A screenshot quickly preserves the feeling of a run: distance, time, a map line and a few important measures. Months later, however, your question changes. You want to examine a climb separately, check a sensor dropout or compare exact lap times. The image may then be insufficient because the visible summary does not contain the full underlying timestamped data series.

Keeping the original running file helps with more than changing services. It can support later analysis, correction or a more precise question if the necessary data were recorded. This article follows a simple personal archive: what to save, how to separate converted copies and how to check that the backup actually works. You need not build a professional data-management system, but a few consistent rules can prevent considerable uncertainty.

A screenshot shows results, rather than the full measurement history

Figures on a run card are selected summaries, often rounded. Graph pixels do not necessarily preserve every recorded timestamp, field unit and pause marker either. You therefore cannot reliably reconstruct the same table from the image. Visual presentation and reprocessable data play different roles in an archive.

The Garmin FIT Activity documentation describes timestamped records, lap summaries, sessions and events separately. An activity file may carry several kinds of detail that can be investigated later. Which fields actually exist depends on the device and recording. The original file is therefore not a complete copy of every possible sensor signal, but may preserve more than a few summary pictures.

In our example, two runs share a ten-kilometre distance and one-hour duration. One is steady; the other alternates faster and slower sections. Their summaries may show identical figures while the underlying histories differ. With only the final card remaining, checking that difference in detail is harder. Timestamped data let you return to the same run later with another perspective.

What does an original file mean in your own data path?

Here, original means a file retained unchanged as a source from the device or a service’s original export. It is not necessarily a raw, unprocessed sensor signal. A watch may calculate and filter data before recording. Part of an archive’s value is clarity about where and how you obtained the file.

The official Strava export information distinguishes exporting the original format from downloading GPX or TCX. The original type depends on the file uploaded; it is not always FIT. Preserving the source format matters for later detailed reanalysis. A route converted to another format may not contain every original field.

Our archive note might therefore say “FIT copied from watch” or “service’s original export, downloaded on this date”. These descriptions are not interchangeable guarantees that every byte and earlier step is identical. The aim is documented provenance. If a field’s contents are questioned later, the short note helps establish whether we are examining the device recording or an intermediary service’s export.

Give original, converted and corrected copies separate roles

Do not overwrite the original with a manual correction. Make a working copy whose name or folder identifies the conversion. In one simple system, “original”, “export” and “corrected” collections are separate. Folder names have no magic power; what matters is being able to return to the source before changes and understand which copy produced an analysis.

In our example, a programme modifies a location point believed to be wrong during a city run. The corrected file may yield a new distance and pace, but the original remains. A short description records which programme made which change and when. “A prettier map” alone is insufficient documentation. The assumption behind the change and its effect should remain understandable later.

CSV, GPX and screenshots can serve different purposes. CSV offers a convenient analysis table, GPX can transfer a route, and an image provides an accessible memory. These can accompany the original rather than necessarily replace it. A good archive does not depend on having many copies of everything, but on clear roles and a traceable source.

A female runner reviews a notebook beside a café window, looking for patterns in her running log.Enlarge image
AI-generated editorial illustration of reviewing a running log; not a documentary photograph.

A few metadata details may explain a jump years later

Device model, sensor and software version can help explain differences between two periods. FIT documentation allows several such details to be stored in Device Info messages, but not every field is present in every file. Your own notes can therefore supplement the recording. Dates of device changes, major updates and important configuration changes are particularly useful.

In our example, you use wrist-based heart rate until September, then a new chest strap from October. If only an annual average-heart-rate graph remains, a difference may easily look like a training change. Recording the sensor switch preserves another possible explanation. It does not prove every difference is technical, but prevents forgetting an important data-path change.

Keep the essential environmental details that are genuinely known too: indoor or outdoor mode, an unusual stop, getting lost or a recording error. Do not invent weather and sensations later to fill every row. “Not recorded” is more accurate than confident reconstruction from memory. An archive’s reliability also depends on honestly marking uncertainty.

A custom sensor field may become visible again later

The official FIT developer-field guide explains how custom data can enter a file with its own description and unit. A receiving interface may not display every such field. If the value really exists in the original, an appropriate processor may still examine it later. Missing from the display and missing from the file are different situations.

