Control Room · Ops

From Alert to Playbook: What to Do When a Delivery Metric Spikes

When a delivery metric spikes, don't just stare at the dashboard: run an if-then playbook. Confirm the signal is real, look at the one underlying cause, make a single corrective move, assign a clear owner, and set a time box to re-check. A metric without an owner and a deadline is noise, not action.

By Control Room · 2026-10-11 · 5 min read

30%typical commission on the ticket

15%margin when nobody measures contribution

$100sample order, broken down below

Illustrative example

• Food $32
• Commission $28
• Packaging & ads $12
• Promo $10
• Margin $18

Food, commission, packaging, ads, and promo eat the ticket before margin. Every restaurant has its own structure.

Every delivery operation generates alerts: prep times creeping up, cancellations stacking, sales flat in one store, rating sliding. The problem is rarely a shortage of data — it's that a spike sits on a screen without becoming a decision. This piece skips another list of KPIs and gives you conditional playbooks: if X happens → look at Y → move Z → who owns it → by when. The goal is simple: turn an alert into a repeatable move you can run during the shift, not after the damage is done.

Why a metric needs a playbook, not just a dashboard

A playbook is a documented, repeatable response to a specific signal: the trigger, what to inspect, the corrective action, the owner, and the time box to verify. Without it, the same spike gets handled differently depending on who's on shift — or ignored until the close-of-day report.

Food apps are part of this picture as a channel and a signal: they show you when prep is slow or orders are being dropped. But the lever is almost always inside your four walls — the kitchen, the shift, the routing, the staffing. The app surfaces the symptom; your operation fixes the cause.

Before you move: is this alert urgent or can it wait?

Not every spike deserves a war room. Sort each alert into three buckets:

• Act now (mid-shift): the problem is live and compounding — delivery times climbing during peak, a cancellation streak, a store going offline unexpectedly.
• Act at shift close: the metric moved but the shift is stable — a single bad rating, a one-off slow order, a minor dip.
• Act this week: a trend, not an incident — sales drifting down over days, rating sliding half a point across two weeks.

The test: is it compounding right now? If each passing minute makes it worse, it's an act-now. If it's a pattern, it belongs in your weekly review, not a panic.

Four mini-playbooks you can run today

Playbook 1 — Delivery time is climbing

• Trigger: prep or total delivery time rises meaningfully above your normal band for 2+ consecutive time slots.
• Look at: kitchen queue depth, number of active couriers waiting, and whether a single item is the bottleneck.
• Move: if the kitchen is backed up, temporarily pause or hide the slowest-prep items; if couriers are waiting, stagger ticket release; if it's one item, 86 it or pre-batch it.
• Owner: shift lead at the store.
• By when: re-check in 20 minutes; if times haven't dropped, escalate to ops.

Playbook 2 — Sales dropped in one location

• Trigger: one store's orders fall well below its own trailing average while sister stores hold steady.
• Look at: is the store actually online and visible? Check availability status, menu outages, and whether items are marked unavailable.
• Move: reopen the store or item if it was auto-paused, restore stock, confirm hours are set correctly.
• Owner: store manager; ops verifies visibility from the outside.
• By when: within 15 minutes — a store silently offline is pure lost revenue.

Playbook 3 — Cancellation streak

• Trigger: three or more cancellations in a short window, or a cancellation rate above your baseline.
• Look at: the reason codes — stockout, store too busy, courier no-show, or address issues.
• Move: if stockouts, update availability immediately; if "too busy," throttle incoming orders instead of rejecting them; if courier gaps, check the handoff zone.
• Owner: shift lead, with ops tracking the pattern across stores.
• By when: act within the shift; document the reason mix for the weekly review.

Playbook 4 — Rating is sliding

• Trigger: rating drops, or negative comments cluster around the same theme (cold food, missing items, slow).
• Look at: packaging, prep-to-pickup time, and order accuracy on recent tickets.
• Move: tighten the item-check step before handoff, adjust packaging for heat retention, and run a short recovery message protocol for affected customers.
• Owner: store manager for execution; ops for the recovery cadence.
• By when: address at shift close; review the comment themes weekly.

What to do when two things spike at once

When times and cancellations rise together, treat them as one incident, not two. Rising times usually cause the cancellations downstream, so attack the time first: relieve the kitchen bottleneck or throttle intake. Fixing the root almost always drains the second alert. The priority rule is simple: handle the metric that is causing the others before the ones that are merely reacting.

Common mistakes that keep you firefighting

• No thresholds. "Times are high" means nothing without a number. Define a normal band per store so an alert is objective, not a gut call.
• No owner. An alert everyone sees and no one owns gets handled by no one. Name the person and the backup.
• No time box. Without a re-check deadline, you never learn whether the move worked.
• Always the same store. If one location eats all your attention, the problem is structural — staffing, layout, or routing — not a daily fire. Log repeat incidents and fix the cause once.

From KPI to documented procedure

To make a metric repeatable, write it down in one line everyone can follow: trigger → inspect → action → owner → deadline. Store these playbooks where the shift can reach them, review them weekly, and update the thresholds as your baselines shift. The first time you run a play it feels slow; by the fifth, it's muscle memory — and that's when firefighting turns into operating.

Frequently asked questions