GymStreak research · Progress data study
How lifters progress
When weekly gym streaks break, what a new lifter’s first week predicts, how often the weight goes up from one session to the next, how quickly new personal bests dry up and how much strength changes over a year. From 723,068 back-to-back sessions and 30,794 new GymStreak lifters.
By GymStreakPublished Data to
- Back-to-back sessions
- 723,068
- Lifters (progression)
- 22,819
- New lifters (retention)
- 30,794
- Lifts analysed
- 15
Key findings
- 1
About half of new lifters never log a second workout
15,784 of 30,794 people who logged a first workout never logged another. Across everyone, 15% were still logging in week 4 and 8% in week 12.
- 2
Streaks break in the first month, then get sticky
A first weekly streak ends after week 1 72% of the time, after week 2 50% and after week 4 26%. By week 20 the chance of it ending that week is 9.7%.
- 3
The first week predicts the next three months
New lifters who trained on 4 or more days in their first week were still logging at week 4 51% of the time, against 6% for those who trained on one day. At week 12 it was 28% against 3%.
- 4
Comebacks happen fast or not at all
31.7% of broken streaks were followed by a return within 12 weeks, and 14.2% came back after missing a single week. After a first streak of one week only 13% returned; after 16+ weeks, 79% did.
- 5
Half the time, lifters repeat last session’s weight
Between back-to-back sessions of a lift the top weight went up 32.5% of the time, stayed the same 52.3% and dropped 15.2%. When it stayed the same, 55.5% of sessions also repeated the same reps.
- 6
New personal bests get rare fast
51.5% of second sessions on a lift set a new estimated best, 18.5% of tenth sessions and 5.4% of sessions 41–80. The curve is almost the same for beginners, intermediates and advanced lifters.
- 7
Bench press: about +14% in a year, and a quarter go backwards
Among lifters still benching at each checkpoint, the median estimated one-rep max was 7% higher after about 2 months, 11% after 6 and 14% after 12. About a quarter were weaker than at the start (25% at 12 months).
Read this first
These are people who chose to use a gym app, not a representative sample. Everything here is correlation: keen people both train more in week 1 and stick around longer. The progress figures only include lifters who were still logging a lift at each point, so they flatter the typical result (survivor bias). Every figure describes a group of at least 100 people. Details in the methodology.
01 / Streaks
When streaks break
A weekly streak here is a run of consecutive Monday-to-Sunday weeks with at least one logged set. We followed the first streak of every lifter whose first logged week fell between January 2024 and June 2026 and asked: once a streak has reached a given length, how likely is it to end that week?
Chance a first weekly streak ends after each week
Weeks the streak has lasted
Show the numbers
| Streak has reached | First streaks | Chance it ends that week | Later streaks: chance it ends |
|---|---|---|---|
| 1 week | 30,831 | 71.9% | 54.8% |
| 2 weeks | 8,659 | 49.9% | 39.9% |
| 3 weeks | 4,338 | 35.5% | 31.1% |
| 4 weeks | 2,798 | 25.6% | 25.5% |
| 5 weeks | 2,082 | 23.5% | 23.7% |
| 6 weeks | 1,592 | 22.1% | 21.6% |
| 7 weeks | 1,241 | 19.2% | 18.7% |
| 8 weeks | 1,003 | 19.0% | 20.6% |
| 9 weeks | 812 | 16.9% | 17.8% |
| 10 weeks | 675 | 16.3% | 15.0% |
| 11 weeks | 565 | 12.7% | 15.0% |
| 12 weeks | 493 | 15.8% | 14.4% |
| 13 weeks | 415 | 12.3% | 12.9% |
| 14 weeks | 364 | 12.4% | 12.3% |
| 15 weeks | 319 | 11.0% | 9.9% |
| 16 weeks | 282 | 11.4% | 13.3% |
| 17 weeks | 249 | 9.6% | 11.8% |
| 18 weeks | 225 | 11.1% | 13.1% |
| 19 weeks | 199 | 11.6% | 10.2% |
| 20 weeks | 175 | 9.7% | 5.8% |
| 21 weeks | 156 | 9.6% | 9.7% |
| 22 weeks | 141 | 7.8% | 6.2% |
| 23 weeks | 130 | 9.2% | 7.5% |
| 24 weeks | 118 | 5.9% | – |
| 25 weeks | 111 | 7.2% | – |
| 26 weeks | 102 | 4.9% | – |
n = 30,831 first streaks. Weeks shown while at least 100 streaks remain.