In our example, today’s log displays pace and heart rate but not a sensor’s custom field. A year later, that field may be exactly what another question needs. Retaining the original offers a chance to investigate if it was actually recorded. An export limited to map points cannot reliably restore the custom sensor number. The opportunity comes from preserved contents, rather than the cleverness of a new programme.

Reanalysis does not necessarily produce a better truth either. A new algorithm may summarise the same file differently using other assumptions. Distinguish old and new results by version and method. The archive’s main advantage is checkability: you can see what produced a conclusion and how. Retaining a source does not guarantee every later estimate’s accuracy.

A simple, human-searchable system may suffice

In our example, a year–month collection gives every run a date, short name and unchanged file type. Recordings on the same day can have a start time or unambiguous sequence number. Alongside the name, a short index records the source and important note. The aim is to avoid guessing between fifteen files named “activity_final” when a later question arises.

Interpret dates consistently: a run’s local day and the file’s UTC timestamp may differ near midnight. Do not alter the original timestamp merely for your own naming scheme. A filename is a human label; internal time information belongs to the recording. Keeping them separate preserves the original reference for later timezone checks.

Track duplication too. A copy downloaded from a watch and one re-exported from a log may be two versions of the same run without necessarily having identical contents. Do not delete one merely because filenames look similar. First clarify provenance, size and each copy’s role. Tidying is useful when it does not discard a source for which another is an incomplete substitute.

A backup is checked only after reading it back

The Canadian Centre for Cyber Security’s official backup guidance discusses regular backups, testing recovery and protecting sensitive data. A sports log is a smaller personal collection, but the principle applies. Seeing a filename on external storage does not establish that the contents are intact, readable and usable.

In our simple trial, we copy several files from an earlier month back from the backup into a separate checking folder. We check that they really concern the expected runs, have the correct start and distance, and open in the appropriate programme. We do this without deleting the sources. The trial need not mean manually analysing every file, but should include actual reading back rather than only a successful-copy message.

In a more advanced personal archive, a checksum can help detect whether a file’s bytes have changed. This does not prove a measurement’s physiological accuracy or replace a sound provenance description. Checking that a file is unchanged differs from assessing the data’s scientific credibility. Keeping those questions separate leaves the checksum as a useful technical reference without overstating its evidential role.

A runner adds a separate note about an uncertain measurement while reviewing training-log data.Enlarge image
AI-generated editorial illustration of traceably marking suspicious sensor data; not a documentary photograph.

Route data need protection in your own archive too

A running file can reveal a home starting point, regular routes and training times. Hiding locations on a public map does not automatically mean an original file shared later has the same masking. Preserving the full source therefore calls for more responsible sharing. Every technical question does not require publishing the full personal route.

The Canadian guidance also recommends separate copies and appropriate access protection, and discusses encryption of sensitive information and strong account security. Use the actual protections available in your chosen personal-archive device and service. Two folders on the same drive of one laptop do not protect against losing the computer. Where copies are stored and who can access them matter too.

A synchronised folder is not necessarily the same as a backup retaining earlier versions. If a faulty change or deletion spreads everywhere connected, several views may conceal the same loss. Check your service’s recovery options and retention rules. The aim is genuinely recoverable, protected data, rather than the largest number of linked copies.

A short personal archiving routine

After running, first check that recording has finished and the intended activity exists. At intervals suited to your own system, save the available original files and add short notes for important events. Put converted or corrected files elsewhere. The archive can then result from small, consistent steps rather than one large tidying session.

Occasionally read back an earlier backup and check that you can still interpret the necessary format. This is particularly useful before changing a service or device. You need not produce new analysis for every possible data field. What matters is that provenance, role and readability remain known beyond the time your current app reminds you of them.

A screenshot is therefore a good memory and quick overview, while the original running file preserves detail for later questions. They complement each other. An unchanged source, separately handled conversions, a few important metadata details and a tested backup together make the archive durable. Your training history can remain an understandable personal data collection rather than merely today’s interface display.

Sources

Helpful? Pass it on.