Garmin Daily Suggested Workouts: How the Algorithm Works

The Garmin/Firstbeat Daily Suggested Workouts (DSW) algorithm recalculates your recommended training load every day based on heart rate variability, sleep, and training stress. Longevity Resort programmes are built around it.

Updated 2026-08-28 · Sample procedures Luxury Spa & Medical Wellness Hotel Prezident

Contents

Why the platform needs this breakdown
Conceptual architecture and functional purpose of the algorithm
Input biometric and workload parameters of the system
EPOC calculation model and aerobic/anaerobic Training Effect
Modelling and calculating VO₂max
Evidence base: HRV-guided training
Workout taxonomy and session types
Target Metric Selection: Pace vs Heart Rate
Comparison of Garmin's Adaptive Training Approaches
Algorithm limits, systemic vulnerabilities, and edge cases
Conclusion
What the algorithm cannot know: the Karlovy Vary medical tradition
How this comes together in the platform
What this brings to the hotel and investors
Sources
Further reading

Why the platform needs this breakdown

The foundation of the Longevity Resort programmes is displaying the data recorded by the Garmin watch and scales, and processing it with the Daily Suggested Workouts (DSW) algorithm from Garmin/Firstbeat. The guest wears the watch and steps on the scales; the portal shows sleep, heart rate, heart rate variability, stress, Body Battery, weight and body composition (what the watch measures), while the algorithm converts these streams into a daily readiness score and a safe exercise load.

We did not rewrite physiology from scratch: under the hood are Firstbeat mathematical models backed by patents and clinical studies. The breakdown below is given in full to show the foundation the platform is built on: what exactly the algorithm calculates, from which data, where its strengths lie, and where its limits are.

One caveat about applicability. DSW was created for runners and triathletes, and a spa guest does not need its running mesocycles. The platform takes its physiological core — recovery assessment based on heart rate variability and sleep, accumulated load, daily readiness — and applies it to what the guest actually does at the resort: walks along marked spa trails, procedures, and the drinking cure.

Conceptual architecture and functional purpose of the algorithm

The Daily Suggested Workouts (DSW) feature, integrated into the Garmin ecosystem starting with the Forerunner 745, is an adaptive expert system for training micro-periodisation [1]. The algorithm's computational core is based on developments by Firstbeat Analytics, specialising in the mathematical modelling of the autonomic nervous system and cardiorespiratory physiology [2].

Unlike traditional calendar programmes with a rigid, fixed structure, DSW operates as a closed-loop feedback system [1]. Every 24 hours, the system recalculates the training plan based on a dynamic comparison of the athlete's current metabolic status, cumulative training stress, and overnight recovery biomarkers [1].

The main purpose of the algorithm is to optimize cardiorespiratory performance (maximum oxygen uptake VO₂max, lactate clearance rate, and neuromuscular endurance) while minimizing the risks of overtraining and orthopedic injuries [5]. The system distributes training stimuli according to polarized training principles, balancing low-intensity aerobic volume with threshold and high-intensity anaerobic intervals [8].

When a specific target event is added to the calendar, DSW automatically shifts to a cyclic periodisation model, structuring base, specific endurance, peak load, and pre-race tapering phases [2].

Input biometric and workload parameters of the system

The DSW computing pipeline aggregates diverse streams of physiological and telemetry data gathered from continuous 24-hour monitoring and active workout sessions [1]. The accuracy of the generated recommendations depends directly on uninterrupted biosignal recording and correct baseline calibration of the user profile [1].