Later streaks, after someone has already broken one and come back, are a little sturdier at the start: 55% end after their first week. From about week 4 the two curves are almost the same.
Comebacks happen fast or not at all
Of 46,192 broken streaks with 12 weeks of follow-up, 31.7% were followed by a return within 12 weeks, and 14.2% came back after missing just one week. The longer the streak that broke, the more likely the lifter was to come back.
Back within 12 weeks after a first streak breaks, by how long it lasted
- 1 week13%
- 2–3 weeks30%
- 4–7 weeks57%
- 8–15 weeks66%
- 16+ weeks79%
Show the numbers
| Streak | Length | Breaks | Back after 1 missed week | Within 2 weeks | Within 4 weeks | Within 12 weeks |
|---|---|---|---|---|---|---|
| First | 1 week | 22,172 | 4.8% | 7.0% | 9.3% | 12.6% |
| First | 2–3 weeks | 5,861 | 14.0% | 19.8% | 23.9% | 30.2% |
| First | 4–7 weeks | 1,794 | 29.1% | 40.0% | 47.3% | 56.6% |
| First | 8–15 weeks | 707 | 40.0% | 50.1% | 57.1% | 65.6% |
| First | 16+ weeks | 203 | 50.7% | 63.5% | 71.4% | 78.8% |
| Later | 1 week | 8,730 | 17.3% | 25.0% | 33.1% | 44.7% |
| Later | 2–3 weeks | 4,104 | 28.3% | 39.9% | 50.6% | 62.3% |
| Later | 4–7 weeks | 1,785 | 38.5% | 52.7% | 63.4% | 73.9% |
| Later | 8–15 weeks | 642 | 45.8% | 60.1% | 69.2% | 77.3% |
| Later | 16+ weeks | 194 | 52.1% | 69.1% | 76.3% | 80.9% |
If you miss a week, the next one matters most. Plans that fit the days you really have, such as a 3-day full body plan, are easier to keep going than an ambitious one. See also how many times a week to work out.
02 / First week
What the first week predicts
We took the 30,794 people whose first logged workout fell between 1 January 2024 and 9 June 2026, so everyone had at least 16 weeks of follow-up. “Week 4” means they logged anything 28–55 days after their first workout; “week 12” means 84–111 days. Overall, 15% were still logging at week 4 and 8% at week 12. 15,784 never logged a second workout at all.
Still logging at week 4, by workout days in the first week
- 1 day6%
- 2 days18%
- 3 days32%
- 4+ days51%
Coming back at all matters more than coming back quickly. Among people who did log a second workout, the share still logging at week 4 ranged from 27% to 38% however many days later it came (32% when it came 1–2 days later). The other big signals are a fuller first week (more sets, more exercises) and lifting some weights: 6% of people with no weighted sets in week 1 were still logging at week 4, against 25% of those who mixed weighted and bodyweight sets.
