What should stop needing your approval?

Run the exception audit

Before changing anything, get the facts. For one week, write down every question that reaches you — every call-me, every “quick question,” every thread you were pulled into for a yes. Capture it verbatim, with who asked and roughly when. Don’t fix anything yet; a fix mid-audit changes the data.

At the end of the week, sort every entry into three piles:

• “Needed me” — judgment only the owner can supply: a precedent-setting price, a new commitment, a one-way door.
• “Needed a rule” — a recurring question with a knowable answer, asked because the rule doesn’t exist in writing, or exists but nobody trusts it enough to act without cover.
• “Needed information someone lacked” — the asker had the authority but not the data: the margin table, the customer’s history, the real cost of the change.

Run it honestly and the first pile is usually the shortest. The second pile is your rulebook backlog, already sorted by frequency. The third pile isn’t a delegation problem at all — it’s an access problem, usually fixed by opening a report to the people who keep asking for its contents. Only the first pile is your actual job.

Write decision rules that scale

“Use your judgment” is not a rule; it’s a liability transfer, and your team knows it. That’s why they keep asking. A decision rule has three parts: who decides, within what limits, and what gets logged — written down, in a place people can cite while a customer is on the phone.

Start with dollar thresholds, because they’re the fastest to write and the fastest to relieve pressure. Refunds under a set amount — the service manager grants them on the spot. Discounts up to a stated percentage — the salesperson applies them without asking, never below the pricing floor. Rush fees waived under a fixed figure — dispatch decides. The numbers are yours to set; put each threshold where a wrong call stings but doesn’t scar, and revisit the levels every 90 days as the company grows.

Then turn your audit’s second pile into playbooks. Every exception that showed up more than once gets one paragraph: the trigger, the allowed responses, the limit, and what to write down. Build these from real cases you’ve already decided — you’re not inventing policy, you’re transcribing judgment you already apply, so the next occurrence doesn’t need you to repeat it.

Sequence matters, and there’s a clean test for it: the two-way door. A reversible decision — a refund, a schedule swap, a goodwill credit — can be unwound if it’s wrong, so it gets a rule and a named decider first. An irreversible one — a signed lease, a fired employee, a public price cut — cannot be unwound. Delegate the two-way doors first, and the one-way doors last or never. Some decisions should keep needing you:

• The pricing floor itself. Anyone can price above it inside their limits; only you move the floor.
• Hiring and firing. Who is on the team is a commitment, not a transaction.
• New standing obligations — leases, credit terms for a new account, multi-year contracts.
• Anything that bets the company: a new market, a new line, a customer so large they change your risk.

These stay on your desk not because your team is weak, but because they’re one-way doors. Keeping them is exactly what makes releasing everything else safe.

Design the escalation path before you need it

A rule without an escalation path dies on its first strange case. Someone hits a situation the playbook doesn’t cover, walks it to your desk, and quietly concludes the rule was decorative — and the old reflex returns for everything. Escalation has to be designed, not improvised.

Three elements do the work. First, a named decider with a clock: for every decision type, one person owns the call within a stated window — who decides, and within how many minutes — because a deadline turns “I’ll ask around” into a commitment the customer can feel. Second, automatic triggers: when a case crosses a dollar threshold, touches one of your largest accounts, or outlasts its clock, it moves up one level on its own. Nobody has to be brave, and nobody has to guess. Third, a logging duty: whoever decides writes down the decision, the reason code, and the dollars, in the same minute they decide it.

Notice what this preserves. The exception still gets decided quickly, close to the customer, by someone with the authority to decide it. What changes is that every decision now leaves a record — and a record is something you can actually supervise.

The control loop that replaces cc

Being cc’d on everything feels like control and delivers none of it. The threads arrive faster than anyone can verify them, so you end up skimming everything and auditing nothing. Replace the cc line with a loop that has three parts.

• Reason codes on every exception. A short, fixed list — customer-save, competitor-match, our-error, goodwill — one code per logged decision. Codes turn a pile of stories into a pattern you can see: which code is growing, which rep leans on which, what each one costs.
• A weekly one-page readout, by name. Every exception granted last week: who decided, which code, what amount, what happened next. One page, real names, no summaries of summaries. It replaces a hundred interruptions.
• Spot-checks. Pick a few lines from the readout, pull the underlying thread, and confirm the story matches the code. Not many — just enough that everyone writing a reason code knows a check can land on it.

Compare the two regimes honestly. Before: you saw requests one at a time, decided under interruption, and kept the record in your head. After: rules you wrote, deciders you named, a code on every exception, a page every week, and an audit trail you can walk. You didn’t give up control — you upgraded its instrument, from your inbox to a system that reports. The triage from here is simple. You can run the exception audit yourself this week: a notebook, one honest week, three piles. Where the pattern is contested — you believe approvals come back fast, your team says quotes sit — the question stops being a matter of opinion, because your systems already hold timestamps between request and approval. The Revenue Review maps that decision latency from your own records — quotes, refunds, change orders, message threads — and shows in writing where work waits and what the waiting touches. Fixed fee $4,500, two weeks. There’s a sample report on the method page, and a free 20-minute call if you’d rather talk it through first.