How We Calculate
Sundr is a repair intelligence platform that analyzes hundreds of thousands of real repair outcomes to tell you what lasts, what breaks, and when fixing it is smarter than replacing it. Every recommendation is the result of a weighted scoring model that evaluates five factors using real-world data.
Sundr's Repair Breakpoint model evaluates five factors: cost, remaining lifespan, environmental impact, parts availability, and repair success rate. Using data from 305,000+ community repair records, iFixit teardowns, and the French Indice de Réparabilité to determine whether repairing a device is economically rational.
The Repair Breakpoint
The commonly cited 50% rule says if repair costs more than half the replacement price, you should replace. It's a useful starting point, but it only looks at cost. Sundr evaluates five weighted factors to determine your Repair Breakpoint , the actual point where repair stops making economic sense.
Each factor produces a score from −1 (strong repair signal) to +1 (strong replace signal). These scores are multiplied by their weights and combined into a single weighted average. A negative final score means we recommend repair; positive means replace. The further the score is from zero, the more confident the recommendation.
The Five Factors
1. Cost (40% default weight)
Uses the 50% rule as a baseline: if the repair cost exceeds 50% of the replacement cost, the cost signal favours replacement. This is one input to the broader Repair Breakpoint.
- Repair ≤ 25% of replacement cost → strong repair signal
- 25-50% → moderate repair signal
- 50-75% → moderate replace signal
- ≥ 75% → strong replace signal
- Under warranty → repair is free, maximum repair signal
2. Lifespan (20% default weight)
Compares the device's current age against its expected total lifespan. Newer devices are worth repairing; devices near end-of-life favour replacement.
- < 25% life used → strong repair signal
- 25-50% → moderate repair signal
- 50-75% → moderate replace signal
- > 75% life used → strong replace signal
Support lifecycle: When security or software update end dates are known, the effective remaining lifespan is capped by those dates. A device without security updates has reduced practical value regardless of physical condition.
3. Environmental Impact (15% default weight)
Repair almost always has a lower carbon footprint than manufacturing a new device. This factor quantifies the COâ‚‚ difference.
- > 75% CO₂ savings from repair → strong repair signal
- 50-75% → moderate repair signal
- 25-50% → slight repair signal
- < 25% → neutral to slight replace signal
Energy efficiency: For appliances, we account for energy savings from newer, more efficient models. If a new appliance saves significant energy over its lifetime, that COâ‚‚ reduction offsets some of the manufacturing emissions.
4. Parts Availability (15% default weight)
A device can't be repaired if parts don't exist. This factor checks whether replacement parts are accessible and from which sources.
- Available (multiple sources) → strong repair signal
- Aftermarket only → moderate repair signal
- Limited (scarce or expensive) → slight repair signal
- OEM only (manufacturer parts, typically pricier) → neutral
- Unavailable → strong replace signal
The score is scaled by a confidence multiplier reflecting how certain we are about parts availability for your specific device.
5. Repair Success Rate (10% default weight)
Historical data on how often this type of repair succeeds, drawn from community repair records.
- 95-100% success → strong repair signal
- 85-95% → high repair signal
- 75-85% → good repair signal
- 65-75% → moderate, some failures
- < 65% → significant failure risk, favours replace
Rates are shown for both DIY and professional repair paths. When data confidence is low, the score impact is reduced.
Weight Presets
You can adjust how much each factor matters by choosing a preset that matches your priorities. Each preset changes the weight distribution:
| Preset | Cost | Lifespan | Environ. | Parts | Success |
|---|---|---|---|---|---|
| Balanced | 40% | 20% | 15% | 15% | 10% |
| Budget-first | 45% | 15% | 5% | 23% | 13% |
| Eco-friendly | 15% | 15% | 30% | 25% | 15% |
| Longevity | 20% | 35% | 10% | 20% | 15% |
Confidence Scoring
Every recommendation includes a confidence score (0-100%) that reflects both data quality and decision clarity. Higher confidence means we have better data and the decision is more clear-cut.
