01The ten that hold on hardest
Every used-car conversation in Malaysia eventually reaches the same question: which cars actually keep their money? We can answer it from our own market data rather than from folklore—and the answer is less about badges than most buyers expect.
340,782
Listings in the dataset
Malaysian advertised asking prices, snapshot of 14 August 2026
1,587
Make-and-model cohorts examined
392 produced a valid curve; 52 cleared the publication guards
30
Models corroborated by two methods
Only these are eligible to be named in a ranking
| # | Model | 5-year retention | Curve method | Price-ladder method | Listings |
|---|---|---|---|---|---|
| 1 | Perodua Myvi | 79.4% | 79.4% | 79.5% | 5,811 |
| 2 | Perodua Alza | 77.2% | 80.8% | 73.6% | 3,320 |
| 3 | Honda City | 77.2% | 75.3% | 79.0% | 4,536 |
| 4 | Toyota Yaris | 75.9% | 75.3% | 76.5% | 1,047 |
| 5 | Proton Iriz | 75.2% | 73.8% | 76.5% | 878 |
| 6 | Perodua Axia | 74.0% | 70.1% | 78.0% | 3,647 |
| 7 | Toyota Corolla Altis | 73.4% | 77.7% | 69.1% | 1,280 |
| 8 | Mazda CX-30 | 72.8% | 70.4% | 75.3% | 668 |
| 9 | Toyota Vios | 72.2% | 69.4% | 75.1% | 5,016 |
| 10 | Perodua Aruz | 71.4% | 65.2% | 77.7% | 1,295 |
02Why cheap, common cars win this table
Every one of the top ten is a national or mainstream Japanese model, and most of them are the kind of car you see in every car park in the country. That is not a coincidence, and it is not really about build quality.
What the top of this table has in common:
- Deep, continuous supply. The Myvi alone is backed by 5,811 listings across 223 model-year cohorts. Thick markets price efficiently and leave less room for a seller to be badly wrong.
- Predictable running costs. Servicing, parts, and road tax on a 1.3-litre national hatchback are known quantities, so a buyer five years from now does not have to price in uncertainty.
- A large pool of next buyers. Value retention is ultimately about who will want the car later. Broad affordability keeps that pool wide.
- Low purchase price to begin with. There is simply less value available to lose, which flatters the percentage—see the ringgit section below.
The two genuine surprises
The Mazda CX-30 at 72.8% is the strongest non-national, non-Toyota/Honda entry in the table, and it does it on the smallest sample here (668 listings). Treat it as a real but less certain result than the Myvi's.
The MINI Cooper shows the widest disagreement between our two methods of any car that still qualified: 62.1% on the curve against 76.8% on the price ladder. We publish it at the midpoint—69.4%, just outside this top ten—and flag the spread rather than picking the flattering number.
03Retention in percent versus retention in ringgit
This is the part that changes decisions. A high retention percentage on an expensive car can still cost you far more money than a mediocre percentage on a cheap one.
| Model | Typical near-new ask | Typical 5-year-old ask | Value lost | Per year |
|---|---|---|---|---|
| Perodua Axia | RM35,166 | RM27,444 | RM7,722 | RM1,544 |
| Proton Saga | RM30,909 | RM22,943 | RM7,966 | RM1,593 |
| Proton Iriz | RM34,558 | RM26,421 | RM8,137 | RM1,627 |
| Perodua Myvi | RM46,050 | RM36,605 | RM9,445 | RM1,889 |
| Perodua Aruz | RM56,877 | RM44,184 | RM12,693 | RM2,539 |
| Honda City | RM68,765 | RM54,352 | RM14,413 | RM2,883 |
| Perodua Alza | RM60,409 | RM44,455 | RM15,954 | RM3,191 |
| Toyota Vios | RM69,215 | RM51,980 | RM17,235 | RM3,447 |
| Mazda CX-30 | RM104,310 | RM78,509 | RM25,801 | RM5,160 |
| Toyota Corolla Altis | RM106,150 | RM73,299 | RM32,851 | RM6,570 |
The Perodua Axia sits sixth on the percentage table but loses the least money of any car we ranked—about RM1,544 a year. The Toyota Corolla Altis sits seventh on percentage and loses more than four times as much per year. If your goal is to protect ringgit rather than to win a percentage, read this table first and the ranking second.
