Your watch showed a step rate around 170, but the downloaded table contains values around 85. At first this looks like a serious measurement error, as though the exporter lost every second step. An exactly twofold difference can, however, have a simple explanation in units. One view may count every right and left foot contact, while another describes a complete right–left cycle.
The solution is not to multiply every low-looking number by two. First identify the field’s meaning, its transformation history and the period being compared. Original calculations show how the same run can become 85 cycles per minute and 170 steps per minute, why fractional values matter and when a completely different explanation should be sought. This is data interpretation, not a recommended running rhythm for everyone.
What are we counting: one contact or a complete repetition?
Imagine a simple sequence: the right foot touches the ground, then the left, then the right again. Between two successive right-side contacts, one complete same-side cycle has passed, with contacts from both feet in between. Counting every alternating step separately gives twice as many events as counting complete right–left repetitions. The relationship follows from the definition, not a difference in fitness.
The official Forerunner 165 data-field description defines running cadence as the total steps per minute from both right and left feet. Cycling cadence on the same page means crank revolutions. The same field name therefore counts different events in different sports. Correct recognition of the app’s sport type is part of interpreting the number too.
Everyday use of “step” is not always consistent either. Someone may mean a single foot contact, another person a complete alternation of both feet, and the English term stride can follow different naming traditions in some products. A clear table therefore states whether it counts all contacts of both feet or complete cycles. The definition attached to the label says more than how familiar the word sounds.
85 in a FIT file is not necessarily an incorrect 170
The verified professional answer on Garmin’s FIT developer forum explains the relationship between FIT cadence and the steps per minute displayed to a runner. For running, the described FIT field’s cycles-per-minute value is doubled to give total steps from both feet, including the fractional part. This is interpretation of a particular format, not a rule for every arbitrary “cadence” column.
In our original example, a decoder writes 85 into the appropriate cycle field at one moment, with a zero fractional part. The display may turn that into 170 steps per minute. If another exporter already outputs the second value, both tables may be correct in their own clearly stated units. The mistake would be labelling both columns identically as steps per minute.
A filename or FIT extension alone does not tell you whether the view shows a raw field, decoded value or user-facing conversion. An intermediate library and exporter may also transform it. Original data, decoding and display are separate layers. When troubleshooting, therefore, record which device produced the data and which programme and version created the table.
Half a cycle can produce an odd number of steps
In Garmin’s developer explanation, fractional_cadence matters partly because doubling the integer cycle value alone would produce only even steps-per-minute values. The described conversion first adds integer cadence and the already appropriately decoded fractional part, then multiplies by two. Half a cycle can therefore correspond to one additional step. Both operation order and field units matter.
In our original calculation, the integer value is 84 and the decoded fractional part 0.5. The sum is 84.5 cycles per minute, doubled to 169 steps per minute. Omitting the fraction gives 168. Doubling 84 first and then adding only 0.5 gives 168.5. Only the first follows the formula specified in this example. This is different processing, not a large physical change.
A raw byte value and an already scaled fraction are different concepts. An exporter may show an immediately interpretable 0.5, while a lower-level view may still display the encoded number. Apply decoding according to the format profile in that case; do not randomly add the raw integer. Programme documentation and the field description determine whether decoding has already occurred.
Enlarge imageThe Bluetooth label does not define cadence uniformly
The Bluetooth SIG GATT supplement dated 9 September 2026 separately describes the RSC Measurement characteristic for transmitting running speed and cadence. Its instantaneous-cadence field uses units of 1/min. In the same document, Indoor Bike Data cadence has a different description, including a field with half-revolution-per-minute resolution. Without the specific service and field name, “Bluetooth cadence” is not a complete technical description.
Decoding a running sensor’s message and displaying running data in a FIT file are therefore not necessarily the same task. An app can create its own internal variable from a standard transmission field and later save it into another format. Follow the whole chain. Two devices being pairable does not prove every exported field is passed on with identical names and units.
In an original fictional data chain, a sensor provides running data equivalent to 170 steps per minute, the app internally stores 85 complete cycles per minute and the view displays 170 again. There is nothing inherently wrong if every transformation is documented and correct. If an intermediate CSV labels the internal 85 as steps per minute, the error is in labelling. Drawing the data chain can identify this faster than replacing the sensor.
An instantaneous value and a workout average differ too
The FIT Activity documentation distinguishes timestamped Record messages, lap summaries and the whole workout’s Session summary. An instantaneous series and a summary average answer different questions. Field availability can also depend on the device, so not every possible cadence field must be sought in every run. First compare at the same level.
