What if a brand had to survive 1,000 years? The survival maths

A typical listed company has a half-life of about ten years. To reach a thousand, a brand needs a death rate 84 times lower, and eleven Japanese firms show how.

The five-storey pagoda of Shitenno-ji temple in Osaka, the temple Kongo Gumi was founded to build
Photo: Pufacz, CC0, via Wikimedia Commons

A brand can survive 1,000 years only if its yearly chance of dying stays below about 0.07%, which is roughly 84 times lower than the 5.8% a year measured for 26,561 US-listed companies. At the normal rate, the odds of lasting a millennium are about 1 in 1025. Yet Teikoku Databank counts 11 Japanese companies that are already over 1,000 years old, and they share a pattern: they are tied to something that itself lasts (a temple, a hot spring, a mint), they carry little debt, and they pick successors by merit rather than by birth.

Key takeaways
  • A typical listed company has a half-life of about 10 years and a death rate that does not fall with age (Daepp et al., 2015).
  • At 5.8% a year, surviving 1,000 years has odds of about \( e^{-58} \approx 6\times10^{-26} \).

    Calculated in this post from the 5.8% yearly death rate (Daepp et al., 2015) and a 1,000-year target.

  • For even odds over a millennium, the total yearly hazard must be 0.069% or less.
  • Japan has 45,284 firms over 100 years old and 11 over 1,000 (Teikoku Databank).
  • The oldest survivors hang off a durable anchor: temples, springs, abbeys, the state's coinage.
  • Kongo Gumi lasted 1,428 years and then died of one modern disease: debt.

    Calculated in this post from the founding year 578 and the January 2006 sale of its assets (Wikipedia: Kongo Gumi).

Most brand strategy is written on a three-year horizon, and most brand "heritage" is a font choice. This post asks the opposite question with a straight face. If a board told you the brand must still be trading in the year 3026, what would the numbers say, and what would you change on Monday?

The answer comes from survival analysis, the same maths that actuaries use for human lives and engineers use for bearings. It has two candidate models, a clean dataset of real millennium-old firms, and one famous death that teaches more than all the survivors.

How long does a company normally live?

A publicly listed company lives about a decade in the median case: the measured half-life is 10.5 years (Kaplan-Meier; Daepp et al., 2015) and the yearly death rate about 5.8%. The surprising part is that the rate does not drop as the firm gets older.

The mortality of companiesMadeleine I. G. Daepp, Marcus J. Hamilton, Geoffrey B. West, Luis M. A. Bettencourt · Journal of the Royal Society Interface · 2015

The team took every company in the Compustat database of US-traded firms from 1950 to 2009, 26,561 after cleaning, and treated each as a life that ends when it stops reporting sales. Lifespans followed an exponential distribution across sectors. The maximum-likelihood death rate was 0.058 a year; Kaplan-Meier gave a half-life of 10.46 years. Mergers and acquisitions caused 45.1% of deaths and bankruptcy only 4.5%, so "death" here often means being bought. The limit: these are listed firms in one country over 60 years, and a takeover is counted as a death even when the brand lives on.

Read the paper (free)

An exponential lifespan means a constant hazard rate. In plain words, a 40-year-old listed company is as likely to die next year as a 4-year-old one. Age buys nothing. That is the first model, and it is brutal for our question.

The story

In 578, a carpenter named Shigemitsu Kongo founded a family guild of temple builders, the firm whose work is tied to Shitenno-ji, the Buddhist temple in what is now Osaka. It would build and rebuild Japan's temples for the next fourteen centuries. A family scroll about 3 metres long traces 40 generations of leaders; when no son fit the job, a son-in-law took the Kongo name. The firm rebuilt Shitenno-ji's pagoda after a 1934 typhoon under Yoshie Kongo, its first woman head carpenter.

Then came the 1980s property bubble. Kongo Gumi borrowed against land that later fell in value. By 2005 it had revenue of about ¥7.5 billion and roughly 100 staff, and it could not service its debt. In January 2006 Takamatsu Construction bought the business, and the old entity filed for bankruptcy that July owing about ¥4 billion. The name still trades, as a subsidiary.

