Return to Running After Injury: Building a Plan Your App Won’t Build for You

Your app is very good at one job: taking a fitness level and a race date and filling the gap between them with sessions. That job assumes your tissue can absorb whatever your aerobic system can handle. After an injury, it can’t, and the gap between those two things is where most re-injuries live.

This page is about the audit. You already know how to read your own Strava and your watch. What you need is a set of checks you can run against a Runna block, a Garmin Coach plan or a ChatGPT-built spreadsheet, so you can tell within thirty seconds whether week 1 is going to put you back on the physio’s table.

Why the first four weeks are where plans break

Every mainstream plan generator works backwards from a goal. Runna asks for your race, your date, your days per week and a recent time or estimated pace. Garmin Coach asks something similar and then adapts off your performance and heart rate data. A ChatGPT plan does whatever your prompt implied. None of them have a model of a healing tibia.

The practical failure is specific and repeatable. Tell an app you’re currently running 20 km a week and it will build a week 1 that contains a quality session, a long run at roughly 30 to 40% of weekly volume, and continuous running on consecutive days. Every one of those three things is a reasonable default for a healthy runner and a bad idea for someone six weeks out from a bone stress injury. The app isn’t wrong about aerobic fitness. It’s answering a question you didn’t ask.

There’s a second problem that’s easy to miss: your history has been erased. Adaptive plans lean on your recent load to decide how hard to push, and after five weeks off the bike-and-swim routine you’ve been doing has left your running-specific load near zero. Garmin will read your acute load as low and your Training Readiness as high, because by its own metrics you are extremely well rested. You are. Your calf still isn’t.

Start from the tissue, not from the fitness

The single most useful thing you can do before you touch a plan is name the tissue and its timeline. They are not interchangeable, and the differences drive everything about how the first six weeks are structured.

Bone. Bone stress injuries are the ones where the plan needs to be most conservative and least negotiable. Low-risk sites (posterior medial tibia, metatarsals 2 to 4, fibula) typically need a period of no running until you can walk 30 minutes and hop pain-free, then a graded return over 6 to 10 weeks. High-risk sites (femoral neck, anterior tibial cortex, navicular, talus, sesamoids, base of the fifth metatarsal) need imaging and specialist input, not an app. Bone wants 48 hours between impact sessions in the early weeks and it wants zero pain, not manageable pain. If you had a bone stress injury and you haven’t had a conversation about energy availability, iron and vitamin D, the plan is not the limiting factor.

Tendon. Achilles and patellar tendinopathy behave almost oppositely. They respond to load rather than rest, they tolerate some pain, and they report their status on a 24-hour delay. The standard monitoring approach allows discomfort up to about 5/10 during the run, provided it settles within 24 hours and your morning stiffness the next day is no worse than the day before. Morning stiffness is the honest metric here. If you needed 30 seconds of hobbling on the stairs last Tuesday and you need three minutes today, your last three sessions were too much regardless of what your watch says.

Muscle. Calf and hamstring strains heal faster than either of the above but have the highest reinjury rate, mostly because people return on the strength of feeling fine at easy pace. A grade 1 calf strain can feel resolved at 5:30/km and fail at 4:10/km, because the demand on the soleus scales with speed in a way easy running never tests. The return criterion is not comfort, it’s strength symmetry and speed exposure.

Joint and cartilage. Patellofemoral pain and similar presentations are typically load-distribution problems rather than tissue-damage problems, which means cadence, hill exposure and volume distribution matter more than total volume.

Three gates before you run a step

These are cheap to test at home and they catch a lot of premature returns.

1. Walk 30 minutes continuously, pain-free, on the flat. No limp, no next-morning soreness. If this fails, nothing else matters.
2. Single-leg calf raises to fatigue, both sides. Aim for the injured side within 90% of the good side, and ideally 20 to 25 reps through full range with a straight knee. Count honestly and stop when the heel height drops.
3. 20 single-leg hops in place, then 10 forward. Pain-free during, and pain-free the next morning. This is the closest home proxy for the impact loading of running.

Add a fourth for calf and Achilles cases: a 10-second single-leg heel hold with your knee bent to 30 degrees, which shifts load to the soleus and catches the deficit that straight-knee raises hide.

Pass all of those and you’re ready for a walk-run progression, not for continuous running. The structure, the ratios and the session-by-session increments are laid out in detail in the walk-run protocol for coming back from injury, and it’s worth following that rather than improvising, because the walk intervals are doing real work: they cap the number of consecutive loading cycles your tissue sees before it gets a break.

What a real rebuild looks like in numbers

Here’s a worked six weeks for a specific runner: 42 years old, marathon background, 55 km per week before injury, grade 2 posterior medial tibial stress injury, five weeks without running, now passing all three gates.

