A founder of a 35-person engineering consultancy on a Friday afternoon. He has just rejected the operations director’s recommendation to extend payment terms with a key supplier. The director had spent two days analysing the cash impact and brought a clear recommendation. The founder asked three follow-up questions, asked for a revised analysis, and ended the call by saying “let me think about this over the weekend.“
The decision will not get made. He is doing this thirty times a week. He created a delegation matrix nine months ago. Nobody references it. He thinks the matrix is the problem. The matrix is not the problem.
Why do most authority matrices fail in practice?
The founder commissions a RACI chart, populates it with decision categories and approval thresholds, posts it on the intranet, and leaves it unchanged for two years. The matrix is rarely read, even less frequently consulted before decisions are made, and frequently contradicted by founder behaviour. The team learns that the matrix is theatre, not law. They go on asking for permission anyway because that is what gets decisions actually made.
The reason is structural. The matrix sits in isolation. There is no governance around when decisions are made, how they are escalated, or what happens when the matrix is breached. Without consistent founder behaviour and explicit consequences, the matrix is aspirational documentation. The framework is fine; the surrounding behaviour is the gap.
The four-tier authority pattern
Tier one: decisions only the founder makes. Existential strategic calls, major capital commitments, key person hiring or termination. Few in number. Tier two: decisions the founder approves after the leadership team recommends. Critical operational changes, contract variations beyond standard, significant departures from process. Tier three: decisions the leadership team makes collectively, with the founder informed after the fact. Tier four: decisions individual role holders make within their domain, no founder involvement except boundary breach.
This four-tier structure mirrors the Bain RAPID framework but with explicit founder behaviour attached. The founder consciously acts as a single-point decider for tier one, actively reviews and approves for tier two, receives summary reports for tier three, and does not intervene in tier four except through the boundary conditions. The tiers are operational, not honorary.
Monetary thresholds as the starting point
Tier one decisions might include purchases over £100,000 or contracts with annual value over £250,000. Tier two might be £25,000 to £100,000. Tier three might be £5,000 to £25,000. Tier four everything below £5,000. The numbers are illustrative; calibrate to the firm’s revenue and risk tolerance. The point is that the thresholds are explicit and written down, not implied or remembered.
A founder who relies on implicit thresholds finds them drift. Two years in, the same purchase that originally needed approval at £15,000 is now being approved by a manager at £40,000 and the founder no longer sees it. Or the reverse: thresholds harden through founder anxiety and the team is escalating purchases at £3,000 because nobody is sure where the line sits. Explicit numbers prevent both drifts.
Reversibility as the second axis
Jeff Bezos’s framework distinguishes one-way door decisions (irreversible, consequential, meriting careful deliberation) from two-way door decisions (reversible, allowing for rapid experimentation). A one-way door belongs in tier one or two even if the financial value is not large. A two-way door can sit in tier three or four even if the initial commitment looks substantial. The cost of backing out is the test, not the cost of the decision.
Examples: exiting a service line is a one-way door at any value. A trial vendor swap on a small project is a two-way door even at a six-figure cost. Changing compensation structure is a one-way door even when the immediate budget impact is modest, because reversing it is expensive culturally and operationally. The reversibility test sits underneath the monetary one and often overrides it.
Customer-impact thresholds in services firms
A decision that affects contract terms with a top-twenty client is tier two regardless of financial value. A decision that affects a brand-new or marginal customer can be tier three. The threshold is defined explicitly before publishing the matrix. In professional services, customer concentration risk drives a layer of authority that is separate from the cash impact.
A £8,000 service-line change for a £1.2 million client is a tier two decision; the same change for a £40,000 client can sit in tier three. The founder defines the customer tier explicitly. Top twenty by revenue is the simple version. More careful firms add strategic-importance markers: long-tenure clients, referenceable case studies, sector-anchor relationships. The matrix does not need to capture the nuance perfectly. It needs to capture the boundary clearly enough that the leadership team knows which side of it any given decision sits on.
Role-based, not person-based
The matrix says “the Operations Manager can approve contracts up to £50,000.“ Not “Stephen can approve contracts up to £50,000.“ Person-based authority creates bottlenecks when Stephen leaves or moves. Role-based authority transfers with the role and lets you create the role on paper before you have hired into it.
This matters most in growing firms. The role of Operations Manager might be informal today but the matrix can define what authority sits inside that role. When the firm hires properly into the role, the authority is already specified. When the existing manager leaves, their replacement inherits the same authority tier without renegotiation. Person-based matrices look reasonable in firms of fifteen but break the moment the team turns over.
Build it from an audit, not from theory
The error founders most frequently make is publishing an authority matrix based on what decisions the team should make, in theory. Reality bears no relation. The correct approach is to audit actual founder decisions for one or two typical weeks. Note every decision that was made (not merely every meeting attended), the category, the time taken, and whether someone else could have made the decision.
Once the audit is complete, the founder categorises each decision: irreplaceable, replaceable with a recommendation process, replaceable with a team decision, or replaceable with a role decision. The count typically reveals that 60 to 70 percent of founder time is spent on decisions that could be made elsewhere, usually because no one else knows they are allowed to decide. The matrix is then published as a one-page card: for each decision category, the tier, the threshold, the reversibility test, and the approval path.
What to do this week
Track every decision you make for the next five working days. A simple spreadsheet works: time, category, what was decided, who else could have made the call. By Friday you will have a sample of forty to sixty decisions. Read down the column “who else could have made the call.“ That column is the spec for the matrix.
If 60 to 70 percent of your decisions show up as replaceable, the matrix will free meaningful time within two months of being lived. If under 30 percent show up as replaceable, the matrix is not your bottleneck and the gap is somewhere else (probably in the role design or the documented procedures the team needs to act inside their tier).
If you would like a second pair of eyes on what your audit is telling you, book a conversation.
