The 7pm problem
A customer wants a refund on an order that falls outside your returns window. It is seven in the evening, the owner is not answering, and the person on shift has two options: say no and risk a complaint, or say yes and risk being told off tomorrow.
Most of the time they do neither. They say "let me check with someone and come back to you", and the customer waits until morning for an answer that takes thirty seconds to give.
That delay is not a communication failure. Everyone knew who to ask. The failure is that nobody had decided in advance who is allowed to say yes.
This post is about that: decision rights and routing. If your problem is that escalations arrive as unhelpful forwarded emails, what a good escalation actually contains covers the content side.
Escalation is two separate things
Teams conflate them and then design a process that solves neither.
Authority escalation is "I am not allowed to decide this." The person knows what should happen. They lack permission. This is the common case and it is almost entirely fixable in advance.
Knowledge escalation is "I do not know what to do here." The person would decide if they knew how. This is fixed with documentation and training, not with a routing chart.
Sort last month's escalations into the two piles. If most were authority escalations, you do not have an escalation problem, you have an undelegated-authority problem, and no amount of process will fix it because the process is working exactly as designed: it is routing decisions to the only person allowed to make them.
Writing the thresholds down
The fix for authority escalation is unglamorous. Decide the limits once, in advance, and write them where the person on shift can see them.
The format that works is a value or a condition, not a judgement. "Use your discretion" is not a threshold, because the whole problem is that people are unsure how much discretion they have.
| Decision | On shift can | Escalate above |
|---|---|---|
| Refund a damaged item | Any value | Never |
| Refund outside the returns window | Up to a set value | Above it |
| Replace and reship | Yes, one per order | Second replacement |
| Cancel and refund a flagged order | Always allowed | Never |
| Fulfil a high-risk order | No | Always |
| Apply a goodwill discount | Up to a set value | Above it |
| Promise a delivery date | Standard lead times | Anything faster |
Fill in your own values. The specific numbers matter far less than the fact that a number exists, because a person with a number can act and a person without one has to ask.
The asymmetry worth exploiting
When you set those limits, set them higher than feels comfortable.
The cost of a staff member approving a refund you would have declined is the value of the refund. The cost of a customer waiting overnight for a decision is a worse experience, a follow-up message, staff time on both sides, and some chance of a public review. The second is usually larger and it is much easier to overlook because it never appears as a line item.
Most small stores set thresholds based on what they can afford to lose, which is the wrong frame. Set them based on what a delay costs, and they end up meaningfully higher.
A threshold that is too low does not prevent bad decisions. It converts them into delays, and delays are harder to see on a P&L than refunds are.
Routing when it genuinely has to go up
Some things really do need the owner. For those, three rules keep it working.
- Name one person, not a group. "Escalate to management" means escalate to nobody. One name, with a named backup for when they are unavailable.
- State the response time you actually offer. If the honest answer is "by 10am tomorrow", say that rather than implying immediacy. Staff can then set customer expectations correctly instead of guessing.
- Say what to do while waiting. The most commonly missing piece. Does the customer get told "we are looking into it"? Is the order held? Does anything get cancelled? Without this the item stalls entirely while everyone waits.
Out of hours, honestly
Most small stores have no genuine out-of-hours cover, and pretending otherwise produces a policy that quietly fails at 9pm.
The honest version is a short list of what actually warrants contacting someone outside working hours. For most stores it is a very short list: the site is down, payments are failing, or there is a safety issue. Everything else waits, and staff should be told explicitly that it waits, because otherwise conscientious people agonise over whether to call.
Give the out-of-hours person a holding action for everything else: acknowledge the customer, hold the order, log it for the morning. That is usually enough to preserve the outcome without waking anyone.
Making it findable at the moment of use
An escalation policy in a document nobody opens is decoration. The test is whether someone new, alone on a Sunday, can find the answer in under a minute.
Practically: keep it short enough to fit on one screen, put it where the team already works rather than in a folder, and pin it. If your team coordinates in a chat tool, the thresholds table belongs pinned in the room where customer issues get raised, not in a drive nobody has bookmarked.
Review it when someone new joins and after any escalation that went badly. Those two moments will surface more gaps than a scheduled annual review.
Test it before you need it
A policy nobody has rehearsed is a document, not a process. Twenty minutes, once, will tell you more than any amount of drafting.
Take three situations from the past few months and hand them to whoever is most likely to be alone on a Sunday. Not as a quiz, framed openly: "If this landed while I was unreachable, what would you do?"
You are listening for one of three responses. They answer confidently and correctly, which means that path works. They answer correctly but check with you first, which means the authority exists on paper and they do not believe it, and that gap is worth a direct conversation. They do not know, which is a genuine finding and the cheapest one you will ever get.
The middle case is the most common and the most overlooked. People frequently have permission they do not feel they have, usually because it was granted once verbally and never repeated. Written thresholds fix this only if someone points at them and says, out loud, "this is yours, you do not need to ask me."
The three that should never wait
Whatever else your policy says, these three should be actionable by anyone on shift, at any hour, without permission:
- Holding an order. Reversible, costs nothing, and the window is short. Requiring approval to hold something guarantees things ship that should not have.
- Telling a customer you are looking into it. An acknowledgement is not a commitment. Staff who feel they cannot even acknowledge a problem will go silent instead, which is worse.
- Escalating something that feels wrong. Explicitly including cases where they turn out to be mistaken. If being wrong carries a cost, people stop raising things, and you lose the early warnings.
What to do about the answer you regret
Delegate real authority and someone will eventually make a call you would not have made. How you respond to the first one determines whether the policy survives.
If the decision was inside their limits, back it publicly, even when it cost you. Reversing a decision that was theirs to make teaches everyone the authority is nominal, and afterwards they will check with you regardless of what is written down. You will have paid for the delegation and kept none of the benefit.
If the decision was genuinely outside the limits, that is a different conversation and worth having plainly. Usually it reveals the threshold was unclear rather than ignored.
Either way, treat it as information about the policy rather than about the person. A limit that produces a regrettable outcome is a limit set at the wrong level, and adjusting it is a five-minute job. Quietly withdrawing trust instead is how stores end up back at every decision routing through one unreachable person.
Common questions
What should a small store escalate to the owner?
Genuinely very little. Decisions above a set financial threshold, anything with legal or safety implications, and anything that sets a precedent you would not want repeated. Most day-to-day judgement calls should be delegated with a written limit.
How high should refund thresholds be?
Higher than most owners set them. Weigh the cost of an occasional wrong approval against the cost of every customer waiting overnight for an answer, including the follow-up messages and staff time. The second usually wins.
What if we have no out-of-hours cover?
Say so explicitly and define the very short list of things that justify contacting someone. Then give staff a holding action for everything else, so nothing stalls while they agonise over whether to call.
Should escalation go to a group or a person?
A person, with a named backup. A request addressed to a group is addressed to nobody, which is the same diffusion problem that breaks group chats as teams grow.
How often should the policy be reviewed?
When someone new joins, and after any escalation that went badly. Those two triggers surface more real gaps than a calendar reminder ever will.