iFixit has scored the repairability of consumer electronics—on a 0–10 scale—since 2017, across smartphones, laptops, tablets, wearables, and more. Each teardown is graded mainly on how easy the device is to open up and put back together (roughly 80–90% of the score), plus whether the manufacturer publishes free repair documentation and sells replacement parts at a reasonable price. A device also has to be identifiable while non-working to earn full credit on those last two, and can lose up to all of its points if software or "parts pairing" locks block repairs even when the physical work is easy. The result is a single, comparable-at-a-glance number that gets picked up widely in tech coverage—outlets like MacRumors and TidBITS reported individually on the iPhone 14's score drop, and it shows up regularly in phone-buying guides ranking models by how repairable they are. iFixit's repair advocacy has reached lawmakers directly, too—CEO Kyle Wiens gave written testimony to the U.S. House Judiciary Committee on right-to-repair in July 2023, though the score itself isn't formally adopted by any regulator; France runs its own separate, government-mandated repairability index.
Right-to-repair advocacy and legislation have picked up considerably since 2017—and so has iFixit's own scoring methodology, which has been revised roughly 15 times over that span (see the appendix for what changed, version by version). Has the hardware actually gotten easier to fix as a result? I pulled 172 scored products (2017–2026, ten product categories) from iFixit's public repairability hub to find out.
The headline: yes, but unevenly
The average score rose from about 5.3 (2017) to 6.9 (2025)—roughly +0.25 points per year, a small but statistically reliable trend (p = 0.006). But averaging across everything hides the more interesting story: the rise is concentrated in a handful of product categories, not spread evenly across the electronics industry.
Laptops, smartphones, and e-readers are climbing. Laptops show the clearest, most reliable rise (+0.70 pts/year, p = 0.018)—driven largely by Framework's modular design and Lenovo's ThinkPad line, both of which now regularly score 8–10. Smartphones rose more modestly (+0.53 pts/year), and that rise is really one event: Apple's iPhone jumped from 4/10 in 2022 (glued-in battery) to 7/10 from 2023 onward, coinciding with mounting right-to-repair pressure in the EU and several US states (see the appendix for one relevant piece of EU law). E-readers show the steepest slope of all, though with only 10 products scored, that estimate is noisy.
Tablets and wireless earbuds haven't moved at all. The iPad has scored 2–3 out of 10 in nearly every generation since 2017; AirPods have scored 0 in two of the last three generations. Whatever pressure is nudging laptops and phones upward hasn't reached these categories.
Brand matters more than year
Once you look at averages by manufacturer, calendar year explains relatively little. Fairphone has scored a perfect 10/10 on every product it has ever submitted—phones, earbuds, and headphones alike. Lenovo averages 8.9. On the other end, Apple, Samsung, and JBL average around 3.5–4.5 and have stayed there across a full decade of product cycles.
| Brand | Avg. score | Products scored |
|---|---|---|
| Fairphone | 10.0 | 5 |
| Lenovo | 8.9 | 17 |
| Microsoft | 5.1 | 7 |
| 4.8 | 19 | |
| Samsung | 4.6 | 11 |
| Apple | 4.5 | 38 |
| JBL | 3.6 | 5 |
Is this just the rubric changing?
iFixit has revised its scoring rubric about 15 times since 2017—adding a parts-availability criterion in 2019, a parts-pairing penalty in 2020, category-specific criteria for tablets, wearables, and handhelds along the way. That's a legitimate wrinkle: a 5/10 scored under the 2017 rubric and a 5/10 scored under the 2026 rubric aren't graded on quite the same basis.
Ruling out rubric drift
I re-ran the trend controlling for which rubric version was used (12 versions in the dataset). The upward trend not only survives—it holds essentially unchanged (+0.28 pts/year, p = 0.017). Rubric drift alone doesn't explain the rise. Category composition is a harder factor to fully separate from rubric version, since several rubric revisions were introduced specifically to formalize scoring for one category (handheld consoles, headphones, and speakers each got their own dedicated version)—so the two can't be cleanly disentangled in a single model. Controlling for category alone, the trend still holds (+0.22 pts/year, p = 0.031), just somewhat attenuated.
So: what we observe—an upward trend—isn't just an artifact of the rubric changing. In fact, looking more closely at what each revision actually changed, we'd argue the estimate above is if anything understating the real improvement, not overstating it.
That's because the changes almost all push toward stricter, not more lenient, scoring. Walking through them (full list in the appendix):
- v1.4 added the Parts Availability criterion and the Unique Product Identifier gate—a new way to lose credit that didn't exist before.
- v1.6 added the Parts Pairing penalty, which can zero out up to 100% of the score. No product could be marked down for this before 2020ish, and it's exactly the kind of software lock manufacturers have leaned into more since, not less.