| First week | People | Week 4 | Week 12 |
|---|---|---|---|
| Workout days in week 1 | |||
| 1 day | 19,004 | 5.7% | 2.5% |
| 2 days | 5,353 | 18.3% | 9.8% |
| 3 days | 3,214 | 32.4% | 16.7% |
| 4+ days | 3,223 | 51.2% | 28.4% |
| Days until the second workout | |||
| 1–2 days | 8,836 | 31.5% | 16.8% |
| 3–4 days | 2,015 | 31.6% | 18.0% |
| 5–7 days | 1,325 | 26.6% | 13.7% |
| 8–14 days | 882 | 30.3% | 14.2% |
| 15–28 days | 649 | 37.8% | 10.3% |
| 29+ days | 1,303 | 35.8% | 17.9% |
| Never | 15,784 | 0.0% | 0.0% |
| Working sets in week 1 | |||
| Under 10 | 7,625 | 4.5% | 1.9% |
| 10–19 | 5,482 | 10.0% | 4.9% |
| 20–39 | 10,517 | 11.0% | 5.8% |
| 40–79 | 5,265 | 32.7% | 17.1% |
| 80+ | 1,905 | 52.0% | 28.1% |
| Weights in week 1 | |||
| No weighted sets | 13,171 | 5.9% | 2.8% |
| Weighted and bodyweight sets | 14,311 | 24.8% | 13.1% |
| Weighted sets only | 3,312 | 12.9% | 6.6% |
| Length of the first workout (app-timed) | |||
| Under 30 min | 5,529 | 15.6% | 8.1% |
| 30–44 min | 4,459 | 22.9% | 11.3% |
| 45–59 min | 3,334 | 25.6% | 13.2% |
| 60–89 min | 3,151 | 28.1% | 15.2% |
| 90 min or more | 7,341 | 8.3% | 4.1% |
| Not timed | 6,980 | 7.5% | 4.0% |
n = 30,794 new lifters, first workout 1 Jan 2024 to 9 Jun 2026. Week 1 = the first workout day plus the next 6 days.
Correlation, not cause
People who were already keen do more in their first week and also stick around longer. These figures show who stays, not that forcing extra sessions would make someone stay.
03 / Session to session
Session to session: up, same or down
For the 15 most-logged weighted lifts we compared each session’s top working weight with the lifter’s previous session of the same lift, when the two were 1–21 days apart: 723,068 pairs of sessions from 22,819 lifters.
Top weight compared with the previous session of the same lift
- Weight went up32.5%
- Same weight52.3%
- Weight went down15.2%
When the weight does go up, it goes up a lot. The median jump is 4.5 kg, or 13.3% of the load, and on small dumbbell lifts the relative jump is bigger still: 24.9% on lateral raises, where the next dumbbell up is a big step.
What comes before a jump
On machine and isolation lifts about half of weight increases follow a high-rep session, the classic “double progression” of adding reps, then weight: 48.2% of leg extension increases and 51.5% of lateral raise increases came after a session of 12 or more reps. On the barbell bench press, squat and deadlift, about a quarter of increases came after a session of fewer than 8 reps (27.5%, 26.3% and 26.3%).
| Lift | Lifters | Up | Same | Down | Median jump | Jump % | Same weight, same reps |
|---|---|---|---|---|---|---|---|
| Barbell bench press | 8,800 | 34% | 50% | 16% | 5.0 kg | 11% | 52% |
| Barbell squat | 5,022 | 39% | 45% | 16% | 5.0 kg | 13% | 56% |
| Barbell deadlift | 4,085 | 39% | 45% | 16% | 6.8 kg | 14% | 57% |
| Standing overhead press | 3,129 | 30% | 56% | 13% | 4.0 kg | 14% | 53% |
| Lat pulldown | 10,278 | 31% | 52% | 16% | 5.0 kg | 11% | 56% |
| Seated cable row | 7,123 | 33% | 52% | 16% | 5.0 kg | 13% | 57% |
| Bent-over row | 5,431 | 33% | 54% | 13% | 4.5 kg | 14% | 55% |
| Dumbbell bench press | 7,002 | 30% | 58% | 12% | 2.3 kg | 13% | 52% |
| Incline dumbbell press | 6,529 | 30% | 56% | 14% | 2.3 kg | 13% | 53% |
| Machine chest press | 5,718 | 35% | 48% | 17% | 4.5 kg | 13% | 51% |
| Leg press | 6,924 | 39% | 45% | 15% | 10.0 kg | 12% | 59% |
| Leg extension | 6,995 | 36% | 50% | 15% | 5.0 kg | 11% | 60% |
| Rope tricep pushdown | 7,046 | 32% | 50% | 18% | 2.5 kg | 15% | 56% |
| Hammer curl | 7,820 | 26% | 61% | 13% | 2.0 kg | 17% | 54% |
| Lateral raise | 6,920 | 25% | 62% | 14% | 2.0 kg | 25% | 59% |
Pairs of sessions of the same lift 1–21 days apart. Top weight = the heaviest working set; weights in pounds were converted to kilograms, which is why some dumbbell jumps are 2.3 kg (5 lb).
