How to do intervals outside: the road, the hill and the route that fit the workout

At a glance

Why outdoor intervals fail: the road decides

Outdoor intervals fail on the road, not in the legs. Twenty competitive cyclists held the same 20-minute power outside as on an ergometer (284 vs 280 W), but their second-by-second power varied about twice as much outdoors, with a within-rider standard deviation of 69 W against 33 W [Jeffries et al. 2019]. The engine is the same. The road keeps moving the target.

That result has a long history. In 2001, eight trained cyclists rode three 40 km time trials indoors and three on a local course; mean power was not different (303 vs 312 W) and the day-to-day spread was about 2% in both settings [Smith et al. 2001]. A rider can produce threshold power outside as reliably as on a trainer. What the outdoor ride adds is everything around the power: a junction at minute six, a descent that drops the effort to zero, a crosswind that pushes it up by 40 W for a kilometre. Jeffries and colleagues found the extra variability across almost every time scale, from second-to-second surges to minute-long swings [Jeffries et al. 2019].

That is the core of the thesis. Indoors, the trainer holds the watts and the road is irrelevant. Outdoors, the road decides whether the session runs as written. A 4×8-minute threshold session on a road with a set of lights every 2 km is not a 4×8; it is 12 pieces of 2 to 3 minutes with unplanned rests, and the training effect of that is not the same. Whether to keep separate indoor and outdoor FTP numbers is a different question, and the indoor-vs-outdoor FTP guide in this series answers it. This guide is about the road.

So the planning step changes. A rider who trains inside picks a workout. A rider who trains outside picks a workout and a road together, and when they disagree, the road wins. The rest of this guide sorts roads into four shapes, matches each workout to the shape it needs, and gives rough minimum lengths. The minimum lengths are arithmetic, not research: they assume a rider of about 75 kg at 250 W of threshold, who rides about 35 km/h on the flat and 16 km/h on a 6% climb at that power. A lighter or stronger rider needs a longer road; a slower rider needs a shorter one.

The four road types and what each one gives

Almost every road that matters for intervals is one of four shapes: a flat uninterrupted stretch, a steady climb, a rolling road, or a loop with a turnaround. Each one supports some workouts and breaks others. A climb also changes the work itself: in one study, the same riders produced 4-6% more power in a 20-minute uphill time trial than on the flat [Nimmerichter et al. 2012].

The flat uninterrupted stretch is the hardest to find and the most useful. It is a road with no lights, stop signs or junctions you must yield at, long enough to hold the whole rep. Its risk is speed: at threshold on the flat a rider covers about 580 m a minute, so an 8-minute rep needs about 4.5 km of clear road. Its other risk is wind, which turns one direction into a climb and the other into a descent. The flat stretch suits threshold, sweet spot and Zone 2 best, and it is where most riders do VO2 work when they have no hill.

The steady climb is the easiest place to hold a target. Gravity gives a constant load, speed is low, and there are rarely junctions on the way up. Riders tend to produce more power on a grade: the uphill time trials in Nimmerichter's training study were 4.4% higher than the flat ones before training and 6.3% higher after [Nimmerichter et al. 2012]. In a second study, the same power cost about the same oxygen on a 5.1% grade as on a 1.1% road (2,594 vs 2,536 mL/min), although gross efficiency was 1.3 points lower uphill and was strongly affected by cadence [Nimmerichter et al. 2015]. A watt on a climb trains the same engine as a watt on the flat; it is easier to hold, and it feels different.

The rolling road is what most riders have. It breaks a constant target into pieces: the rider pushes on each rise, freewheels on each dip, and the average power hides how little of the rep was at target. A rolling road can still carry a threshold rep if the rider pushes a little harder up each rise and keeps pedalling over each top, which is also the faster way to ride it [Swain 1997]. It carries Zone 2 badly when the dips are long enough to coast.

The loop with a turnaround is the solution when no single road is long enough. It can be an out-and-back on a quiet stretch with a safe place to turn, or a short circuit of 2 to 5 km that the rider repeats. The rep length sets the loop: a 4-minute VO2 rep fits a 2.5 km out-leg, the recovery covers the turn and the return. The test for any loop is that the turnaround sits inside a recovery, never inside a rep.