Three things about that table are worth arguing with, so let me get there first.

The week 1 to week 2 jump is +100%, which breaks every version of the 10% rule you’ve read. That’s fine. Percentage rules are meaningless at low absolute volumes, because going from 18 to 20 minutes a week is not a meaningful stimulus and going from 18 to 36 is still only 18 extra minutes of running spread over seven days. Percentage caps start to matter around 150 minutes a week and become genuinely important past 250. Below that, cap the absolute increment (roughly 15 to 25 minutes of running added per week) and cap the session count separately.

Notice also that nothing gets faster until week 5, and that the fourth run day appears only after continuous running is established. The order is fixed: frequency, then duration, then intensity. Adding two of those in the same week is the most common self-inflicted setback, and it’s exactly what adaptive plans do when they see a good week.

Intensity comes back in this sequence, one step per week or two: 6 × 15 second strides on flat grass, then short hill reps (uphill only, walk down), then steady or threshold running, then intervals at 5K pace, then fast downhill running last. Downhill is last because it’s eccentric and it’s the single highest per-step load in normal training. If you had a bone or tendon injury, keep downhill reps out until week 10 or later.

Reading your own data when your baseline has been wiped

Your normal metrics are compromised for about four weeks after any extended layoff, and knowing exactly how they’re compromised is most of the skill here.

Acute:chronic workload ratio. The logic is sound (compare your last 7 days against your last 28) and the arithmetic falls apart after time off, because the denominator is close to zero. Come back with 20 minutes of running in week 1 and your ratio can read 3.0 or higher. The ratio is also genuinely contested in the research literature as a predictive tool. Ignore it entirely for the first four weeks, then start watching it once your 28-day average reflects actual running.

Garmin. Acute Load, Load Ratio and Training Status are all downstream of the same problem. What you’ll see in week 2 of a return looks like this:

Training Status    Unproductive
Acute Load         48       (low)
Load Ratio         2.4      optimal range 0.8–1.5    HIGH
Training Readiness 84 / 100
HRV Status         Balanced

Contradictory, and both halves are right. The load ratio is high because the baseline is empty; readiness is high because you genuinely are recovered. Use HRV Status and resting heart rate, which are not corrupted by the layoff, and disregard Training Status until you have four weeks of consistent running behind you. Garmin Coach will also try to hand you a benchmark run or a time trial early in the plan. Delete it.

Strava Fitness and Freshness. More useful during a rebuild than most people expect, because the Fitness curve tells you your ramp rate, and ramp rate is what you’re actually managing. A rebuild that looks like this is well judged:

Date      Fitness   Fatigue   Form
Aug 25       8         5       +3
Sep 01      12        14       -2
Sep 08      17        21       -4
Sep 15      23        29       -6
Sep 22      28        33       -5

Fitness climbing 4 to 6 points a week with Form sitting between 0 and -10 is a sustainable rebuild. A week where Fitness jumps 11 points and Form drops to -25 is the week you will regret. Keep the ramp under roughly 5 to 7 points a week during a return, which is more conservative than the 8-plus you might run in a healthy build.

intervals.icu is free, handles ACWR and ramp rate more transparently than either, and lets you tag sessions. Worth setting up purely for the ramp-rate chart if you don’t pay for Strava.

Auditing an AI-generated plan in five checks

Take whatever your app produced for week 1 and run these against it.

Check 1: Are there back-to-back running days in the first three weeks? For bone and tendon, there shouldn’t be. Alternate days, always. Cross-train on the off days instead; the aerobic work is genuinely useful and carries no impact cost.

Check 2: What percentage of weekly volume is the long run? Apps routinely build a long run at 35 to 45% of the week. During a rebuild, cap it at 30%. In the week 6 row above, that’s 35 minutes out of 115, which is 30% exactly.

Check 3: Is there a quality session before week 5? If yes, remove it and replace with the same duration easy. This is the single highest-value edit you’ll make.

Check 4: Does the plan contain a time trial, benchmark run or race-pace assessment in the first month? Delete it. Your app wants the data. Your tendon is paying for it.

Check 5: Does the weekly increment exceed 25 minutes of running, or add a run day in the same week as a duration increase? Split it across two weeks.

Here’s what that audit looks like in practice. Tell Runna you’re running 20 km a week and point it at a half marathon 12 weeks out, and week 1 typically arrives as something like: 5 km easy, a session of 6 × 800 m with recoveries, a 5 km easy, and a 10 km long run. That’s roughly 25 km with a quality session and a long run at 40% of volume, on a four-day pattern that will put two runs back to back. For the runner in the table above, all four checks fail. The fix isn’t to abandon the app, it’s to lie to it productively: enter a much lower current volume (8 to 10 km), set three days a week rather than four, push the goal race four weeks later than you want it, and manually swap the interval session for easy minutes until week 5.