Work out your own numbers with the one rep max calculator, or read the progressive overload definition.
04 / Personal bests
How often lifters set a new best
For each session we estimated a one-rep max from the best set of 10 reps or fewer (Epley formula) and counted the session as a new best if it beat every earlier session of that lift. Then we grouped sessions by their number on that lift: the 2nd time someone logged it, the 3rd, and so on.
Share of sessions that set a new estimated best, by session number on the lift
- Session 251.5%
- Session 340.7%
- Session 433.8%
- Session 529.1%
- Session 626.0%
- Session 723.2%
- Session 821.7%
- Session 919.7%
- Session 1018.5%
- Session 11–2014.3%
- Session 21–409.0%
- Session 41–805.4%
- Session 81+2.6%
Show the numbers
| Session on the lift | Lifters | Sessions | All | Beginner | Intermediate | Advanced |
|---|---|---|---|---|---|---|
| 2 | 20,086 | 80,194 | 51.5% | 49.6% | 51.7% | 52.4% |
| 3 | 13,301 | 56,379 | 40.7% | 39.3% | 40.9% | 41.6% |
| 4 | 10,279 | 43,926 | 33.8% | 32.0% | 33.8% | 34.9% |
| 5 | 8,329 | 35,654 | 29.1% | 27.3% | 29.5% | 29.4% |
| 6 | 7,016 | 29,758 | 26.0% | 25.2% | 26.3% | 25.9% |
| 7 | 6,051 | 25,247 | 23.2% | 21.5% | 23.3% | 24.2% |
| 8 | 5,337 | 21,970 | 21.7% | 21.2% | 21.7% | 21.9% |
| 9 | 4,792 | 19,399 | 19.7% | 19.1% | 19.7% | 20.1% |
| 10 | 4,308 | 17,245 | 18.5% | 17.5% | 18.5% | 19.1% |
| 11–20 | 3,908 | 105,837 | 14.3% | 14.7% | 14.3% | 14.3% |
| 21–40 | 1,904 | 81,322 | 9.0% | 9.4% | 9.2% | 8.4% |
| 41–80 | 735 | 48,410 | 5.4% | 6.1% | 6.0% | 4.2% |
| 81+ | 223 | 21,439 | 2.6% | – | 2.6% | – |
n = 20,086 lifters with at least 2 sessions of one of the 15 lifts.
The curve barely depends on the experience level people picked when they joined: 17.5% of tenth sessions set a new best for self-described beginners, 18.5% for intermediates and 19.1% for advanced lifters. What matters is how long someone has been doing that lift, not the label. A run of sessions without a new best after a few months on a lift is normal, and a cue to change reps, sets or the exercise rather than a sign that something is wrong.
05 / Strength gains
Strength change over a year
Here the starting point is each lifter’s best estimated one-rep max in their first 14 days on a lift. We compared it with their best about 2 months (days 49–70), 6 months (days 168–196) and 12 months (days 350–380) later.