- v1.7 raised the documentation bar from "free, public" to "free, public, complete."
- The one exception: v2.1 widened Parts Availability to count off-the-shelf standard parts, not just OEM—a loosening, but on a single 10%-weighted sub-criterion, against three tightenings elsewhere.
Put plainly: the same physical design would very likely score the same or lower under the current rubric than it would have under the 2017 one—essentially never higher. There's a hint of this in the regression, too: controlling for rubric version nudged the year slope up slightly (+0.25 → +0.28 pts/yr), not down. That's the direction you'd expect if later rubrics are systematically harsher on the same design—leveling the playing field revealed a bit more trend, not less.
I can't turn this into a precise correction, though. iFixit doesn't publish the point-level deduction tables behind each version, and rubric version is too entangled with product category to isolate "stricter scoring" cleanly from "which products happened to get scored when" (v2.1, v2.3, and v2.4 each map almost one-to-one onto a category being formalized for the first time). So take it as a directional argument, not a corrected estimate: if anything, the +0.25–0.28 pts/year trend above is a floor on the real improvement in repair-friendly design, not a ceiling.
Takeaway
The repairability gains are real, but they're not an industry-wide shift—they're concentrated in a few product lines with strong incentives to design for repair: Framework and Fairphone built their brand on it, Lenovo's ThinkPad line has a long enterprise-serviceability tradition, and Apple's post-2022 iPhone redesign lines up with mounting right-to-repair law in the EU and several US states. Tablets and wireless earbuds, where no comparable pressure exists yet, haven't budged.
Appendix: iFixit's scoring rubric, version by version
iFixit's repairability score has run 0–10 (10 = most repairable) since it launched in 2017, but the rubric behind that number has been revised roughly 15 times—12 distinct versions appear in this dataset. Every scored product links to the specific rubric version used to grade it. Below is what changed, version by version, oldest to newest.
The core structure (holds across nearly every version)
| Criterion | Typical weight | What it measures |
|---|---|---|
| Design for Repair | 80–90% | Ease of disassembly/reassembly—the bulk of the score |
| Service Manual | 10% | Free, public, complete repair documentation via the OEM support site |
| Parts Availability & Cost | 10% | Whether replacement parts (OEM, later OEM-endorsed/standard) are available at reasonable cost |
| Unique Product Identifier | Gating, no direct weight | Whether the device can be identified (model number, service tag) while non-working—gates full credit on the two criteria above |
| Parts Pairing | Penalty only | Deduction of up to −100% when repairs are blocked by software/parts-pairing or calibration locks with no free remedy |
"Duration of software security updates" is flagged repeatedly across versions as a factor iFixit intends to add but has not yet incorporated.
Version-by-version changes
| Version | What changed |
|---|---|
| v1.0 (2017) | First published scorecard. Design for Repair 90% / Service Manual 10% only. Parts availability, parts-pairing penalties, a unique-identifier gate, and software-update duration are all explicitly flagged as not yet scored. |
| v1.3 | Laptop modularity refinement: replaced I/O-port-specific disassembly/reassembly proxy times with a simple modularity score for laptops. Weights unchanged (90/10). |
| v1.4 | Parts availability introduced (10%, OEM parts only), design weight drops to 80%. New Unique Product Identifier gating function: a device must be identifiable in a non-working state to earn full documentation/parts credit. Documentation standard tightened to require tools lists, numbered parts lists, exploded diagrams, troubleshooting procedures, and schematics—all free and public. |
| v1.6 | Parts Pairing penalty introduced (global, up to −100%). Refined proxy times for various de-gluing actions. |
| v1.7 | Tablets & foldables formalized under the smartphone framework; added proxy times for tablet-specific design features. Service Manual standard raised to "free, public, complete." Hardware-troubleshooting section now required for full documentation credit. |
| v1.8 | Wearables (smartwatches, wireless earbuds) formalized and given their own proxy times. More specific guidance for individually scored repair procedures. |
| v1.9 | E-readers split out from tablets as their own product group, with their own proxy times. Schematics scoring refined based on feedback from board-level repair professionals. |
| v2.0 | Tooling & action refinements: adjusted tool categorizations and scaling factors, replaced deprecated actions, tightened internal grading guidance for consistency. First version to specifically flag "software security update" duration (vs. generic software updates) as a reserved future factor. |
| v2.1 | Handheld game consoles formalized. Parts Availability & Cost widened from OEM-only to also include off-the-shelf standard parts. Adjusted tool multipliers for heat sources. |
| v2.2 | Proxy-time overhaul: actions and proxy times revised broadly for clarity and consistency, some actions retired/replaced. Adjusted T2 Torx/Torx Plus multipliers to reflect increased market availability of those tools. |
| v2.3 | Wireless headphones formalized. Added proxy times for common soldering actions and adjusted existing soldering-action times. |
| v2.4 (current, Q3 2026) | Portable wireless speakers formalized. Added proxy timing benchmarks for speaker repairs, adjusted thermal-compound separation times, and modified tool multipliers for combination tools requiring multiple simultaneous uses. |
Two caveats this raises for the trend above
- Provisional scores. Some products in the dataset are flagged as carrying a "provisional" score—typically assigned shortly after a teardown, before the full documentation/parts-availability review is complete. These run marginally higher on average (5.8 vs. 5.4) but the difference isn't meaningful given the small provisional sample (n = 10).
