terrain-narrow set

This set has usable HR coverage, but most files come from a narrow terrain regime.

Strengths:

may support some effort interpretation

Limitations:

weaker grade-shape learning
less confidence that the shape generalizes outside that narrow terrain band

Converted or reconstructed FIT set

This set comes from GPX exports, third-party conversions, or reconstructed FIT files rather than original device FIT recordings.

Strengths:

can still be useful when original FIT files are unavailable
may preserve enough track and timing structure for limited review

Limitations:

often weaker than original device FIT for calibration
heart-rate data may be missing, synthetic, flattened, or absent altogether
pauses, timestamps, sensor fields, and sampling behavior may not reflect the original outing faithfully
effort-intent interpretation is especially vulnerable when the physiological story was not preserved

Converted files are not automatically wrong.

They deserve more skepticism, especially if they ingest cleanly but later calibration results remain weak or unstable.

What to exclude first

If you need to prune quickly, remove the files most likely to distort the story:

unrelated sports
obviously broken or incomplete recordings
sessions with very poor heart-rate coverage
activities dominated by stop time or weird recording behavior
extreme one-offs that do not represent your normal planning use
duplicate or near-duplicate exports of the same outing

What to keep first

If you need a starting point, keep:

hikes and backpacking outings you trust
files with usable movement and elevation data
files with good heart-rate coverage
outings across a useful range of terrain and real effort levels
outings that reflect the routes you most want TRIPS to help plan

How to decide whether a workset is good enough

A good enough workset is not perfect.

It is one that gives you stable enough outputs to continue.

Good enough often means:

the ingest summary looks substantial rather than thin
terrain coverage looks relevant
heart-rate coverage is usable
the grade and HR relationships are consistent enough to review
easier and harder outings are distinguishable

If those conditions are not true, changing the workset is more productive than forcing later controls.

Build order

If you are unsure where to start, use this sequence:

Start with a small clean set of obviously relevant hiking or backpacking files.
Check whether the ingest summary and diagnostics look substantial enough.
If the set looks too thin or narrow, expand with more files from the same activity world.
If the set looks noisy or contradictory, prune aggressively before tuning controls.
Save a reusable profile only after the route preview and effort-intent outputs are stable enough to keep.

The simplest decision rule

When in doubt, ask:

If TRIPS learns from this file, will future plans become more accurate or less accurate for the trips I care about?

If the answer is less accurate, leave it out.

What to read next

For the basic workflow:

For evaluating the resulting dataset: