Intervals.icu can interest several kinds of runner: someone reviewing repetitions in an interval workout, someone using a shared calendar with a coach, or someone seeking to understand longer-term training-load changes. Its views help when you know where the Load number comes from and distinguish planned tasks from actually recorded movement.
This guide draws on official product pages and the developer’s own explanations checked on 21 November 2026; it is not an app test. It introduces calendars, intervals and model inputs through specific analysis situations. Calculations illustrate settings’ meanings rather than recommend individual workloads.
1. Check data direction when making the first connection
The current integrations page lists connections with Garmin, Polar, Suunto, COROS, Amazfit and other platforms; receiving activity and wellness data and sending planned workouts to devices are separate possibilities. A supported brand does not mean identical two-way behaviour for every data type and model. Check what arrives from your watch and which tasks can be sent back.
After importing, first find a workout you know well. Are total time, distance, heart rate or power, and important sections visible? If something is missing, analysis cannot produce a measurement absent from the original file. A manual note can usefully supplement it, but does not turn a missing sensor series into an actual measurement.
Your data-path note might say “watch → manufacturer app → Intervals.icu”. If you also use Strava or another intermediary, draw a separate arrow. This helps locate faults: a run missing despite being correct in the manufacturer’s log differs from a recording not yet transferred from the watch. Start with one clear, working path rather than many parallel connections.
2. Interval detection: inspect what it selected
The Activity Timeline offers automatic interval detection, detection and correction of heart-rate and power spikes, and detailed interval data. These analyses derive from the recorded trace. A section identified by the program may not match what you mentally considered one work interval, especially with interruptions, turns or noisy data.
In our interval example, you run three longer straights, walking back to the start between them. Use the map and pace trace to check whether the selected work interval really covers its start to finish. If a brief GPS jump split it, do not draw separate training conclusions from the two tiny parts’ averages. Clarify selection and the event’s sequence first.
Use the same interval definition for comparisons. If one repeat includes the slow lead-in while another includes only the steady middle, their average paces answer different questions. Processing has value when you can apply it at the next similar session too. One attractive table is insufficient if its component sections cannot be traced.
3. Calendar: a plan and completed activity are different records
The Training Calendar brings planned and completed workouts, weekly totals, a workout library and device sync together. Dragging a task to another day changes scheduling. It does not change when a previous run actually happened or what it contained. If similar planned and completed entries appear, identify their status first.
Our example: a track run was planned for Thursday, but you got out only on Friday. The difference between Friday’s recording and Thursday’s plan is not necessarily data loss. Check their days and sports. Calendar organisation should bring actual and planned events together rather than erase the fact that life changed the week.
You can add travel and known interruptions to the weekly overview as personal notes. A log becomes more precise when a missed outing leaves more than a blank field. Later, it can explain whether a lower-volume week reflected planned rest, work pressure or another circumstance. The same weekly distance can conceal different stories.
Enlarge image4. Workout Builder: a target’s value depends on your threshold settings
The Workout Builder supports structured intervals through a text editor. Targets can use power, heart rate, pace or cadence; pace can reference a threshold percentage or zone, and heart rate can reference maximum or threshold. FIT, ZWO, MRC and ERG import/export are also available. A supported format alone does not guarantee every target type on every receiving watch.
Our controls example: 6 introductory minutes, three repetitions of 4 minutes of work and 2 minutes of recovery, and 5 minutes to finish. This totals 6 + 3 × 6 + 5 = 29 minutes. If the group omits recovery after the third effort, it is 27 minutes. Check repetition logic before entering target pace. Copying structure accurately differs from selecting an appropriate intensity for you.
For a percentage target, check the reference value too. Changing threshold pace or threshold power can make the same percentage mean a different actual target. The workout’s name may stay unchanged. Check interval values on the version arriving at your watch, rather than the title alone. For a coach’s programme, clarify shared understanding of thresholds and zones before execution.
5. Load: which data series does the system use?
According to the developer’s own explanation, a future workout’s load is calculated from its selected target, while Training Load Priority determines source order for completed activities. For example, if heart rate comes first but is absent, the next available source may take precedence. Planned and actual Load can therefore derive from different inputs.
The current Fitness Chart overview lists power-, heart-rate-, pace- and RPE-based models. These are not identical measurement procedures. A load number makes sense only with its method: faulty heart-rate data, changing pace estimates or your effort rating may affect it differently. “Fifty Load” alone says less than “fifty Load from the configured heart-rate model”.