04How the ranking was built, and what got a car disqualified
Rankings are only as good as the things they refuse to publish. We started from 1,587 make-and-model cohorts in the 14 August 2026 snapshot and finished with 30 that we were willing to name. This ranking is editorial analysis of asking-price patterns — it is not how a Carvaly valuation is calculated. Understanding your Carvaly valuation explains how to read a result.
- 1
Fit a curve across model years, not across time
For each model we fit a decay curve across its model-year cohorts within one snapshot. This is the only defensible way to estimate depreciation from marketplace data with a short observation history.
- 2
Throw out the fallback assumptions
Where we could not fit a model from observed asking prices, we would otherwise have used a segment-wide assumption. Those rows are an assumption, not evidence about that car, and none of them appear in this ranking.
- 3
Require depth, and require it in the right place
A model needed at least 6 model-year cohorts, 200 listings, and a 6-year age span. Critically, it also needed at least 4 cohorts and 200 listings aged 0–6 years, so a five-year claim rests on cars that are actually that age.
- 4
Reject curves that hit the guardrails
Our estimator clamps implausible rates. If a model's curve was sitting on that floor, the data did not support a rate at all—it produced a boundary, not a measurement. Those models were removed, and several would otherwise have topped this table with false 97% figures.
- 5
Corroborate with a second, independent method
We separately computed the observed price ladder: the weighted median asking price of 0–1 year-old cars against 4–6 year-old cars. Models were published only where the two methods agreed within 15 percentage points.
05What this ranking cannot tell you
Being straight about the limits is what makes the rest of it usable.
- These are asking prices. Malaysian marketplace listings show what sellers want, not what buyers finally paid. Negotiation sits underneath every figure here.
- A model is not a car. A ranking chooses a shortlist. Year, variant, engine, transmission, mileage, accident history, service record, and condition decide what one specific vehicle is worth.
- Cohort mix moves with the market. If a model's newer listings skew towards a higher trim than its older ones, the ladder overstates retention. This is the main reason we run two methods.
- Absence is not failure. Many models—including most Korean marques and most EVs—simply lacked the depth to clear our gates. That is a statement about our evidence, not about those cars.
- Nothing here is a forecast. Retention describes the market as it stands on 14 August 2026. Facelifts, discontinuations, tax changes, and new-car discounting all move it.
Frequently asked questions
Which car holds its value best in Malaysia?
In our 14 August 2026 analysis the Perodua Myvi ranked first, retaining an estimated 79.4% of its value over five years, backed by 5,811 marketplace listings. The Perodua Alza and Honda City followed at 77.2%. These are estimates from asking-price patterns across model-year cohorts, and each is published with its sample size.
Does a high retention percentage mean I lose less money?
No, and this is the most common mistake. Retention is a percentage of the purchase price, so an expensive car with strong retention can still cost far more in ringgit. A Perodua Axia loses roughly RM1,544 a year in our data while a Volvo XC90 loses roughly RM29,617. Always read the percentage next to the price.
Why are only 30 cars ranked when there are hundreds of models?
Because most cohorts cannot support an honest claim. Of 1,587 make-and-model cohorts, 392 produced a valid curve, 52 cleared our depth and quality gates, and 30 were corroborated by a second independent method. The rest were excluded for thin samples, poor fits, boundary-clamped curves, or disagreement between methods.
Is this based on actual sale prices?
No. It is based on advertised asking prices in Malaysia. Asking prices are a strong current signal of supply and seller expectation, but they are not completed transactions. We label them as asking prices everywhere rather than implying we observe final sale values.
Why isn't my car in the list?
Most likely it lacked depth in the right places—particularly recent model years. A five-year retention claim needs cars aged 0–6 in real numbers, and many models are only well represented at older ages. Exclusion here means we could not measure the model confidently, not that it holds value poorly.
Sources and references
- Carvaly: Understanding your Carvaly valuation
- Carvaly Market Intelligence: Malaysia Used-Car Market Report · July 2026
- data.gov.my: Car Registration Transactions
- Malaysian Automotive Association (MAA): Production and Sales for May 2026
- Perodua: Perodua e-Owner's Manuals
- ASEAN NCAP: Vehicle Safety Results