Macro / saved-reply audit checklist

A macro library grows the way a junk drawer grows: every busy week someone adds one, nobody ever removes one. Two years in you have 300 saved replies, agents can't find the right one, a third point to a feature you renamed, and the canned tone is quietly training customers to feel processed. An audit is how you get back to a library that's small, accurate, and worth reaching for.

The tone problem isn't cosmetic. In Zendesk's 2026 research, 74% of customers say they're frustrated when they have to repeat information, and 67% expect support tailored to their prior interactions (Zendesk CX Trends 2026). A macro that ignores the customer's actual situation trips both at once.

Run a full audit [quarterly] (twice a year if you're a small team), plus same-day fixes whenever a product, price, or policy changes.

Before you start — pull the data

You cannot audit from memory. Get the library and its usage numbers side by side first.

• Export the full macro / saved-reply list from [Zendesk / Intercom / Gorgias / Front / Freshdesk / Help Scout]
• Pull usage per macro for the last [90 days]: times applied, and by whom
• Join the outcome signals you can: reopen rate, CSAT, and escalation rate on tickets where each macro was used
• Separate shared macros from personal/private ones — personal macros are where drift and duplication hide
• List folders/categories and who owns each
• Decide scope: whole library, or one team / channel / language at a time
• Name one owner for the audit and set a deadline

The disposition system

Every macro gets exactly one verdict. No "we'll circle back."

Pass 1 — Usage: the dead-macro sweep

• Sort by times-applied. Anything used 0 times in [90 days] → Retire candidate
• Used fewer than [5] times but still accurate and covering a rare-but-real case → Keep, tag "low use, intentional"
• Identify the top [20] by volume — these carry most of your traffic and get the hardest scrutiny in every later pass. (Rule of thumb: a small handful of macros covers the bulk of volume; find your set instead of guessing at it.)
• Personal macros one agent leans on but nobody shares: promote the good ones to shared, delete the rest

Pass 2 — Accuracy and freshness

• Every factual claim still true: pricing, plan names, limits, timelines, policy
• Every link resolves and points to the current article — no 404s, no redirect chains
• No reference to a renamed or removed feature, a sunset integration, or an old UI ("click the gear icon")
• Screenshots and step-by-steps match the product as it ships today
• Legal / compliance wording current: [refund policy], [privacy], [regulatory disclosures]
• Kill stale time-stamps: "as of [2024]" lines that are no longer true

Anything failing → Fix, or Retire if the thing it describes is gone.

Pass 3 — Tone and the canned test

• Read it aloud. Person, or form letter? Rewrite the robotic ones
• Does it open by acknowledging the customer's actual situation, or launch straight into boilerplate?
• Personalization tokens present and correct ([first name], [product], [order #]) — and do they degrade gracefully when empty rather than printing "Hi ,"
• Every placeholder an agent must edit is obvious ([INSERT DATE]), not a silent blank that ships as-is
• No internal jargon, ticket codes, or dead metaphors leaking to the customer
• Plain and short. Cut the throat-clearing ("We truly value your business and...")

Pass 4 — Coverage gaps

• Pull your top [20] ticket drivers / tags. Does each high-volume topic have a current macro?
• Where agents keep hand-writing the same reply (scan for near-identical free-text answers) → Create a macro
• New features or policies from the last [quarter] — is there a saved reply yet?
• Recurring incidents / outages — a holding or status reply ready to go?

Gaps become Create tasks, tracked apart from the audit dispositions.

Pass 5 — Duplication and findability

• Group near-identical macros. Keep the best, Merge the rest, redirect agents to the survivor
• Naming: can an agent find it by typing the obvious word? Rename the cryptic ones ([Billing - refund - EU], not [macro_v3_final])
• One consistent convention: [Topic - Situation - Channel/Lang]
• Folder structure matches how agents think, not how the org chart looks
• Keyboard shortcuts / aliases on the top macros
• Sanity-check the total count. Fewer, well-maintained macros beat a giant unsearchable pile every time

Pass 6 — Actions and side effects

The part everyone forgets: many macros do more than paste text. They set status, tags, priority, assignee, or fire automations.

• For each macro that changes ticket fields, confirm every action is still valid — a tag that no longer exists, a queue that was merged, a status flow that changed
• Auto-solve / auto-close macros: check reopen rate. Are they closing tickets that actually needed a reply?
• Assignment actions pointing at a person who left or a team that was renamed
• Triggers / automations keyed off a macro's tag — still wanted?

Pass 7 — Localization and channel variants

• Every language version says the same thing as the source — did a fix to the English macro get mirrored to the others?
• Channel-appropriate: chat macros short, email macros complete, no email signature bleeding into a chat reply
• Retiring a macro? Retire the whole translated set, not just one orphaned language

Pass 8 — Governance and ownership

• Every surviving macro has a named owner and a last-reviewed date
• A written "how to propose or change a macro" process exists, so the library doesn't rot between audits
• A deprecation path: retired macros are archived and recoverable, not silently deleted mid-conversation
• If macros feed an AI / agent-assist or draft-reply system, they're the reviewed source of truth — not drifting behind whatever the model actually sends

The working sheet

One row per macro. This is the audit's output.

Disposition roll-up

Kill on sight

Don't debate these. If a macro does any of the following, retire or fix it today:

• Links to a 404 or names a feature that no longer exists
• Zero uses in [12 months] and not a rare compliance / edge-case reply
• Ships a visible [INSERT X] placeholder to customers
• Auto-closes tickets that come back (high reopen rate)
• Is a worse duplicate of a macro you already have
• Reads as a form letter on an angry or high-stakes ticket

Keeping it healthy between audits

Metrics to watch over time

• Macro count — flat or falling, not creeping upward
• % of the library used in the last 90 days
• Reopen rate / CSAT on macro-touched tickets vs. the overall baseline
• Broken-link count — trending to zero
• Coverage — share of your top [20] drivers with a current macro

The goal isn't a big library. It's a small one every agent trusts enough to reach for — and lightly edit — instead of writing from scratch or ignoring entirely.