Establishing Causality in product management

# The 9 Bradford-Hill Criteria for Establishing Causality: Industry Examples

The Bradford-Hill criteria, established by Sir Austin Bradford Hill in 1965, provide a framework for determining whether an observed association is likely to be causal. These nine criteria have become fundamental tools in epidemiology, but their application extends far beyond healthcare into various industries. This article explores each criterion with three practical examples from different sectors.

## 1. STRENGTH: How large is the association?

The strength criterion examines the magnitude of the observed association between a potential cause and effect.

**Healthcare Industry:** The exceptionally strong association between smoking and lung cancer (smokers are 15-30 times more likely to develop lung cancer than non-smokers) was one of the first compelling pieces of evidence that smoking causes cancer.

**Technology Industry:** A strong correlation exists between screen time before bed and sleep disruption. Studies show individuals using screens within an hour of bedtime are 2-3 times more likely to experience poor sleep quality.

**Agriculture Industry:** The relationship between neonicotinoid pesticide use and bee colony collapse shows a strong association, with exposed colonies experiencing 50-80% higher collapse rates in controlled studies.

## 2. CONSISTENCY: Can you reproduce the results broadly?

Consistency evaluates whether the same association can be observed across different populations, methods, and circumstances.

**Finance Industry:** The relationship between interest rate increases and decreased consumer spending has been consistently observed across different countries, economic cycles, and demographic groups.

**Manufacturing Industry:** The connection between ergonomic workstation design and reduced repetitive strain injuries has been consistently demonstrated across various manufacturing environments, from electronics to automotive.

**Education Industry:** The positive relationship between early literacy intervention and improved reading outcomes has been consistently reproduced across different countries, socioeconomic groups, and educational systems.

## 3. SPECIFICITY: How many other associations are in play?

Specificity examines whether the observed effect is associated specifically with the suspected cause rather than multiple other factors.

**Automotive Industry:** Vehicle crash investigations show a specific relationship between tire tread depth below 2/32 inch and hydroplaning accidents, with minimal confounding factors on dry roads.

**Food Industry:** Specific food contamination outbreaks can be traced to particular pathogens, such as the clear link between certain batches of romaine lettuce and E. coli outbreaks, with no other explanation for the cluster of cases.

**Pharmaceutical Industry:** Certain medications show highly specific side effects—for example, the direct and specific relationship between clozapine usage and agranulocytosis, which is rarely caused by other factors in the patient population.

## 4. TEMPORALITY: Observe, infer, act, repeat

Temporality confirms that the cause precedes the effect in time, establishing the correct chronological sequence.

**Environmental Industry:** The temporal sequence of industrial chemical release into waterways followed by downstream wildlife population declines establishes correct chronology before inferring causation.

**Advertising Industry:** Companies observe sales increasing only after launching specific marketing campaigns, establishing the temporal relationship needed to attribute the sales boost to the campaign.

**Workplace Safety Industry:** Implementation of new safety protocols precedes the reduction in workplace accidents, establishing the temporal relationship necessary for attributing improved safety to the protocols.

## 5. GRADIENT (BIO.): Does greater exposure to the IV lead to greater effect in the DV?

The biological gradient (dose-response relationship) examines whether greater exposure leads to proportionally greater effects.

**Radiation Industry:** Workers with higher radiation exposure show proportionally higher rates of certain cancers, demonstrating a clear dose-response relationship.

**Nutrition Industry:** Studies show a gradient relationship between sugar consumption and obesity rates, with populations consuming progressively higher amounts of sugar showing correspondingly higher BMI averages.

**Chemical Industry:** Exposure to different concentrations of industrial solvents shows a clear gradient of liver enzyme abnormalities, with higher concentrations producing proportionally greater enzyme disruptions.

## 6. PLAUSIBILITY: Do you have a hypothesis on cause-effect that doesn't seem crazy?

Plausibility evaluates whether there is a reasonable and logical mechanism that could explain how the cause produces the effect.

**Renewable Energy Industry:** The relationship between increased solar panel efficiency and reduced electricity costs has a plausible mechanism: more efficient energy conversion results in more electricity generated per dollar invested.

**Construction Industry:** The association between proper building insulation and reduced energy consumption has a plausible mechanism: better insulation reduces thermal transfer, requiring less energy for heating and cooling.

**Transportation Industry:** The relationship between traffic roundabouts and fewer serious accidents has a plausible mechanism: roundabouts force speed reduction and eliminate the possibility of high-speed T-bone collisions.

## 7. COHERENCE: Do you generally observe the effect elsewhere or just in your experiments?

Coherence assesses whether the cause-effect relationship aligns with existing knowledge and observations in the broader world.

**Telecommunications Industry:** The relationship between increased network bandwidth and improved video streaming quality coheres with the known principles of data transmission and digital media processing.

**Sports Industry:** The association between specialized training programs and improved athletic performance coheres with established principles of exercise physiology seen across different sports and competition levels.

**Retail Industry:** The relationship between store layout changes and customer purchasing patterns coheres with established consumer psychology principles observed in various retail environments.

## 8. EXPERIMENT: Does experimental evidence (at least) support the inference?

The experiment criterion examines whether controlled experimental interventions support the causal relationship.

**Gaming Industry:** A/B testing different reward systems in games demonstrates through controlled experimentation that certain reward schedules increase player retention more than others.

**Healthcare Technology Industry:** Randomized controlled trials show that implementing electronic health record systems with specific design features reduces medication errors compared to traditional paper-based systems.

**Hospitality Industry:** Controlled experiments comparing different room pricing strategies conclusively demonstrate that dynamic pricing algorithms increase overall revenue compared to fixed pricing models.

## 9. ANALOGY: Do you see similar effect-cause patterns elsewhere?

Analogy considers whether similar cause-effect relationships have been established in comparable situations.

**Cybersecurity Industry:** The relationship between implementing multi-factor authentication and reduced account breaches is analogous to the proven effectiveness of multiple locks on physical doors preventing break-ins.

**Aerospace Industry:** The causal relationship between composite materials and fuel efficiency in aircraft is analogous to the same relationship observed in automotive racing and maritime applications.

**Healthcare Administration Industry:** The impact of checklist protocols on reducing surgical complications is analogous to the proven effect of standardized procedures reducing errors in aviation and nuclear power operations.

## Conclusion

The Bradford-Hill criteria provide a robust framework for evaluating causality across diverse industries. While meeting all nine criteria isn't always necessary to establish causation, the more criteria that are satisfied, the stronger the case for a causal relationship. These examples demonstrate how these principles can be applied beyond epidemiology to strengthen causal inference in business, technology, manufacturing, and numerous other sectors.