Survivor bias
Only lifters who were still logging that lift at each checkpoint are counted. People who stopped, got injured or switched exercises drop out, and they are probably the ones who progressed least. So these figures describe people who kept going; the typical result for everyone who starts is likely lower.
| Lift | After 2 months | After 6 months | After 12 months | Lifters at 12 months | Weaker at 12 months |
|---|---|---|---|---|---|
| Barbell bench press | +7% | +11% | +14% | 596 | 25% |
| Barbell squat | +9% | +12% | +14% | 337 | 27% |
| Barbell deadlift | +9% | +15% | +18% | 212 | 24% |
| Standing overhead press | +5% | +13% | +15% | 140 | 24% |
| Lat pulldown | +5% | +10% | +13% | 627 | 28% |
| Seated cable row | +6% | +11% | +17% | 397 | 26% |
| Bent-over row | +6% | +11% | +20% | 189 | 21% |
| Dumbbell bench press | +9% | +13% | +20% | 334 | 23% |
| Incline dumbbell press | +9% | +14% | +20% | 381 | 24% |
| Machine chest press | +8% | +11% | +18% | 283 | 28% |
| Leg press | +13% | +14% | +25% | 371 | 25% |
| Leg extension | +11% | +17% | +20% | 398 | 24% |
| Rope tricep pushdown | +5% | +8% | +13% | 346 | 31% |
| Hammer curl | +5% | +9% | +11% | 407 | 31% |
| Lateral raise | +5% | +12% | +14% | 393 | 28% |
On the bench press the median estimated 1RM rose 7% after about 2 months, 11% after 6 and 14% after 12 months, a median gain of 7.5 kg in a year (596 lifters). At every checkpoint a sizeable minority were weaker than when they started: 24%, 26% and 25%. Progress is real but slow and uneven, even among people who stuck with it.
Lifts and checkpoints with fewer than 100 lifters are not shown. Estimated from logged sets, not tested maxes.
Compare your own lifts
See where your lifts sit in the strength standards from GymStreak lifters, estimate your max with the one rep max calculator or check your level with the strength level calculator. For how people train week to week, read how lifters actually train.
06 / Methodology
Methodology and caveats
What the data is
The workout sets GymStreak users logged in the app, from the same pseudonymised dataset as How lifters actually train (workout-log backup of 29 September 2026; every ID replaced with a keyed hash whose key was destroyed; workouts logged by under-18s removed; processing inside GymStreak’s own Google Cloud project). The queries for this page ran on 30 September 2026 and returned aggregates only.
Definitions
- New lifters: people whose first logged set fell between 1 January 2024 and 9 June 2026. Week 1 is the first workout day and the 6 days after it. Week 4 = any logged set on days 28–55; week 12 = days 84–111. Dates are UTC days.
- Streaks: consecutive Monday-to-Sunday weeks with at least one logged set, for lifters whose first logged week began between 1 January 2024 and 8 June 2026. A streak counts at a given length only if the following week could be observed. Comebacks use breaks with at least 12 weeks of follow-up.
- Lifts: the 15 most-used weighted catalogue exercises, all logs from December 2019 to September 2026. Working sets are ticked, non-warm-up sets with a weight above 0 and up to 400 kg. Pounds were converted to kilograms.
- Session to session: the top working weight in a session compared with the previous session of that lift, when the two were 1–21 days apart. Reps 1–30.
- New best: a session whose best estimated one-rep max (Epley: weight × (1 + reps ÷ 30), sets of 1–10 reps) beats every earlier session of that lift by more than 0.1%.
- Experience: the level people chose when they joined (beginner, intermediate or advanced). Answers that look like the app’s defaults are treated as unknown.
Caveats
- Survivor bias. The progress and strength-change figures only count lifters still logging a lift at each point. Those who stopped are missing, so the figures are likely better than the typical result for everyone who starts.
- Correlation, not cause. The first-week and comeback figures show who stays, not what makes them stay.
- Only people who logged at least one workout are included. People who installed the app and never logged anything are invisible here.