Input data category	Recorded metrics and parameters	Physiological interpretation and role in the DSW algorithm
Acute and chronic load	Acute load (7 days) and chronic load (28 days) [1]	Acute-to-chronic workload ratio (ACWR) assessment; determines the cumulative capacity of allowable training volume for the current day [1].
Load distribution balance	Training load focus (Load Focus): low aerobic, high aerobic, anaerobic capacity [6]	Identifying systemic deficits in energy pathways and selecting the target stimulus focus for the session [8].
Autonomic status	Heart rate variability (HRV Status), daily RMSSD, and 7-day baseline range [1]	Assessment of vagal tone and sympathovagal balance; blocks high-intensity conditioning during vagal suppression [1].
Recovery kinetics	Sleep quality and architecture (Sleep Score), daily stress level, Training Readiness index [1]	Determining the range of the adaptive reserve; reducing the planned pace or duration if recovery is incomplete [1].
Residual metabolic debt	Full recovery time (Recovery Time in hours), residual EPOC [1, 8, 20]	Prevents compounding metabolic acidosis and glycogen depletion before homeostatic equilibrium is reached [1].
Current functional threshold	VO₂max value, lactate threshold (LTHR, LT pace), functional threshold power (FTP) [1]	Serves as the computational baseline for generating target zones for pace (min/km), heart rate (bpm), and power (W) [1].
Individual bioprofile	Maximum heart rate (HRmax), resting heart rate (HRrest), activity class (Activity Class), anthropometrics [14]	Scales non-linear energy expenditure curves and internal relative intensity scales [14].
Calendar context	Target race date, event distance, elevation profile of the course [2]	Modulates mesocycles and distributes training volume to match the specific demands of the target event [2].

EPOC calculation model and aerobic/anaerobic Training Effect

The core metric for quantifying training load in Firstbeat architecture is an indirect mathematical estimate of excess post-exercise oxygen consumption (EPOC) [11]. Unlike traditional ergospirometry, the Firstbeat method calculates EPOC accumulation in real time from continuous heart rate dynamics, current percentage of oxygen uptake (%VO₂max), and ventilatory parameters (patents US7717827B2, US2006/0032315, US10238915B2) [15].

Instantaneous pulmonary ventilation and respiration rate are determined non-invasively through spectral analysis of respiratory sinus arrhythmia (RSA), which modulates beat-to-beat intervals in the high-frequency (HF) band of heart rate variability [23]. The accumulation and clearance of metabolic debt over time is described by a differential equation:

dEPOC/dt = φ(HR(t), Respiration Rate(t), Activity Class) − λ · EPOC(t)

where the function φ represents the rate of oxygen debt generation at a given intensity, and the term λ · EPOC(t) accounts for continuous metabolic clearance during exercise [15].

The peak EPOC value recorded during a session is scaled into an Aerobic Training Effect score on a non-linear scale from 0.0 to 5.0, where 2.0–2.9 indicates maintaining aerobic fitness, 3.0–4.9 provides an improving stimulus, and 5.0 warns of potential functional overreaching [11].

Anaerobic Training Effect is computed by a parallel pattern-recognition algorithm [27]. The algorithm continuously tracks sharp surges in pace or power, comparing them against cardiovascular response and recovery interval durations [27]. A steady run at lactate threshold, despite its high metabolic stress, is logged as a high-aerobic stimulus [31]. Generating an anaerobic effect requires discrete near-maximal or sprint efforts separated by recovery intervals, driving intramuscular phosphagen depletion (ATP-CP) and substantial glycolytic flux [27].

The overall data flow runs through a sequential processing pipeline. Raw heart rate and RR-interval signals from the sensor split into two parallel computational streams: spectral analysis of HRV to reconstruct respiration, and evaluating mechanical motion output against cardiodynamics [23]. These metrics merge into the dynamic EPOC calculation engine, whose output feeds the aerobic and anaerobic Training Effect calculation modules [11]. The resulting load vector categorises the session within Load Focus, assigning it to the low aerobic, high aerobic, or anaerobic category [11].

Modelling and calculating VO₂max

Calculating aerobic power in Garmin algorithms relies on the linear relationship between movement speed, mechanical power, and the oxygen cost of locomotion (patent US20110040193A1) [23]. For every valid running segment, the system determines the theoretical oxygen demand VO₂theor, adjusted for terrain gradient:

VO₂theor = 12 · v + 54 · tan(arcsin(Δh/d)) + c

where v is running speed in metres per second, Δh/d is the sine of the surface slope angle computed from barometric altimeter and GPS data, and c is the baseline resting metabolic cost [33].