Wikipedia: Kongo Gumi (citing Bloomberg, Nikkei Asia and the company's own history)
Exterior of Nishiyama Onsen Keiunkan, a hot-spring inn in Yamanashi, Japan
Nishiyama Onsen Keiunkan in Yamanashi traces its founding to 705. Its product, a hot spring, has not changed in 1,300 years, which is exactly the point.Photo: Boltor, CC BY-SA 4.0, via Wikimedia Commons

What are the odds of a brand surviving 1,000 years?

Under the constant-hazard model, the odds of a normal company reaching 1,000 years are about 6 in 1026. You would need far more companies than have ever existed to expect one survivor. Here is the working.

Calculated in this post from the 5.8% yearly death rate (Daepp et al., 2015) and a 1,000-year target.

Assumptions and variables

  • \( \lambda \) is the hazard rate: the chance per year that the firm dies, given it is alive.
  • \( S(t) \) is the survival function: the probability the firm is still alive at age \( t \).
  • Baseline \( \lambda = 0.058 \) per year, from Daepp et al.; I treat it as constant, as their data show.
  • Target: \( t = 1{,}000 \) years.
\[ \frac{dS}{dt} = -\lambda S \quad\Rightarrow\quad S(t) = e^{-\lambda t} \]
  1. Plug in the baseline. \( S(1000) = e^{-0.058 \times 1000} = e^{-58} \approx 6.5\times10^{-26} \).
  2. Check the half-life. \( t_{1/2} = \ln 2 / \lambda = 0.693/0.058 = 11.9 \) years, close to the paper's Kaplan-Meier half-life of 10.46 years.
  3. Invert for the target. For even odds at 1,000 years, \( e^{-\lambda \cdot 1000} = 0.5 \), so \( \lambda = \ln 2/1000 = 6.93\times10^{-4} \), or 0.069% a year.
  4. Compare. \( 0.058 / 0.000693 \approx 84 \). The millennium brand must be 84 times harder to kill than a normal listed firm.

    Calculated in this post from the 5.8% yearly death rate (Daepp et al., 2015) and the 0.069% a year worked out in the previous step.

\[ \lambda_{\max} = \frac{\ln 2}{1000\ \text{yr}} = 0.000693\ \text{yr}^{-1} \;\approx\; \frac{0.058}{84} \]
A normal listed company is as likely to die at 40 as at 4. Age buys nothing, unless you change what kills you.

Use this to check what you just read. Move the hazard rate and see how fast a millennium disappears.

Survival odds for your brand

Is a company's death rate really constant with age?

For listed firms over 60 years, yes; for family firms over centuries, the Japanese data suggest the death rate falls sharply with age. That points to a second model, Lindy's law, where the older something is, the longer it is expected to last.

Teikoku Databank, Japan's largest corporate credit agency, publishes counts of long-lived firms. The latest figures, reported in January 2026, are below.

Japanese companies by age, Teikoku Databank count
See the data as a table
Companies (firms)
100+ years45284
200+ years1813
300+ years889
500+ years47
1,000+ years11

Source: Teikoku Databank, as reported by Asia Economy (Jan 2026) · Counts are cumulative: a 1,000-year firm is also in every other bar.

Reading the hazard out of the counts

If firms of every era were founded at the same rate, the ratio between two bars is a survival probability. That assumption is shaky, because far fewer firms were founded in 1026 than in 1926, so treat what follows as an estimate with that bias built in.

  1. 300 to 500 years. \( 47/889 = 0.053 \) survive 200 years, so \( \lambda \approx -\ln(0.053)/200 = 0.0147 \), or 1.5% a year.

    Calculated in this post from Teikoku Databank counts of 47 firms over 500 years and 889 over 300 years.

  2. 500 to 1,000 years. \( 11/47 = 0.234 \) survive 500 years, so \( \lambda \approx -\ln(0.234)/500 = 0.0029 \), or 0.29% a year.

    Calculated in this post from Teikoku Databank counts of 11 firms over 1,000 years and 47 over 500 years.