- A missing week means nothing was logged. Some people train without logging, so streaks and retention are measured in the app, not in the gym.
- Fewer new lifters appear in 2025 than in 2024 or 2026. The app may have stored some 2025 workouts elsewhere, which we are still checking. The new-lifter figures pool January 2024 to June 2026, so we don’t report year-on-year trends.
- Estimated one-rep maxes depend on users entering weights correctly, and are less accurate for higher-rep sets.
- Every published number describes at least 100 different people. Breakdowns by sex, age or bodyweight are not published for this study. Percentiles and medians use approximate quantiles.
07 / Data
Download the data
The summary tables behind this page are free to download as CSV under a Creative Commons Attribution 4.0 (CC BY 4.0) licence. They contain aggregates only (every row describes at least 100 lifters). Column notes are in the README.
- When streaks breakChance a weekly streak ends after each week (1 to 26), first and later streaks.r4_streak_hazard.csv
- Comebacks after a breakShare of broken streaks followed by a return within 1, 2, 4 and 12 weeks, by streak length.r5_comeback_after_break.csv
- First week and retentionShare of new lifters still logging at weeks 4 and 12, by what they did in their first week.r2_week1_predictors.csv
- Session to sessionHow often the top weight goes up, stays or drops between sessions of a lift, and the size of the jumps.p2_session_to_session.csv
- New personal bestsShare of sessions that set a new estimated best, by session number on the lift, overall and by experience.p3_pr_rate_by_session_number.csv
- Strength change over timeChange in estimated 1RM after about 2, 6 and 12 months, per lift, overall and by experience.p1_e1rm_change_windows.csv
08 / Cite
How to cite this
You’re welcome to quote and chart these figures. Please credit GymStreak and link to this page. Suggested citation:
GymStreak (2026). How lifters progress: GymStreak product data study. Data: GymStreak app users, 723,068 back-to-back sessions (22,819 lifters) and 30,794 new lifters (first workout Jan 2024–Jun 2026).
Short form for captions: “Data: GymStreak app users, 2019–2026.” Link: https://www.gymstreak.com/research/how-lifters-progress
09 / FAQ
Frequently asked questions
- How often do lifters set a new personal record?
- Often at first, then rarely. In GymStreak’s data 51.5% of second sessions on a lift set a new estimated best, 18.5% of tenth sessions, 14.3% of sessions 11–20 and 5.4% of sessions 41–80. A best here is a new estimated one-rep max from a set of 10 reps or fewer.
- When do gym streaks usually break?
- In the first few weeks. A new lifter’s first weekly streak ends after week 1 72% of the time, after week 2 50%, after week 3 36% and after week 4 26%. Streaks that pass a month are much safer: the weekly chance of ending is 16% at week 12.
- How fast does the bench press go up?
- Slowly, on average. Among GymStreak lifters still benching at each checkpoint, the median estimated one-rep max was 7% higher after about 2 months, 11% after 6 months and 14% after 12 months (median gain 7.5 kg at 12 months). About a quarter were weaker than when they started.
- How much weight do people add between sessions?
- When the top weight goes up, the median jump is 4.5 kg, or 13.3% of the load. Most of the time it doesn’t go up: the weight rose in 32.5% of back-to-back sessions, stayed the same in 52.3% and dropped in 15.2%.
- How many people stick with the gym?
- Among 30,794 people who logged a first workout in GymStreak, 15,784 never logged a second. 15% were still logging in week 4 and 8% in week 12. People who trained on 4 or more days in their first week were still logging at week 4 51% of the time.
- Can I use this data?
- Yes. The summary tables are free to download and reuse under a Creative Commons Attribution 4.0 (CC BY 4.0) licence. Please credit GymStreak and link to this page.
Track your own progress
GymStreak logs every set, shows when you hit a new best and keeps your weekly streak going, so you can see your own progress over time.