- Version drift. Because the weighting, gating rules, and penalties keep evolving, a score from one version isn't perfectly apples-to-apples with a score from another—a 5/10 under v1.0 (90% disassembly weight, no parts-pairing penalty) isn't graded on the same basis as a 5/10 under v2.4 (80% weight, a parts-pairing penalty, and wider parts-availability criteria). As noted in the body of this post, controlling for rubric version statistically doesn't erase the upward trend—but version and product category are too entangled to fully separate, since several versions (v1.9, v2.1, v2.3, v2.4) were introduced specifically to formalize scoring for one new category each.
Source: ifixit.com/repairability and the linked Repairability_Scoring_Rubric_v* wiki pages. Full point-level scoring tables (exact proxy times, tool multipliers, per-action deductions) aren't published in plain text on iFixit's summary rubric pages—iFixit describes categories and weights but keeps the granular scoring guide for internal/reviewer use.
Appendix: the EU's new smartphone/tablet repairability law
One piece of regulation worth knowing about, since it overlaps with the back half of this dataset: Commission Regulation (EU) 2023/1670 and Delegated Regulation (EU) 2023/1669—the EU's Ecodesign and energy-labelling rules for smartphones, cordless phones, and tablets. Adopted in 2023, they became mandatory for products placed on the EU market on June 20, 2025.
What's on the label
Every covered device now needs an energy label (the same style as a refrigerator's or washing machine's) showing energy efficiency, a battery-endurance rating, resistance to drops/scratches/dust/water, and a repairability score from A (best) to E (worst). Here's roughly what one looks like (this is my own mock-up of the layout, not an official EU graphic):
Illustrative layout of the mandatory EU energy label for smartphones/tablets—the letter grades and numbers above are placeholders, not a real product's ratings.
| What it looks at | What it's checking |
|---|---|
| Disassembly | How many steps it takes to get to a given component |
| Access | What tools are needed to reach and replace it |
| Parts availability | Whether spare parts for it are actually obtainable |
| Documentation | Whether repair information is freely accessible |
That's conceptually close to iFixit's own Design-for-Repair/Service-Manual/Parts-Availability structure (see the other appendix)—but it's a separate, government-run methodology with its own weighting, not a version of iFixit's score.
The other requirements
- Spare parts: manufacturers must supply key components within 5–10 working days, for at least 7 years after a model stops being sold—currently to professional repairers only.
- Software support: a minimum of 5 years of OS updates from when the last unit of a model is sold, plus fair access to the firmware/software repairers need to do their job (aimed at exactly the kind of parts-pairing lockouts that motivated iFixit's own penalty in v1.6, see the other appendix).
- Battery durability: at least 800 charge cycles while retaining 80% of original capacity.
- Removable batteries by 2027: a related but separate law, the EU Battery Regulation (EU) 2023/1542, requires portable batteries in devices including phones to be user-removable and replaceable starting February 2027.
This also explains something from the "is this just the rubric changing" discussion above: PIRG's 2026 scorecard graded phones on this new EU-wide EPREL standard for the first time, replacing the French national index it had used for phones previously—and grades dropped across the board as a result. Laptops aren't covered by this regulation yet, so PIRG kept grading them on France's index, which is part of why laptop grades barely moved while phone grades swung sharply. It's the same phenomenon as rubric drift, just one level up: switching which official standard you use can move the numbers as much as an actual design change does.
I'm not claiming this EU law explains any specific score movement in the iFixit dataset above—I haven't verified a causal link, and iFixit's score is a separate, private methodology from the EU's own rating. It's included here as background, not as an explanation. Worth flagging too: a one-year compliance check by Right to Repair Europe found roughly 80% of listed smartphone models still don't actually provide the repair information the law requires—so however this law eventually affects the numbers, enforcement is still catching up to the rule on paper.
Sources: European Commission announcement, JRC methodology summary, Right to Repair Europe's compliance check, and EU Battery Regulation guidance.
Data: iFixit repairability scores and linked scoring-rubric revisions, compiled 2026. Analysis: OLS regression with rubric-version and category fixed effects; robust (HC1) standard errors.