Skip to content
DevOps AI ToolKit
Newsletter
All prompts
Reduce MTTR with AI Difficulty: Intermediate ClaudeChatGPTCursor

Customer-Impact Quantifier Prompt

Turn raw incident signals into a defensible estimate of how many users, which segments, and how much revenue are affected — so severity, comms, and prioritization are sized on impact instead of gut feel, and the response matches the real blast radius.

Target user
Incident commanders and on-call SREs sizing an active incident
Difficulty
Intermediate
Tools
Claude, ChatGPT, Cursor

The prompt

You are an incident analyst estimating customer impact during a live incident. The team needs a fast, honest sizing of who and how much is affected, so severity and comms are proportional — not over-reactions and not dangerous under-reactions.

Paste what you have:
- The failing signal: [ERROR RATE / LATENCY / AFFECTED ENDPOINTS OR REGIONS]
- Traffic and user context: [REQUESTS/SEC, ACTIVE USERS, TENANT/SEGMENT BREAKDOWN IF KNOWN]
- What the failure blocks: [WHICH USER JOURNEYS OR FEATURES ARE DOWN OR DEGRADED]
- Business context, if available: [REVENUE PER TRANSACTION, KEY CUSTOMERS ON AFFECTED PATH]

Work through this:

1. **Define the affected population.** From the signals, state who is plausibly impacted — all users, one region, one tenant tier, users on a specific journey. Separate "confirmed affected" from "at risk downstream."

2. **Quantify with a stated method.** Estimate the number or percentage of affected users and, if data allows, affected transactions or revenue per unit time. Show the arithmetic and every assumption. Give a range, not a false-precise point value.

3. **Attach a confidence level.** Say how much to trust the estimate and what the biggest source of uncertainty is (e.g. "we don't know if retries mask the true failure count").

4. **Translate impact into severity.** Map the quantified impact to the team's severity rubric (or a standard SEV1-4 scale) and state the severity this impact justifies — flagging if it is higher or lower than the current declared level.

5. **List the cheapest confirming queries.** Two or three read-only queries or dashboard checks that would tighten the estimate or refute it fastest.

Output format: an "IMPACT CARD" with fields AFFECTED POPULATION, ESTIMATE (+range +method), CONFIDENCE, BIGGEST UNKNOWN, SEVERITY THIS JUSTIFIES, CONFIRMING QUERIES. Show your working for the estimate. Present the number as a hypothesis to verify, not a fact.

Run this prompt with AI

Test it, get an AI-improved version, or compare models — live in the Prompt Workspace. No copy-paste.

Why this prompt works

Half of a slow incident response is arguing about how bad it is. Without a shared impact number, teams either panic-escalate a minor blip or shrug off a major outage as “probably just a few users” — both of which cost MTTR, one by burning senior responders on nothing and the other by under-resourcing a real fire. The bottleneck is that turning scattered signals into a defensible impact figure is genuinely hard to do under pressure.

This prompt does that translation with its work shown. It forces a stated method and explicit arithmetic instead of a hand-waved percentage, separates confirmed impact from downstream risk, and gives a range with a confidence level so the number carries its own uncertainty. That honesty is what makes the estimate usable for a severity call: the team can see exactly which assumption to attack if the number feels wrong.

The guardrails guard against the specific danger of a quantified answer — false precision. A crisp “2% affected” is dangerously persuasive even when it is wrong, so the prompt labels every figure a hypothesis, surfaces the biggest unknown, and ends with the cheapest queries to confirm it. Impact sized honestly and verified fast is what lets the response match the blast radius, and a right-sized response is a faster one.

Related prompts

More Reduce MTTR with AI prompts & error guides

Browse every Reduce MTTR with AI prompt and troubleshooting guide in one place.

Free download · 368-page PDF

Reading prompts? Get all 500 in one free PDF

500 battle-tested, copy-paste AI prompts engineered by a senior systems engineer — every one with fill-in placeholders and safety/back-out notes. Drop your email and it's yours.

  • 500 prompts: Linux · Kubernetes · Terraform · OpenStack · GitLab · Docker · Monitoring · Incident Response
  • Instant PDF download — yours free, forever
  • Plus one practical AI-workflow email a week (no spam)

Single opt-in · unsubscribe anytime · no spam.