During steady-state running segments (once transient cardiovascular adjustments settle), the calculated VO₂theor is matched against the current fraction of maximum heart rate (%HRmax) [2]. Linearly extrapolating this ratio to 100% HRmax calculates individual VO₂max capacity without requiring an exhaustive ramp test to exhaustion [2]. In laboratory validation, the mean absolute percentage error (MAPE) is 5–8%, though under field conditions with unstable optical heart rate readings, error can rise to 8–10% [6].

Evidence base: HRV-guided training

The adaptive logic in DSW draws on clinical and applied research in heart rate variability-guided training management (HRV-guided training) [17]. The core evidence foundation rests on studies by Kiviniemi et al. (2007, 2010) and confirming meta-analyses of autoregulated training protocols [17].

In the seminal study by Kiviniemi et al. (2007), researchers compared a rigid predefined training programme (Predefined Training) against a protocol adapted from morning heart rate variability values (HRV-Guided Training) [18]. In the adaptive group, high-intensity development intervals were prescribed only when RMSSD fell within or above the individual's baseline range [17]. When RMSSD dropped below the baseline confidence interval—indicating sympathetic dominance and incomplete recovery—high-intensity work was postponed in favour of a low-intensity run or passive rest [17].

The HRV-guided group achieved a statistically significant advantage in maximal aerobic speed gains (+0.9 km/h vs +0.5 km/h) and greater VO₂max improvement despite doing fewer hard sessions overall [17]. The DSW algorithm implements these physiological principles through dynamic Training Readiness: autonomic suppression automatically steps scheduled high-intensity sessions down to base or recovery runs [1].

Workout taxonomy and session types

The workouts generated by the algorithm fall into distinct functional categories, each addressing a specific biochemical and adaptation target to balance the Training Load Focus profile [6].

Workout type	Target physiological mechanism	Metabolic and substrate profile	Training effect vector	Share in base mesocycle
Recovery	Active recovery, metabolite clearance [8]	Capillary perfusion without fibre microtrauma; intensity below the base zone [21].	Low Aerobic (0.0–2.0 TE) [11]	10–20 %
Base	Mitochondrial proliferation, angiogenesis [8]	Predominant oxidation of free fatty acids; intensity at 70–80 % HRmax [21].	Low Aerobic (2.0–3.5 TE) [11]	50–70 %
Long Run	Increased intramuscular glycogen capacity [5]	Musculoskeletal adaptation to sustained mechanical stress; substrate shift towards lipids [5].	Low Aerobic (3.0–4.5 TE) [11]	15–25 %
Tempo	Increasing power output at the first ventilatory threshold (VT1) [39]	Mixed aerobic-glycolytic energy metabolism, lactate clearance; intensity at 80–85 % HRmax [21].	High Aerobic (3.0–4.0 TE) [14]	5–15 %
Threshold	Shifting the second lactate threshold (LT2 / MLSS) [39]	Maximal blood lactate steady state; working at 85–90 % HRmax [21].	High Aerobic (3.5–4.8 TE) [11]	5–10 %
VO₂max	Maximising cardiac stroke volume and pulmonary diffusion [6]	Peak cardiorespiratory load, submaximal acidosis; operating at 95–100 % VO₂max [8].	High Aerobic (4.0–5.0 TE) [11]	5–10 %
Sprint / Anaerobic	Neuromuscular stimulation, recruitment of fast-twitch motor units [8]	ATP resynthesis via phosphocreatine and anaerobic glycolysis; short bouts of maximum-effort sprints [31].	Anaerobic (2.0–4.5 TE) [11]	3–8 %

Target Metric Selection: Pace vs Heart Rate

You can choose your Target Type: by pace or by heart rate [42]. This choice determines how the algorithm manages workout intensity in real time:

Targeting heart rate (Heart Rate Target) compensates for external environmental factors—headwinds, temperature shifts, elevation changes, and accumulated stress [46]. The algorithm keeps the physiological cost within the prescribed zone, preventing you from drifting from base into threshold when running uphill [9].