What contributes to confidence
- Category data (16 points). Having repair data specific to your device category
- Model data (12 points). Model-specific repair costs are more accurate than category averages
- Cost data quality (up to 10 points). Model-specific (10), issue-matched (8), crowdsourced (6), category average (3)
- Parts availability (10 points). Knowing if parts exist is critical for repair feasibility
- Success rate data (10 points). Community repair outcome data from the Open Repair Alliance
- Environmental data (6 points). COâ‚‚ manufacturing and repair figures
- Detailed description (6 points). The more detail you provide about your issue, the better
- Decision clarity (up to 30 points). A score far from zero (clear-cut decision) adds more confidence
Confidence levels
- 80-100%: High. Comprehensive data, clear decision
- 60-79%: Moderate. Some data gaps or borderline decision
- 40-59%: Low. Significant data gaps
- 0-39%: Very low. Limited data, treat as a rough guide
Data Sources
Our recommendations draw from multiple independent data sources to provide the most accurate picture possible.
Repair cost estimates (priority order)
- Warranty. If your device is under warranty, repair cost is zero
- Issue-matched. If your specific issue matches a known common problem for your device, we use the repair cost associated with that issue
- Model-specific. Repair cost range for your exact device model from our database
- Community-reported. Crowdsourced repair costs from real users who reported what they paid
- Category average. Generic category-level estimates when no device-specific data exists
Open Repair Alliance (ORA)
Over 305,000 community repair records from openrepair.org. ORA data provides repair success rates by problem type, common failure patterns, and barriers to repair across device categories.
iFixit repairability scores
Professional teardown-based repairability scores covering 350+ devices on a 1-10 scale. Scores evaluate fastener types, adhesive use, modularity, and component accessibility. We also use iFixit data to estimate parts availability and DIY difficulty.
France Indice de Réparabilité
Government-mandated repairability scores from France's Indice de Réparabilité program. Manufacturers selling in France must publish repairability scores (0-10) with sub-scores for documentation, disassembly ease, parts availability, parts pricing, and category-specific criteria. Our dataset includes 2,109 products from 33 manufacturers, covering laptops, dishwashers, vacuums, washing machines, and smartphones. This is verified, government-regulated data, not estimates. Note: not all manufacturers participate, so coverage varies. Apple and Samsung are not currently in the dataset. Source: data.gouv.fr.
Environmental data
- Manufacturing COâ‚‚. From manufacturer sustainability reports (Apple, Dell, HP) and peer-reviewed life cycle assessment studies
- Repair COâ‚‚. Category-level averages for the carbon footprint of repair activities
- Energy efficiency. EPA Energy Star data for appliances to calculate lifetime energy savings
- Grid carbon intensity. Regional electricity carbon intensity (defaults to 0.4 kg COâ‚‚/kWh, the US average)
Category defaults
When device-specific data is unavailable, we fall back to category-level defaults for typical repair costs, expected lifespans, and environmental figures. These defaults are maintained for electronics, appliances, and kitchen device categories.
About Our Data
We strive for accuracy, but all estimates come with limitations. Here's what you should know:
Where cost data comes from
Repair cost estimates use an additive model: parts cost (from iFixit listings and Keepa pricing data, with standard shop markup) plus a labour tier based on repair complexity (calibrated against real independent repair shop rates). Community-reported costs from Sundr users and device-specific data supplement the model. All prices are stored in USD, with CAD conversion at display time.
Limitations
- Cost estimates are not quotes. Actual repair costs vary by location, technician, and the specific condition of your device. Always get a real quote before committing.
- Category-level data is less precise. When we don't have model-specific data, we use category averages. The confidence score reflects this. Lower confidence means we're working with less specific data.
- Environmental figures are estimates. COâ‚‚ calculations rely on manufacturer reports and academic studies that may not perfectly match your specific device or region.
- Success rates reflect historical averages. Individual repair outcomes depend on the specific device condition, repair skill level, and parts quality.
- Data has varying freshness. Parts pricing syncs hourly via Keepa, repairability scores sync with iFixit and ORA as new data is published, and environmental data is updated annually. See our About Our Data page for the full refresh cadence.
Data Confidence Levels
Every result page shows a data confidence indicator so you know how much data backs your recommendation:
- Strong data. Model-specific repair costs plus at least two supporting sources (iFixit scores, community reports, ORA success rates, or parts data)
- Good estimate. Some device-specific data available (model data, community reports, or ORA records)
- Rough estimate. Relying primarily on category-level averages; results should be treated as a general guide
How This Content Is Made
Sundr is built and maintained by Keith, a software engineer. The scoring model, data analysis, and editorial calls are all mine. I use AI to help draft some of the written content, but everything gets reviewed against the actual data before it goes live. If a number shows up on this site, it traces back to a real source.
Last updated: