How to Structure a Recovery Plan Playbook That Passes PRA Credibility Tests

The PRA does not want a Recovery Plan that reads like a marketing brochure, and your counterparties do not want to see one that reads like a distressed borrower's confession. Getting the balance right is a drafting discipline, not a strategic one. The plan has to be credible to supervisors reviewing it under SS9/17 and the Recovery Planning Statement of Policy, usable by executives at 2am on a Sunday, and defensible if extracts ever reach a CFO on the other side of a repo trade.

Key Executive Takeaways

• The credibility of a Recovery Plan turns on the realism of its indicators and the executability of its options, not on the volume of narrative around them.
• Language discipline matters: describe options in operational terms tied to management actions, not in terms that imply the firm expects to use them.
• Governance triggers, not thresholds alone, are what the PRA tests hardest; the escalation path from amber breach to board convening must be timed, named and rehearsed.

Start with the reverse stress test, not the menu of options

Most firms build the plan forward: pick indicators, list options, calculate capacity. The PRA reads it backwards. Supervisors want to see that you have identified the scenarios that would actually threaten viability, and that your options bite in those specific scenarios. Begin with two or three severe but plausible paths to failure (idiosyncratic, market-wide, combined) and design the option set to answer each. If an option only works in a benign scenario, it is not a recovery option.

Calibrate indicators to give you time, not comfort

The common failure is setting amber triggers so close to red that by the time the board convenes, the window for private-market options has closed. Capital and liquidity indicators should breach amber when you still have optionality: enough runway to execute a rights issue, sell a portfolio, or draw committed facilities without the market noticing. Test each indicator against your last three stress episodes and ask whether it would have fired early enough. If not, move it.

Avoid the trap of over-engineering with fifteen indicators. Five to seven, each with a clear owner, a defined data source, and a stated frequency, is more defensible than a dashboard nobody can maintain.

Write options so they read as management tools, not distress signals

This is where language discipline earns its keep. Every recovery option should be described in three registers:

• The operational mechanics (who does what, in what sequence, with what approvals)
• The financial impact (capital, liquidity, P&L, with sensitivities)
• The execution risk (counterparty reaction, market conditions, legal constraints, time to cash)

What you do not do is characterise options as things the firm will do in a crisis. Describe them as available management actions. A disposal is a portfolio optimisation option. A funding drawdown is a contingent liquidity action. This is not spin, it is precision: the same option might be executed for balance sheet management in normal times. If the language works in both contexts, it will not read as a distress signal if it leaks.

Make the governance section the sharpest part of the document

Supervisors probe governance harder than financials. Specify:

• Who owns each indicator and how breaches are reported
• The exact escalation path from amber to red, with named committees and time limits (e.g., ExCo within four hours, board within twenty-four)
• The decision rights for invoking each option, including who can act before formal board approval
• The communications protocol for regulators, counterparties and staff

Run a live simulation annually and document the lessons. A plan that has been tested against a scripted scenario, with observed timing failures written up, is far more credible than one that has not.

What good looks like

A credible Recovery Plan is short, specific and rehearsed. The option set is diverse across capital, liquidity, cost and structural actions. Recovery capacity is calculated conservatively, with overlaps and dependencies netted off. The playbook annex, the operational document management actually uses, is separable from the supervisory submission and can be updated more frequently.

The next decision

Before your next annual refresh, ask whether the plan you would hand to the PRA is the same document you would hand to your Treasurer at 6am on a Monday. If it is not, you have two plans to write, and only one of them is the one supervisors are testing.

Frequently Asked Questions

How detailed should the scenarios be?

Detailed enough that each option can be tested against them with numbers, not adjectives. Two idiosyncratic, one market-wide, and one combined scenario is a defensible minimum. Each should specify the trigger event, the transmission mechanism to capital and liquidity, and the assumed duration.

Should we include options we would not realistically use?

No. Padding the option list with actions the board would never sanction weakens credibility. The PRA reads through this quickly. Fewer options with genuine executability score better than a long menu of theoretical actions.

How do we handle the confidentiality risk of the plan itself?

Treat the plan as market-sensitive. Restrict distribution, use clear version control, and ensure the language would not read as an admission of weakness if extracts appeared in litigation or press. The operational playbook can be more explicit because it stays inside the firm.

How often should the plan be tested?

A full simulation annually, with a lighter tabletop exercise on a specific option (for example, a rapid disposal or facility drawdown) at least once more per year. Document the timing observed against the timing assumed. Gaps are the most useful output.

What is the most common PRA challenge?

Recovery capacity that has not been stress-tested for overlapping dependencies. If two options rely on the same counterparty, the same legal entity or the same market being open, the combined capacity is less than the sum. Show the netting.