Matching the workout to the road

Each workout has a road it fits. Threshold and sweet spot need long reps without a stop, so they fit a steady climb or a long flat stretch. VO2 reps fit a repeatable 3-8% hill. Short-short sets such as 30/15s need one quiet stretch for the whole block. Zone 2 needs a flat loop with few stops. The table at the top gives rough minimum lengths.

Threshold is where the road matters most, because the evidence favours long accumulated time at a high but sustainable effort. In 35 cyclists training about six hours a week, 4×8 minutes at 90% of peak heart rate gave the largest gains in VO2peak and power, about 11%, against 4×4 or 4×16 [Seiler et al. 2013]. Eight minutes at threshold is about 2 km of a 6% climb or 4.5 km of flat. A climb that takes 5 minutes cannot hold an 8-minute rep, so the rider either finds a longer climb, rides 2 laps of the climb as a 2-part rep with no rest, or moves the rep to a flat stretch. Sweet spot at 2×20 minutes needs 10 km or more of flat road without a stop, and most riders find that only on a loop.

VO2 max reps are where a hill helps most. At a 9.8% grade, well-trained athletes produced more 1-minute power and more work above critical power than on the flat; 4- and 10-minute power was not significantly different, but the authors judged the two conditions too far apart to call them equivalent [Hovorka et al. 2022]. The practical reading: short hard reps are easier to execute on a grade, and long reps run on either. A 3-8% hill of 1 to 1.5 km holds a 4-minute rep, with the descent as the recovery. A rider with no hill can do the same session on a flat stretch with a safe turnaround; the guide to VO2 max intervals outside in this series covers that case in detail.

Short-short intervals are the most road-sensitive session of all. Ten weeks of 30 seconds hard with 15 seconds easy raised VO2max 8.7% against 2.6% for effort-matched 4×5-minute reps in 16 cyclists [Rønnestad et al. 2015], but 15 seconds is too short to stop, look, or turn around. A 10-rep block of 30/30s covers about 3 km. Sprints need less length but more space: 200-300 m on a flat or slight rise, good sightlines, and a run-out to slow down without braking hard.

Zone 2 looks easy to fit outside and is not. It needs time in zone, and every stop, descent and long coast takes time out of it. Over a long steady ride, heart rate also drifts upward at the same power after the first 10 to 20 minutes as stroke volume falls [Coyle & González-Alonso 2001], and that cardiovascular drift is only readable on a ride with long steady segments. A flat or gently rolling loop with few lights does more for a Zone 2 day than a scenic road with a junction every kilometre. The guide to Zone 2 rides without stops in this series covers how to plan that loop.

Pace outside by a range, not one number

Outside, a single power number is a target the road will not let anyone hold. Outdoor power swings about twice as much as indoor power at the same average [Jeffries et al. 2019], so each rep needs a range: a floor and a ceiling. Hold the range, let the power rise on grades and into wind, and let the lap average take care of itself.

A range is not a softer target. It describes what the body does well and what a head unit can show. A threshold rep prescribed as 95-105% of FTP means about 238-263 W for a 250 W rider. On a flat road that window is tight enough to be work and wide enough to survive a gust. On a climb the rider will sit near the top of it; into a headwind, the same. We prescribe every target as a range for this reason, indoors and outdoors.

Push a little harder up and into the wind, and ease a little on the way down. Modelling studies show this is also the faster way to ride a variable course: on a hilly 10 km time trial, varying power by 10% with the gradient saved about 90 seconds against constant power at the same average [Swain 1997], and a newer model with validated coefficients found larger savings, about 126 seconds [Atkinson et al. 2007]. Atkinson and colleagues also note that how much variation a rider can tolerate is still unclear, which is the point of a ceiling. A surge above the range on a short kicker costs more than it buys.