For your own data-review task, select three runs: a continuous one, one with short intervals and one with faulty heart-rate recording. Check each load’s source. The aim is not to choose whichever method gives the most attractive summary each time, but to have consistent, understandable inputs. If you change method, include the change date in your log.
6. Fitness and Fatigue: the model follows at different speeds
The Fitness Chart uses CTL, ATL and Form to describe longer- and shorter-term load, with adjustable time constants. The developer’s calculation explanation shows exponential weighting: the new value combines the previous state and today’s load. Its stated defaults are forty-two days for Fitness and seven for Fatigue. This is mathematical memory, not two precisely measured physiological processes.
In our calculation, both previous values are 50 and today’s load is 80. Using the described weighting with 42/7-day parameters, the longer-term measure becomes approximately 50.7 and the shorter-term one approximately 54.0. The same new day thus appears more quickly in the shorter-memory curve. This illustrates model behaviour; your body does not “store” fitness or fatigue in these points.
A higher Fitness value is therefore not a goal in itself. Input quality, mixing sports and completeness of history matter too. A new account containing only a few days may start its curve from a different background than an old, complete log. Resolve your own data continuity and settings first rather than pursuing another runner’s score with different recording.
7. Changing sports and reanalysis: retain the reason for changes
The Multisport support provides separate sport-specific zones and calculations, including running-pace and cycling-power settings. Sports are not interchangeable merely because each has time and load points. A cycling-filled week and running-filled week can involve different activity despite identical totals.
The developer’s reanalysis response notes that previous activities need reanalysis to receive new Load values; retaining existing intervals is a separate choice. Check a setting change first on a few familiar recordings. Before rewriting extensive history, be clear what you want to change and which manually organised intervals to preserve.
Percentage compliance with a plan is not a complete technical assessment of workout quality either. According to the developer’s specific example, compliance compares planned and actual load; it does not prove perfect execution of the whole interval sequence and intensity. Open the details in your own review too. Two runs with identical overall load may involve a longer steady section in one and many short interruptions in the other.
Enlarge image8. Free core and Supporter: Connector belongs here too
The current pricing table lists Fitness/Fatigue/Form tracking, activity analysis, interval detection, calendar, workout builder, multisport and custom charts among free features. The Supporter list includes, for example, an annual-plan builder, importing full Strava history, route matching and advanced weather analysis. The page currently names four US dollars per month; check your final payment currency and conditions when buying.
The Intervals MCP Connector is currently part of Supporter rather than listed as an additional Intervals charge. It connects your account to the AI assistant you already use; that assistant’s access and pricing conditions are separate. The provider emphasises that the assistant interprets Intervals numbers, so two systems can formulate different conclusions from the same week.
For AI connectivity, review authorised activity, wellness, calendar, settings and library data too. The Connector description says Strava-sourced activities provide the AI only with dates; direct device connections matter for detailed data. Routes transfer on separate request, and the connection can be revoked. Read the change list shown before edits; asking for analysis and writing next week into a calendar are different operations.
9. Three separate questions for a working weekly analysis
First check log completeness: did runs arrive, are there parallel records, and do known errors have notes? Second, examine intervals in individual key sessions. Third, view the longer curve with knowledge of its method and sport settings. This order helps you catch a faulty outlying input before building a whole week’s explanation around it.
If using an AI assistant too, first ask a traceable question: which days formed the week, which run lacked heart rate, what changed in the settings? Confident text based on incorrectly imported data helps less than an accurate list of omissions. Add how you felt as brief, specific notes, rather than asking for a health-suitability verdict from a curve’s colour.
Intervals.icu’s strength is a detailed, adaptable workspace: calendars, intervals and different load views can form one story. Consistent data paths and understood settings provide usability. Knowing which recording you are analysing, which target you are comparing with and how Load is calculated turns many graphs into a few genuinely answerable running questions.
Sources
- Intervals.icu – current plans
- Intervals.icu – data connections
- Intervals.icu – activity timeline and intervals
- Intervals.icu – calendar
- Intervals.icu – workout builder
- Intervals.icu – Fitness Chart
- Intervals.icu – multisport settings
- Developer explanation – Load priority
- Developer explanation – Fitness/Fatigue weighting
- Developer explanation – reanalysis
- Intervals.icu – official MCP Connector
- Developer explanation – plan compliance
How Futószív articles are prepared: our editorial guidelines.