Targeting pace (Pace Target) uses mechanical running speed, calculated directly from your current VO₂max and critical speed profile [1]. This mode is ideal for race preparation on flat courses, but carries a risk of metabolic overload in challenging weather or high temperatures [2].

Firstbeat's internal classification logic is strictly tied to percentages of maximum heart rate (%HRmax) and lactate threshold, bypassing any custom heart rate zone displays set in the Garmin Connect app (such as %HRR zones based on heart rate reserve) [14]. Because of this, the actual load breakdown in the Training Load Focus widget is based on Firstbeat's physiological constants rather than the visual zone scale on your watch [21].

Comparison of Garmin's Adaptive Training Approaches

The Garmin platform includes several independent workout planning engines that differ fundamentally in algorithmic flexibility and planning horizons [1].

Parameter	Daily Suggested Workouts (DSW)	Classic Garmin Coach (Jeff, Greg, and Amy plans)	Garmin Run Coach (Updated adaptive architecture)
Planning horizon	Rolling 7-day window with daily micro-adjustments [1]	Static 6–26 week macro-plan with a fixed schedule [1]	Long-term dynamic plan targeted at a specific race date [45]
Recovery sensitivity	Immediate: reacts to overnight HRV trends, sleep, and cumulative EPOC [1, 5, 6]	Minimal: adjusts only based on completed workout pace and performance [49]	High: combines a macrocycle plan with daily readiness analysis [50]
Operational flexibility	Fully autonomous: automatically recalculates if workouts are missed or rescheduled [1]	Low: skipping sessions reduces the system's estimated Confidence Score [1]	High: algorithmic redistribution of key sessions across the week [50]
Event periodisation	Automatic (Base → Build → Peak → Taper) once a race is scheduled in the calendar [2]	Dedicated templates for a limited set of distances (5K, 10K, 21.1K) [49]	Comprehensive macro-periodisation for any running distance, including the marathon [50]
Optimal athlete profile	Experienced runners, triathletes, and athletes with irregular work schedules [1]	Beginner athletes who need step-by-step written guidance and motivation [1]	Athletes of any level preparing for official events with a target finish time [50]

Algorithm limits, systemic vulnerabilities, and edge cases

A thorough review of long-term DSW use during real-world training reveals several physiological and cybernetic limits:

The first critical factor is a downward positive feedback loop (the fading stimulus effect) [10]. Because recommended session duration and intensity depend on current readiness and autonomic state, non-training stress (sleep deprivation, systemic inflammation, emotional strain) persistently lowers HRV and elevates resting heart rate [1]. The algorithm responds by steadily reducing volume and intensity, steering the athlete exclusively into recovery and base runs [5]. An extended absence of productive stimuli leads to genuine cardiorespiratory detraining and a dropping VO₂max, which the system misinterprets as a signal to scale back workload even further [10].

The second source of systemic error is the latency and optical noise of photoplethysmography (PPG) sensors [1]. Accurately calculating anaerobic training effect and EPOC accumulation kinetics requires instant logging of sharp heart rate spikes and precise RR-interval timing during sprint accelerations [23]. Optical sensor lag across sharp workload gradients and the risk of cadence lock lead to an underestimation of the anaerobic stimulus [1]. As a result, the system registers a false chronic deficit in the Anaerobic segment of Load Focus and keeps prescribing sprint protocols regardless of the athlete's actual physiological fatigue [14].

The third functional limit is the conservative long-run cap built into the algorithm [5]. To reduce the risk of overuse injuries to the periosteum and tendons, the system rarely generates continuous low-intensity runs lasting more than 100–110 minutes [5]. For runners training for a marathon with a target time over 3.5 hours, this volume is insufficient to build complete enzymatic adaptation for lipolysis and neuromuscular resilience to extended impact stress [5].

The fourth issue is the misinterpretation of unstructured cross-training and strength workouts [52]. Gym strength sessions cause local muscle fatigue and mechanical microdamage to fibres while generating only moderate cardiovascular strain [20]. By relying primarily on heart-rate-derived EPOC, the DSW algorithm either underestimates peripheral muscle fatigue from strength work or mistakenly logs strength EPOC into the general acute aerobic load pool, cancelling key running sessions without good reason [15].

