Astrazeneca is a large pharmaceutical company focused on the development, manufacture and sales of ethical pharmaceutical products. Over a number of years, it has developed a sophisticated approach to managing its supply chains. This business interruption risk management programme aims to identify and measure key product supply chain exposures, so the inevitable trade-offs involved in risk mitigation can be made in an informed manner. The process is based on a rigorous analysis of supply chain dependencies, and it is embedded with the company's global teams responsible for the continuing supply of products.
This systematic and consistent risk quantification process allows formulation of risk improvement plans to address unacceptable risks in a prioritised manner. Estimates of risk exposures, therefore, need to be credible.
Historically AstraZeneca had adopted a traditional insurance industry approach of considering risk accumulations based on a worst-case loss estimates within specific CRESTA1 earthquake zones. Senior management had difficulty comprehending the magnitude and likelihood of such events and so tended not to respond. At the same time, reinsurers were starting to recognise that such approaches were difficult to reconcile with the way a natural catastrophe event might develop.
Japanese case study
Continuity is critical; the pharmaceutical industry is unique in the level of regulation that governs the registration of product manufacturing and the long lead times needed to gain approval for changes in intermediate sourcing.
Many of the intermediates used in the manufacture of pharmaceutical products are sourced from the fine chemical industry. Japan is a leading producer of technologically advanced fine chemicals and as such is a prime market for sourcing intermediates. The case study outlined here relates to the business exposures in Japan and specifically in the Osaka area (Kinki Region). The approach has, however, been applied elsewhere with equal success.
Detailed analysis of the company's supply chains identified over 50 dependencies on suppliers in Japan. Of these, over half were located in the greater Osaka area. A number of key global brands depended on these companies for supply. Clearly, a single earthquake event could not credibly affect 30 different locations with worst-case impact at each. But we wanted to find out what a credible scenario might be and what could be done about it.
Detailed study methodology
The AstraZeneca risk and insurance group, therefore, worked closely with seismic engineers from long term partners ABS Consultants and Jacobs Gibb to evaluate the multi-site impact potential of a range of earthquake events on the business. This complex analysis firstly involved locating the key dependencies and seeking detailed information about these production facilities.
The data collection challenge cannot be overstated. Of the 50 locations considered, only two were under AstraZeneca's direct control. Working with purchasing and quality assurance professionals was, therefore, critical in gathering site specific data.
We developed a modelling approach with the consultants which consisted of the following steps:
1) Accurately locate the exact geographic position of each site, then characterise each site according to the following criteria:
- Building construction type (for example, steel, reinforced concrete, masonry, low tensile concrete etc)
- Age of construction and building codes used
- Seismic resistance in design (foundation design, design methodology)
- Prevailing soil type (rock, clay, sand etc)
The potential for reclaimed, soft soil conditions for facilities in coastal locations was also taken into account.
2) Collect data by a variety of means including direct contact with the supplier organisations through site visits or email questionnaires.
Where possible we obtained site and plant layouts, photographs or processing information. The potential for earthquake induced fires increasing the severity of an incident was also considered, particularly in relation to the chemical plant sites.
3) Develop a model linking these risk factors to the potential for building damage under a range of earthquake intensities.
For each type of generic building, defined in terms of construction type and building height, we defined a range of results in terms of building damage as a function of localised earthquake damage intensity described by the Mercalli Modified Intensity (MMI) and Japanese Meteorological Association intensity (JMA).
The model produced a series of generic building damage curves as a function of earthquake intensity for each combination of building type and height.
We validated the model using the work of the Applied Technology Council (ATC), extrapolating from data available for California.2
4) Translate these building damage estimates into production stoppage times for which business interruption estimates could be calculated.
We estimated a physical damage outage time (in months) by reference to the ATC research. This outage time estimate was based on an average time to reinstate 100% of activity in the building. Each site was plotted against a generic building damage curve, based on its building characteristics.
5) Estimate two further delay factors to this stoppage time.
The first additional delay factor related to the time required to regain validated status under pharmaceutical industry regulations under good manufacture practice (GMP) guidelines. This reflects the potential delay associated with facilities being re-inspected by international medicines authorities, such as the US Food and Drug Administration (FDA).
Finally, an allowance was made for the impact of general economic disruption as a result of the earthquake. Detailed studies identified that the regional infrastructure could be severely disrupted, potentially leading to a delay before repairs could begin in the area of any facility.
6) Define earthquake damage intensities for a range of earthquake faults identified by consultants for each site at a number of earthquake return periods in the range 50 years through to 5,000 years.
7) Model the business interruption impact for the various product supply chains implicated in this analysis.
We used our business interruption risk management models for each supply chain to assess the net effect of each specific scenario at each location.
The total outage time estimate was used to calculate a gross margin loss allowing for existing mitigation. This mitigation took into account a range of factors including:
- Existing safety stocks (months forward cover)
- Location of these stocks, especially if in Japan
- Lead time to instigate alternative manufacturing facilities
- Capacity of these alternative manufacturing facilities relative to total demand
- Estimated market share impact following the loss over a protracted period
Where different production stages of the same product went through sites within the area studied, we evaluated the aggregate impact on that product supply chain to ensure the effect was not over-estimated.
Study results
The study proved extremely valuable. The initial finding was that the previous worst-case loss estimates represented very unlikely events of the order of 5,000 to 10,000 year return periods. More realistic exposure estimates were generated, and more importantly, we were able to show senior management a range of outcomes across a range of return periods in a highly visual manner.
Management was, thus, able to identify five or six key sites from the portfolio of suppliers that represented an unacceptable business exposure.
Highly targeted risk mitigation plans were developed focusing on these specific peak exposures. This involved a range of options including:
- Identifying and registering secondary suppliers, in some cases in Europe and the United States
- Agreeing business continuity plans with existing Japanese suppliers
- Creating strategic safety stocks of certain key intermediates outside Japan
In other words, the magnitude and likelihood of earthquake events described in the risk management process was credible, and the management was able to commit to significant risk mitigation plans on this basis.
Insurance implications
At a time when the insurance market was hardening and capacity for Japanese earthquake cover was shrinking, it proved very valuable to able to differentiate the company's risk management approach. We could describe our exposure to Japanese earthquake and present the risks to underwriters in a sophisticated manner which allowed us to secure significant insurance capacity.
We subsequently developed the original business interruption risk management model into a simulation model using Monte Carlo techniques. This allowed the scenarios generated for each fault line to be sampled to provide probability distributions. These distributions were interpreted to determine the cost-of-risk exposure to each layer of the company's property damage/business interruption programme to natural catastrophe risk. This exposure enabled us to generate an internal pricing model for earthquake insurance, which in turn informed the intelligent purchasing of reinsurance and also allowed effective reserving within the group's insurance subsidiaries.
References
1. CRESTA: Catastrophe Risk Evaluation and Standardising Target Accumulations - Swiss RE methodology.
2. ATC-13 document - Earthquake Damage Evaluation Data for California.
- Alex Hindson is associate director, enterprise risk management, IRMG, part of AON. He was previously senior risk partner, AstraZeneca.
Email: alex.hindson@irmg.aon.co.uk, Web sites: +