Two habits keep the range honest. First, watch a 3-second or 10-second average, not the instant reading: instant outdoor power jumps with every pedal stroke and every change in gradient, and chasing it makes the effort jerky. Second, judge the rep by lap average and time in range, not by a single moment. If a rep's lap average lands inside the range and the rider never had to stop, the rep did its job, however untidy the trace looks.

Heart rate is a weaker guide outside for short reps. It lags the effort by a minute or more, so on a 3-minute VO2 rep it is still rising when the rep ends. For reps of 8 minutes or more it is a useful ceiling, especially on hot days. For a rider without a power meter, perceived effort and a heart-rate ceiling are a fair way to run threshold and Zone 2; VO2 and short-short sets are better done by feel and time.

Safety and traffic: what the rep cannot cost

A hard rep narrows attention, and that is the risk outdoors. No study we could find measures how traffic stops change interval quality or crash risk during intervals, so this section is reasoning, not evidence. The rules follow from one point: the rep must never be the reason a rider runs a light, misses a car, or turns in the road.

The reasoning starts with the effort. At VO2 power a rider breathes hard, the head drops, and attention moves inward to the effort and the numbers. That is fine on a closed climb and dangerous at a junction. So reps go on roads with long sightlines and no junctions inside the rep. A rep that meets a light ends at the light: the rider stops, then restarts the rep or counts it short. Never ride through a junction or a light to finish a rep.

The turnaround is the second risk. A U-turn in the middle of a road at the end of a hard rep, tired and looking down, is the most dangerous moment of an out-and-back session. Turn only in the recovery, at a lay-by, a car park, a quiet side road or a roundabout, and look behind before you turn. A loop that needs no turn at all is better.

The hour matters as much as the road. A stretch that is quiet at 6 a.m. on Sunday can be busy at 5 p.m. on Tuesday. Visibility helps: in a controlled trial with 3,845 Danish cyclists, bicycles fitted with permanent daytime running lights had a 19% lower rate of injury crashes over 12 months, from self-reported data [Madsen et al. 2013]. At dusk and at night, reflective material on the ankles and knees lets drivers recognise a cyclist at much longer distances than a reflective vest alone [Wood 2023].

Two more rules follow. Do not run intervals in a group on an open road; the rider at the front sets the pace for everyone and the rider at the back is in the gap. And do not treat a descent as a recovery sprint: the recovery belongs on the way down at an easy spin, with hands near the brakes. A session that cannot meet these rules on the roads available is a session to move, or to take inside.

Finding the road, or taking the session inside

Finding the road is the slow part. Climb databases list climbs, map apps list routes, and a few tools now try to match a route to a workout. When no road fits — dark, wet, windy, or only stop-start streets near home — take the key session inside, where the trainer holds the watts, and keep the easy ride outdoors.

The tools differ in what they know. Climb databases such as PJAMM and climbfinder list climbs with length and gradient, which answers 'is this hill long enough', but not which workout it fits. Garmin ClimbPro and Karoo Climber show the climb profile during the ride. Strava, Komoot and Ride with GPS build routes but do not read the day's workout. Gritmaps, a free beta launched in October 2026 for intervals.icu users, builds loops that line climbs and headwind sections up with the effort blocks of a planned workout and pushes them to Garmin and Wahoo; it is the closest thing available to placing each interval on a road. Suggested Routes in AdaptCycling, a paid feature, offers loops sized to the day's planned time from the rider's own speed, checks them for straight stretches and 3-8% climbs of 200 m or more, drops loops with too much busy road, and sends the loop to a Wahoo ELEMNT, a Hammerhead Karoo or Garmin Connect courses. It picks a loop where the work fits; it does not mark where each rep starts.

Some days, no road fits. Darkness, wet or icy roads, strong gusty crosswinds and rush-hour traffic all break the rules in the previous section. So does a home that sits among lights and junctions, with the nearest quiet road 30 minutes away. On those days the trade is simple: take the session that needs control inside, and keep the session that needs time outside. A 4×8 threshold session or a block of 30/15s runs better on a trainer in ERG mode than on a road that breaks it into pieces. A 2-hour Zone 2 ride runs better outdoors, where the airflow is free and the hours pass faster.