  3. Result. The implied hazard drops about fivefold between the middle and the top of the age range. A constant-hazard world would show no drop at all.

The Lindy model, worked

Lindy's law says the expected remaining life is proportional to current age: \( E[X - L \mid X \ge L] = aL \). Balbi and Grimaldi show this forces a power-law survival curve.

\[ S(L) \propto L^{-(1+1/a)} \quad\Rightarrow\quad \frac{S(T)}{S(t_0)} = \left(\frac{T}{t_0}\right)^{-(1+1/a)} \]
  1. Fit the exponent from 100 to 1,000 years. \( 45{,}284/11 = 4{,}117 \), and \( \ln 4117/\ln 10 = 3.61 \). So \( 1 + 1/a = 3.61 \) and \( a \approx 0.38 \).

    Calculated in this post from Teikoku Databank counts of 45,284 firms over 100 years and 11 over 1,000 years.

  2. Fit it from 500 to 1,000 years. \( \ln(47/11)/\ln 2 = 2.09 \), so \( a \approx 0.92 \): an old firm can expect to live almost as long again as it already has.
  3. Apply it. A 100-year-old brand with \( a = 0.38 \) reaches 1,000 years with probability \( 10^{-3.61} \approx 0.025\% \): one in about 4,000, not one in \( 10^{25} \).

    Calculated in this post from Teikoku Databank counts of 45,284 firms over 100 years and 11 over 1,000 years.

Technosignatures longevity and Lindy's lawAmedeo Balbi, Claudio Grimaldi · The Astronomical Journal 167, 119 · 2024

The authors ask how long an alien radio signal is likely to last if signal lifetimes follow Lindy's law. They formalise the law as expected remaining life equal to a constant times current age, and derive that the lifetime distribution must be a power law with exponent \( \alpha = 2 + 1/a \). With \( \alpha = 3 \), a detected signal has a 95% chance of lasting under 20 years. Their maths is general survival theory, which is why it transfers to firms. The limit: it is a theoretical paper about signals, and applying it to companies is my step, not theirs.

Read the paper (free)
Two ways to age: constant hazard vs Lindy 100% 10⁻⁶ 10⁻¹⁵ 10⁻²⁶ 0 500 1,000 years Constant 5.8%/yr Lindy, from age 100 0.069%/yr keeps 50% Vertical axis is logarithmic; the curves are illustrative of the fitted numbers
On a log scale, a constant hazard is a straight dive. A Lindy curve flattens, because each century survived makes the next more likely. The dashed line is the target.

Which companies have already lasted 1,000 years?

The world's oldest operating firms are mostly Japanese inns, temple builders and religious-goods makers, plus two European mints, an abbey restaurant and a bell foundry. The ages below are for 2026, from the founding years in the standard list.

Oldest companies still operating, age in 2026
See the data as a table
Age (years)
Kongo Gumi (construction, Japan)1448
Nishiyama Onsen Keiunkan (inn, Japan)1321
Koman (inn, Japan)1309
Hoshi Ryokan (inn, Japan)1308
Genda Shigyo (ceremonial paper, Japan)1255
St. Peter Stiftskulinarium (restaurant, Austria)1223
Staffelter Hof (wine, Germany)1164
Monnaie de Paris (mint, France)1162
Tanaka-Iga (religious goods, Japan)1141
The Royal Mint (mint, UK)1140
Pontificia Fonderia Marinelli (bells, Italy)986

Source: Wikipedia list of oldest companies (founding years 578 to 1040) · Founding dates this old often rest on tradition; Keiunkan restructured into a new company in 2017.

CompanyFoundedWhat it sellsThe anchor it rides on
Kongo Gumi578Temple constructionBuddhist temples that need rebuilding every few generations
Nishiyama Onsen Keiunkan705Hot-spring innA geothermal spring
Hoshi Ryokan718Hot-spring innA spring; the Hoshi family is in its 46th generation
St. Peter Stiftskulinarium803RestaurantSt Peter's Abbey, Salzburg
Monnaie de Paris864Coins and medalsThe French state
The Royal Mint886CoinsThe British Crown
Pontificia Fonderia Marinelli1040Church bellsThe Catholic Church

Read the last column. Every survivor is a supplier to something older and slower than any market: a religion, a state, a piece of geology. Their demand curve was set by institutions that do not change their minds every quarter.