In an original simple example, a runner moves for twenty minutes at an average of 170 steps per minute, then stands for five minutes. The moving twenty minutes contain 3400 steps. Dividing by twenty minutes gives 170; dividing by the full twenty-five gives 136. This is not a twofold difference and cannot be explained by a cycle-to-step conversion. The denominator and treatment of stops changed.
A walking warm-up, standing recovery or stairs may enter the summary differently. Excluding zeros and interpreting missing data as zero are different operations too. If the difference is not consistently twofold, examine matching timestamps from the same running section. This helps separate unit conversion from different averaging rules.
A brief video can also provide a checking calculation
For a simple check, review a short running video made by someone else in a safe place afterwards. The runner does not need to watch a phone or table while moving. Ground contacts should be visible, and record the time window precisely. This checks that specific detail, not laboratory sensor certification or a technique rating.
In an original example, we count sixty total right and left contacts over twenty seconds. The rate is 60 × 60 / 20 = 180 steps per minute. Counting only thirty right-foot contacts in the same section gives 90 same-side events per minute. The twofold relationship describes the same alternating-foot movement in two ways. Here, 180 is a chosen example, not a target cadence.
Choosing start and end times can noticeably affect a short window. One additional or missing contact over twenty seconds means three steps per minute. Do not therefore declare a fault from a one- or two-unit difference. A persistent difference between 85 and 170 has a very different scale and calls first for checking definitions.
Step-length checks can also go wrong by a factor of two
Cadence interpretation also matters when calculating step length from speed. In an original example, a runner moves at 3 metres per second with 180 steps per minute, or three steps per second. Dividing gives one metre per alternating step. The same movement has 90 complete cycles per minute, or one and a half cycles per second, giving two metres for a full cycle.
If you mistakenly label the 90-cycle value as steps per minute, you calculate a two-metre “single-step” length. The result fits the given speed mathematically, but the event definition is wrong. This can mislead later technique assessment: a table may claim excessively long steps when it actually assigned the distance of two steps to one. Put the units right first.
Speed and cadence should also use matching time windows. Dividing instantaneous speed by a whole workout’s average cadence produces a ratio that may describe no actual step. The two fields can also have different filtering and delays. Meaningful comparison requires a shared section, known units and a consistent event definition.
Enlarge imageHow should you examine a suspect export?
Keep the original file unchanged and make a separate working copy. Note the sport, full field name, unit, programme and whether the value is instantaneous or summarised. If the view says only cadence, find the exporter’s documentation. A missing label is no reason to reinterpret the number by intuition.
Then select a brief, steady running section and inspect the same times in the original system and exported view. In an original worksheet, three pairs might be 84.5 and 169; 85 and 170; 86 and 172. This pattern fits the cycle-to-step relationship well. If pairs instead are 85 and 170, then 85 and 140, later 85 and 180, other processing or different time windows may contribute.
Make the correction a documented derived column with a clear name, such as “steps from both feet per minute, calculated from cycle value”. Do not overwrite the original column. Source and formula then remain verifiable later. If the exporter already converted the value, doubling again would immediately create a suspiciously high number. Apply the transformation once, in the appropriate place.
When is multiplying by two not the solution?
If the value halves only in one brief section, a constant field-definition difference is a less likely explanation. Examine movement in that section, the sensor connection and how missing data is marked. Zero, an empty field and an invalid value are different. A run–walk transition or stop may represent real movement change that must not automatically be “corrected”.
Similar numbers for heart rate and cadence on a graph do not prove the fields were swapped either. They measure different events even if both remain near 170 for a while. Resolve a misnamed column through field identifiers and programme operation, not an appealing numerical match. One curve alone is insufficient to assess the whole data chain.
Exported cadence that is half as large can therefore be perfectly meaningful cycle data. Correct interpretation requires knowing what the field counts, in which units and over which period. Appropriate conversion of the known FIT field, its already decoded fractional part and documented original data together provide a traceable explanation. Assess the runner’s movement only afterwards; a labelling error should not become a mistaken judgement of someone’s technique.
Sources
- Garmin Forerunner 165: running and cycling cadence fields.
- Garmin FIT developer answer: cycles, steps and fractional_cadence.
- Bluetooth SIG GATT supplement, September 2026: RSC Measurement and Indoor Bike Data.
- Garmin FIT Activity: timestamped and summary messages.
How Futószív articles are prepared: our editorial guidelines.