Conclusion

Garmin and Firstbeat's Daily Suggested Workouts algorithm is a physiologically grounded software system for automated training stress management, successfully bringing the fundamental principles of polarised periodisation and HRV-guided autoregulation to mainstream wearable devices [8]. Its mathematical framework of continuous EPOC modelling, training effect differentiation, and extrapolated VO₂max estimation provides a solid foundation for maintaining overall cardiorespiratory fitness [1].

Even so, the reliability and validity of the system's prescriptions are strictly limited by input accuracy. Preventing algorithmic glitches and avoiding downward detraining loops requires regular data collection via a chest strap heart rate monitor, accurate empirical or lab verification of maximum heart rate, and manual adjustments to long-run volume when preparing for marathon and ultramarathon distances [1].

What the algorithm cannot know: the Karlovy Vary medical tradition

The conclusion above candidly points out the system's limits: downward detraining loops, optical sensor noise, undercounted strength work, and conservative caps on long workouts. A spa hotel guest has even more variables. The algorithm knows nothing of age, comorbidities, or medications; it cannot tell if someone develops a spa reaction on day three of their drinking cure, or that they underwent surgery six months ago. For an athlete, these simplifications are acceptable. For a spa guest, they are not.

This is where the resort's medical tradition comes in — something no software licence can provide. Longevity Resort programmes at Luxury Spa & Medical Wellness Hotel Prezident were built not by programmers, but by doctors who have observed patient recovery across decades of clinical practice.

MUDr. Milada Sárová — internal medicine and balneology. One of the leading authorities in mineral water spa therapy and Karlovy Vary drinking cures, having lectured on Karlovy Vary balneology in over 20 countries. Across decades of practice, she has overseen thousands of recovery courses for metabolic disorders, digestive tract conditions, and rehabilitation after intensive medical treatments. This expertise defines the platform's indications and contraindications, procedure sequencing, and drinking cure schedules.
Docent Petr Kolisko — physiotherapy and kinesiology. A leading Czech specialist in musculoskeletal functional diagnostics, a pioneer and senior kinesiotaping instructor in the Czech Republic and Central Europe, a longtime lecturer at medical and sports faculties, and the author of research papers and monographs on movement biomechanics. This expertise defines exercise dosing in the platform: what load is safe for a diagnosed muscle imbalance, how to increase volume after an injury or surgery, and where a spa walk ends and overload begins.

How this comes together in the platform

The platform works as an interpreter between two worlds. On one side, it receives objective figures from Garmin: heart rate and its variability, sleep stages, stress, Body Battery, steps, acclimation, weight, and body composition. On the other side stand the resort's medical rules: thresholds for rescheduling a procedure, target heart rate corridors for walks, and the daily sequence of therapies.

The doctor sees a dashboard with every stay metric and adjusts bookings; the guest receives a day-by-day protocol in their own language. The numbers do not prescribe treatment on their own — they give the doctor a clear clinical picture they never had before, and show the guest their results in hard figures rather than general impressions. The platform is classified as Wellness & Lifestyle Intelligence — lifestyle support rather than medical diagnostics (learn more).

What this brings to the hotel and investors

The hotel gains a methodology, not just software. Your own procedures — saunas, baths, massages, mineral springs, and menus — are transferred to the platform, while a daily protocol created by Karlovy Vary medical experts on top of Garmin/Firstbeat research arranges them day by day. Staff are trained on a turn-key tool rather than having to build a programme from scratch (what onboarding includes).
Guests see the cure's effect in clear numbers. Before-and-after metrics for sleep, resting heart rate, HRV, and body composition are what bring premium guests back and give them something to talk about. Recovery stops being a vague feeling and becomes a measurable outcome.
An investor gets a combination that is difficult to replicate. The algorithmic side can be licensed — its physiology is documented in patents and research papers. Decades of clinical observations by spa doctors cannot be licensed. The value of the platform lies right at this intersection: cutting-edge Garmin research on one hand, and the medical school of a world-renowned spa resort on the other.