Historic minting room inside the Monnaie de Paris with coin presses
The Monnaie de Paris, founded in 864. Its one customer, the French state, has changed regime many times but has never stopped needing coins.Photo: DiscoA340, CC BY-SA 4.0, via Wikimedia Commons

What kills old brands, and how do you remove each risk?

The total hazard is the sum of separate hazards, so a 1,000-year brand needs every single one of them under control. One leak caps the whole tank.

\[ \lambda_{\text{total}} = \lambda_{\text{demand}} + \lambda_{\text{succession}} + \lambda_{\text{debt}} + \lambda_{\text{catastrophe}} + \lambda_{\text{acquisition}} \le 0.000693 \]

Sensitivity: one bad term is enough

  • If any single term is 1% a year, survival over 1,000 years is at most \( e^{-10} = 0.0045\% \), however perfect the rest.
  • If the five terms share the budget equally, each must be under 0.014% a year: one failure in about 7,200 years per risk.
  • Halve the target to 500 years and the budget doubles to 0.139% a year; the problem stays hard, not easy.
  • Kongo Gumi beat demand, succession and catastrophe risk for 14 centuries (Wikipedia: Kongo Gumi), then a single burst of leverage undid it.
The hazard budget: five taps, one bucket Demandanchor to an institution Successionadopt by merit Debtkeep it near zero Catastrophespread sites, records Acquisitionlock ownership Total hazard ≤ 0.069% a year Kongo Gumi's leak was the debt tap, opened in the 1980s bubble
Hazards add. A brand can be perfect on four of them and still die of the fifth, which is what happened to the oldest firm in the world.

The five controls, in practice

  1. Demand: sell to an anchor. Tie the product to a need set by something slow: ritual, geography, law, the body. Hot water, bells and coins have outlasted every fashion in the same centuries.
  2. Succession: widen the gene pool. Japan's custom of adopting a capable son-in-law, mukoyoshi, let Kongo Gumi pass leadership on ability for 40 generations without leaving the family name.
  3. Debt: refuse leverage. A firm with no debt cannot be foreclosed in a crash. This is the term that killed the record holder.
  4. Catastrophe: duplicate. Fire, war and flood are local; skills, records and sites should not be.
  5. Acquisition: decide who may buy you. In the Daepp data, 45% of "deaths" were takeovers. A charter, a trust or a family holding company removes that tap.
Traditional wooden building of Hoshi Ryokan in Komatsu, Ishikawa
Hoshi Ryokan, founded in 718, has been run by one family for 46 generations. Its brand is the place itself.Photo: Namazu-tron, CC BY-SA 3.0, via Wikimedia Commons

Can a brand outlive its company?

Yes, and that is the most practical finding here: a brand name can survive the death of the firm that made it, because a buyer pays to keep it. The Kongo Gumi name has traded under Takamatsu since 2006. The Daepp data count that as a death; a brand manager would count it as a transplant.

That changes the maths. If a brand has a second life whenever its owner dies, the brand's hazard is the owner's hazard times the chance nobody wants the name. A name with real memory in buyers' heads, the kind Indian brands measured in BrandZ rankings carry, has a high rescue value.

\[ \lambda_{\text{brand}} = \lambda_{\text{firm}} \times (1 - p_{\text{rescue}}) \]

With \( \lambda_{\text{firm}} = 0.058 \), reaching the 0.000693 target needs \( p_{\text{rescue}} \ge 0.988 \). In words: every time the owner dies, a buyer must step in to keep the name about 99 times in 100. That is plausible only for names with a meaning that buyers can't get from anyone else.

Calculated in this post from the 5.8% yearly death rate (Daepp et al., 2015) and the 0.069% a year target worked out above.

