Fragility audit

What it is

Every plan depends on things that must keep working: a supplier, a customer, a platform, a key person, a price. In a fragility audit you list them and ask what the plan loses if each one fails. You look for the one whose loss would hurt far more than its size suggests.

Forecasting is guessing which future will come. Here you look for the one thing that could ruin the plan in many different futures. You predict nothing. You decide what to risk.

Where it comes from

The name has no single author. The method is built from two older ideas.

The idea of fragility comes from Nassim Nicholas Taleb, a writer on risk. He set it out in Antifragile (2012), the fourth book in his series Incerto. For Taleb, fragile means that past a certain point, harm grows faster than the shock that causes it. Without predicting the shock, he says, you can see where a plan would be hurt out of proportion. In a 2012 paper, Taleb and the mathematician Raphael Douady put this into formulas. They offer a "fast-and-frugal", "model-free, probability free" test. You measure how harm grows as the shock grows, without forecasting whether it will come.

The steps are older. Engineers call them Failure Mode and Effects Analysis (FMEA). You list the parts of a machine, describe how each can fail and rate the result. The US military wrote this into procedure MIL-P-1629 on 9 November 1949. Later, NASA contractors used it on Apollo and Voyager. In a fragility audit you do the same to a business plan. Like Taleb, you rate the damage before the probability.

What it corrects

Say one customer brings in 40% of your revenue. A careful analyst forecasts whether the customer will leave. They gather evidence, estimate the odds and get a number. The forecast may be good. But one wrong forecast could still ruin the company. Either way, 40% of revenue stays at risk.

There is a second problem. Some dependencies hide in the stable parts of a plan, where nobody writes them down. Think of the firm that handles your card payments, or the visa rules. Think of the one engineer who understands the billing software, or the shelf space a retailer keeps for you. None of them is in the figures, because none has ever changed.

How it works

1. Write the recommendation as a commitment with a date and a size, so you can see what could break.
2. List what must keep working for it: suppliers, customers, sales channels, platforms, regulators, key people, a price or rate that must hold.
3. For each, say in one sentence how it fails: what stops, and how soon.
4. Rate the damage first, then the likelihood, and sort by damage. A rare but ruinous failure ranks above a likely but cheap one.
5. For the top two, ask: could the plan recover, or would this be fatal? Fatal means the strategy is gone by next year.
6. Act on the fatal ones now: remove the single point of failure, limit your losses, or buy a back-up while nobody else wants it.
7. Name the first visible sign of each failure, so you have something to watch.

Worked example

On 17 March 2000, lightning struck a power line in Albuquerque, New Mexico, and started a fire at a Philips chip factory. The fire burned for ten minutes. The factory made radio frequency chips for Nokia and Ericsson phones. Philips needed weeks to get back to full output.

Nokia bought some of these chips from more than one supplier. Its staff noticed the drop in deliveries before Philips reported a problem. Within two weeks Nokia sent about thirty managers across Europe, Asia and the United States. It redesigned some of the Albuquerque chips so other factories could make them, and pushed other suppliers for more. Nokia's production stayed on target.

Ericsson bought the chips from that one supplier alone. Jan Ahrenbring, its marketing director for consumer goods, later told the Wall Street Journal: "We did not have a Plan B." Ericsson estimated its lost sales at $400 million or more. By January 2001 Nokia held about 30% of the phone market, up from about 27% a year earlier. Ericsson's share had fallen from about 12% to 9%. On 26 January Ericsson announced it would hand all its phone manufacturing to Flextronics.

Neither company predicted the lightning. But Nokia had checked this dependency. It had paid, in design work and ties with suppliers, for the ability to survive the failure.

In a Business Case Weekly case

In the Nike case, the plan is Consumer Direct Acceleration. From mid-2020 Nike cut back on the shops that resold its shoes and put its own stores and apps first. Direct sales did earn more per pair. In a fragility audit you ask instead: what dependency does the plan create? Casual buyers would meet the brand only in Nike's own stores and apps. The shelf space Nike gives up stays with the shops, which can give it to rivals: On, Hoka or New Balance. The deciding question: if someone else fills that space, can Nike buy it back, at what price, and for how long?

In your answer

• "This recommendation depends on X. If X fails, … stops."
• "The damage would be fatal, not just expensive, because …"
• "I am not predicting this failure. I am limiting its cost by …"
• "The first sign would be …, so I would check it every month."

Common misuse

The usual mistake is a risk register: twenty rows, each with a likelihood score, an impact score, a colour and a person in charge. The plan goes ahead unchanged. The table looks like an audit, but giving a risk a score does not make it smaller. If you changed no commitment and bought no back-up, you only described the risks.

The opposite mistake is buying back-ups for everything. Second suppliers, spare capacity and standby contracts cost money and add complexity. A dependency whose failure would only be annoying deserves none of them. Look at the damage first, and pay only where a failure would be fatal.

References

• Nassim Nicholas Taleb, Antifragile: Things That Gain from Disorder, Random House, 2012. The practical method is in the chapters on the barbell and via negativa.
• Taleb & Douady, "Mathematical Definition, Mapping, and Detection of (Anti)Fragility" (2012). The formal version, with the detection test.
• Almar Latour, "Trial by Fire: A Blaze in Albuquerque Sets Off Major Crisis for Cell-Phone Giants", Wall Street Journal, 29 January 2001. The original report behind the worked example; a twenty-minute read.
• Sunil Chopra & ManMohan S. Sodhi, "Managing Risk to Avoid Supply-Chain Breakdown", MIT Sloan Management Review, Fall 2004. On rating dependencies without buying back-ups for everything.