Garmin Coach needs different handling. It won’t let you strip intensity out of a plan, so the cleanest approach is to not start the plan at all. Run your rebuild as unstructured activities for the first five or six weeks, then load a Garmin Coach 5K or 10K plan once you’re running continuously four days a week. Starting it on day one means the adaptive logic anchors to a baseline that doesn’t exist.

For a ChatGPT-built plan, the model will produce whatever the prompt allows, so make the prompt do the work:

Build me a 10-week return-to-running plan.

Constraints, all of which are hard:
- Weeks 1-4: exactly 3 run days, never on consecutive days.
- Weeks 1-4: walk-run intervals only, no continuous running.
- Total weekly running time increases by no more than 20 minutes per week.
- No quality sessions, strides, hills or time trials before week 5.
- Long run never exceeds 30% of weekly running time.
- Add a 4th run day only in a week where total volume does not increase.
- Two strength sessions per week, listed with sets and reps.
- Output as a table: week, day, session, running minutes, weekly total.
- Add a column showing weekly total as a % increase on the prior week.

Do not add a race date. Do not add taper logic.

That last output column is the useful trick. It makes the model’s own arithmetic visible, and you’d be surprised how often a plan that reads sensibly in prose turns out to contain a 60% week.

The strength work no plan schedules

Every app in this category is a running plan generator, and even the ones that bundle strength give you generic circuits rather than injury-specific rehab. This is the gap you fill yourself, and it matters more than anything in the running plan.

For Achilles and calf, heavy slow resistance is the best-evidenced option: calf raises loaded to around an 8-rep max, 3 to 4 sets, three times a week, for 12 weeks. Slow means roughly 3 seconds up and 3 seconds down. Both straight-knee (gastrocnemius) and bent-knee (soleus) versions, because runners are typically much weaker in the second. Progress the load, not the reps.

For bone stress injuries, the strength priority is hip and calf capacity plus, bluntly, eating enough. Soleus strength reduces tibial bone load meaningfully because the soleus is the primary shock absorber at the ankle during stance. Add impact progression deliberately: two-foot hops before single-leg, in-place before travelling, 30 to 50 contacts per session building over four weeks.

Patellofemoral cases respond to heavy quad and hip work plus a cadence audit. If your cadence at easy pace is 158 and your last 20 easy runs agree, taking it to 168 (a 6% increase, achievable with a metronome app over two weeks) reduces per-step patellofemoral and hip loading noticeably. Check your own Strava cadence history rather than guessing, and look at your easy runs specifically, since your cadence at 5K pace tells you nothing about your cadence at recovery pace.

The rules that override the plan

Write these down somewhere you’ll see them, because in the moment you will negotiate with yourself.

Stop the session if pain increases during the run, if your gait changes, or if pain appears in a new location. Not “reassess at the next lamp post.” Stop, walk home.

The 24-hour rule is the one that catches almost everything: pain during the run should be back to baseline within 24 hours, and next-morning symptoms should be no worse than the previous morning. Two consecutive sessions that fail this means you repeat the previous week rather than progressing. Three means you go back a week and get the injury looked at again.

Repeating a week is not a setback and your app will fight you on it, because plans are built as monotonic progressions toward a date. Build the repeat in deliberately: every third or fourth week of a rebuild should hold volume flat rather than increase it. In the table above, week 5 holds at 95 minutes against week 4’s 96, and that flat week is doing more for the outcome than any of the increases.

Where the walk-run structure earns its place

The reason walk-run works isn’t that it’s easier. It’s that it lets you decouple total aerobic time from consecutive loading cycles, which is the variable that actually matters for a healing tissue. Forty minutes of 4-minutes-run / 1-minute-walk gives you 32 minutes of running with five resets, and those resets change what the tissue experiences even though your watch records nearly the same distance.

This is also why the walk-run phase should run longer than feels necessary. Most people feel ready for continuous running in week 2 and they’re usually wrong, because feeling ready is an aerobic judgment and readiness is a structural one. The walk-run protocol page covers the specific ratios, how to progress them and how to handle a session that goes badly midway through, and it’s worth reading before you build the plan rather than after.

By the time you’re eight or nine weeks in and running four or five days a week with a quality session, your app’s assumptions have become true again. Your 28-day load is real, your Training Status means something, your Fitness curve has a slope worth trusting. Hand the plan back at that point and let it do the job it’s good at.

What you should not do is hand it back earlier because you had a good week. A good week during a rebuild is evidence that the rebuild is working, not evidence that you can skip the rest of it.