India's own long runners

India's oldest firms on the standard list are far younger: the Wadia Group dates to 1736, Forbes & Company to 1767, EID Parry to 1788, and Chennai's Higginbotham's bookshop to 1844. At about 290 years old, the Lindy fits above give Wadia roughly a 1 in 13 chance of seeing 1,000 (500-to-1,000-year fit) or about 1 in 90 (100-to-1,000-year fit). India's ad history shows how fast brand memory forms and fades; see Hamara Bajaj and the Amul girl for two that keep renewing it.

Calculated in this post from the 1736 founding year (Wikipedia: List of oldest companies in India) and the fitted exponents 2.09 and 3.61 above, which come from Teikoku Databank counts.

Verdict: possible with changes

A normal company cannot last 1,000 years: at the measured 5.8% yearly hazard, the odds are about 6 in 1026 (calculated above). It becomes possible if the total yearly hazard falls below 0.069%, about 84 times lower, which eleven Japanese firms have already done. The design that works costs nothing to start but a great deal to keep: sell to a durable anchor, carry almost no debt, pick successors by merit, duplicate every critical asset and lock ownership. Treat the brand name as the part that can be rescued even when the firm cannot.

If you are building a brand for decades rather than quarters, the trade-off between long-term memory and short-term sales is the subject of brand vs performance marketing. I work on that balance in practice; see my work.

What if next?

Questions people ask

What is the oldest company in the world?

Kongo Gumi, a Japanese temple builder founded in 578, is usually listed as the oldest. It was absorbed by Takamatsu Construction in 2006 and still operates as a subsidiary. The oldest still independently run is often given as Nishiyama Onsen Keiunkan, a hot-spring inn dating to 705.

Why does Japan have so many old companies?

Teikoku Databank counts 45,284 Japanese firms over 100 years old. Common explanations include the adoption of capable heirs into family firms, a focus on one craft, and a culture that values continuity over growth. The data in this post add one more: many are tied to temples, springs and rituals with very stable demand.

How long does the average company last?

For US-listed firms from 1950 to 2009, the half-life was about 10.5 years, and the death rate did not fall with age. Around 45% of those "deaths" were mergers or acquisitions rather than failures, so many brands lived on under new owners.

What is the Lindy effect?

The Lindy effect is the idea that the expected remaining life of a non-perishable thing grows with its age. Mathematically it implies a power-law survival curve. Japan's old-firm counts fit it better than a constant death rate, though the fit is biased by how few firms were founded in early centuries.

Sources

  1. Daepp, Hamilton, West, Bettencourt: The mortality of companies, J. R. Soc. Interface (2015) (source for the 0.058 hazard, 10.46-year half-life, 26,561 firms and 45.1% M&A deaths)
  2. Balbi and Grimaldi: Technosignatures longevity and Lindy's law, The Astronomical Journal (2024) (source for the Lindy power-law derivation)
  3. Asia Economy: Japan faces succession crisis for 1,000-year companies (2026) (source for Teikoku counts: 45,284; 1,813; 889; 47; 11)
  4. Wikipedia: Kongo Gumi (source for 578 founding, 40 generations, ¥7.5 billion revenue, January 2006 transfer and ¥4 billion debt)
  5. Wikipedia: List of oldest companies (source for founding years in the chart and table)
  6. Wikipedia: List of oldest companies in India (source for Wadia 1736, Forbes 1767, Parry 1788, Higginbotham's 1844)
  7. Wikipedia: Hoshi Ryokan (source for 718 founding and 46 generations)
  8. Nippon.com: Japanese companies celebrating major anniversaries in 2026 (2026) (source for 2,371 centennials)
  9. Nippon.com: Japan's corporate bankruptcies top 10,000 in 2025 (2026) (source for 10,300 bankruptcies)
oldest companiesbrand longevityKongo Gumishinisehazard rateLindy effectsurvival analysis
Sapta

Sapta (Saptarshi Sarkar) is a mechanical engineer turned marketer in Bengaluru and a solo touring rider. He writes the seven Saptaverse blogs. Work with him or read more of Marketblog.

Keep reading in Marketblog

From the other blogs