Moving a session inside does not need a Zwift subscription. A Wahoo ELEMNT or Hammerhead Karoo can drive a smart trainer in ERG from a planned workout with no app running. Ride Live, in beta on the paid plan, runs the day's planned session on a Bluetooth smart trainer from Chrome or Edge on a laptop or desktop. Zwift and TrainerRoad run structured sessions very well too, and Zwift remains the best indoor riding experience. The indoor cycling training guide in this series covers how to make those sessions work: fan, heat, and how much is enough.

The last piece is the plan itself. A week that loses its threshold day to rain should not simply lose the stimulus. The plan should move the session to the next suitable day, or swap it with an easier day, and keep the rest of the week coherent. That is what our plan does after a missed or moved day: it reads the ride that actually happened and rebuilds the rest of the week around it. Pick the road for the workout when a road exists; when none does, move the workout, not the goal.

Quick answers

Sources cited in this guide

1. 01
   Jeffries et al. 2019. An Analysis of Variability in Power Output During Indoor and Outdoor Cycling Time Trials. International Journal of Sports Physiology and Performance.
2. 02
   Smith et al. 2001. Reliability of mean power recorded during indoor and outdoor self-paced 40 km cycling time-trials. International Journal of Sports Medicine.
3. 03
   Nimmerichter et al. 2012. Effects of low and high cadence interval training on power output in flat and uphill cycling time-trials. European Journal of Applied Physiology.
4. 04
   Nimmerichter et al. 2015. Gross efficiency during flat and uphill cycling in field conditions. International Journal of Sports Physiology and Performance.
5. 05
   Hovorka et al. 2022. Effects of Flat and Uphill Cycling on the Power-duration Relationship. International Journal of Sports Medicine.
6. 06
   Seiler et al. 2013. Adaptations to aerobic interval training: interactive effects of exercise intensity and total work duration. Scandinavian Journal of Medicine & Science in Sports.
7. 07
   Rønnestad et al. 2015. Short intervals induce superior training adaptations compared with long intervals in cyclists — an effort-matched approach. Scandinavian Journal of Medicine & Science in Sports.
8. 08
   Coyle & González-Alonso 2001. Cardiovascular drift during prolonged exercise: new perspectives. Exercise and Sport Sciences Reviews.
9. 09
   Swain 1997. A model for optimizing cycling performance by varying power on hills and in wind. Medicine & Science in Sports & Exercise.
10. 10
    Atkinson et al. 2007. Variable versus constant power strategies during cycling time-trials: prediction of time savings using an up-to-date mathematical model. Journal of Sports Sciences.
11. 11
    Madsen et al. 2013. Safety effects of permanent running lights for bicycles: a controlled experiment. Accident Analysis & Prevention.
12. 12
    Wood 2023. Improving the conspicuity and safety of pedestrians and cyclists on night-time roads. Clinical and Experimental Optometry.

Specific questions inside this topic

Focused reads — each answers one question end-to-end.

1. 01

   How long should a hill be for hill repeats?

   Rep length sets the hill. A table from 30-second to 10-minute repeats: how long, how steep, and which workout each hill serves.

   →
2. 02

   The best road for intervals: what to look for

   Long enough for the rep, no lights, a safe turnaround. How to find a stretch of road that lets an interval session run as written.

   →
3. 03

   VO2 max intervals outside: hill or flat road?

   Uphill and flat VO2 reps are not the same work. What the research says about each, and how to ride 5x4 when you have no hills.

   →
4. 04

   Threshold intervals without hills: rolling and flat roads

   No long climb nearby? How to hold threshold and sweet spot on flat or rolling roads: pace by range, use the wind, fit reps to the road.

   →
5. 05

   Zone 2 outside: how to plan a ride without stops

   Lights and junctions break an easy ride into pieces. How to plan a Zone 2 loop that stays steady, and what to do when you cannot.

   →

Related guides

• Indoor vs outdoor FTP: why the numbers differ →
• Heart Rate Rising at the Same Power? Cardiac Drift Explained →
• Indoor cycling training: the self-coached rider's winter guide →