Scope: what this audit covers and what it does not
The question this audit answers is whether customer service gives customers correct answers and fair fixes through channels they can use, stays in step with what marketing promises, and sends what it hears to the teams that can remove the causes, and which weakness should be fixed first. It covers the support operation as customers meet it, from the channels on offer to the tagging, metrics and quality review that turn contacts into decisions.
It does not re-teach the work: the methods sit in the guides and tutorials of the service and support hub. Scoring a sample of replies is the tutorial on reviewing support conversations for quality; this scorecard asks whether that review happens. An assistant about to answer its first customer is checked with the AI customer service launch checklist.
Why customers leave is scored by the customer retention audit, which takes service as one part of the system that keeps customers. Public reviews are scored by the customer reviews and reputation audit, and replies on the brand's social profiles by the social media presence audit, with moderation left to the community management guide. Staffing levels, workforce planning and legal compliance are out of scope. Every scorecard sits in the audit library.
Before you score: access and evidence
Ask for read access to the support tool, with its tags, statuses and saved replies, and to the help centre with its search log; without them the measurement rows stay provisional. Fix an evidence windowThe fixed period of data an audit scores, agreed before scoring, long enough to cover a full cycle for the business and kept the same at every rescore. that includes at least one campaign or promotion and the contacts it caused, and keep the same window for every rescore.
Collect:
- the written service standards for each channel, the routing rules and the tone guidelines;
- the complaints procedureThe written process for complaints: what counts as one, how every channel feeds one log, the response promised, who decides and how customers escalate., the complaint log and the remedy table with its approval levels;
- contact reason counts, the weekly service report and the latest quality review scorecards;
- the briefs support received before recent campaigns, and the messages sent about a recent disruption;
- the instructions, source content and a sample of conversations of any assistant that answers customers;
- notes from a walk through each channel as a customer, including reaching a person and making a complaint.
Before looking at any counts, write down what the team believes customers contact it about most, and compare it with the tagged reasons. As the expectation-disconfirmation model suggests, contacts cluster where the experience falls short of what customers were led to expect.
Criteria and scoring
Rate each row missing (absent or broken), partial (present but incomplete, inconsistent or out of date) or solid (matches the "what good looks like" column). The scale is your judgement against that column, not a number. Rows marked gate decide the band on their own when they score missing, because they are the records that show what customers need and where the business fails them.
Channels and self-service
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Channels matched to questions | Each channel offered suits the customers and the questions that arrive in it, has an owner and a standard, and is staffed whenever it is shown as open | Missing: unowned channels, or chat shown open with nobody there / Partial: owned, but questions forced into the wrong channel / Solid: chosen, owned, staffed |
| Help articles found at the point of need | Help articles are titled the way customers phrase the question, open to search engines, linked from the pages and messages where the question arises, and kept current by an owner | Missing: no help content, or hidden behind a login / Partial: stale or unlinked / Solid: owned, dated, linked |
| A person within reach | Customers can reach a person at any point in more than one way, written and spoken, through forms and chat usable by keyboard and screen reader | Missing: the route is hidden behind a bot / Partial: one route only, or not accessible / Solid: several routes, tested |
| Assistant answers and handover | Any bot or AI assistant answers only from approved content, says when it does not know, hands the whole conversation to a person and has a sample of its answers read | Missing: answers from general knowledge or stale content, or traps customers / Partial: handover loses the history, or nobody reads its answers / Solid: grounded, honest, sampled |
The support channels guide owns the first and third rows, with accessibility as it states it from the Web Content Accessibility Guidelines. Findability follows the help centre SEO guide, any assistant the guide to AI in customer service, and wrong answers already given the diagnosis for an assistant giving customers wrong answers.
Standards and replies
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Standards written per channel | Response and resolution standards are set from customers' expectations and the team's capacity, published where customers write in and reported by channel with an owner | Missing: none written / Partial: unpublished, unowned or set from ambition / Solid: published, owned, reported |
| Routing on written rules | Contacts are routed by channel, type of question and the customer's situation on a few written rules, so complaints and urgent cases reach people who can act | Missing: one queue in arrival order / Partial: rules in one person's head / Solid: written, complaints first |
| Tone guidelines in use | Tone decisions are written once on top of the brand voice, by situation and channel, with approved examples from real replies, and updated from what quality reviews find | Missing: each agent decides / Partial: adjectives without examples / Solid: decisions, examples, kept current |
| Saved replies owned and current | Saved replies start from contact reasons, answer one question each, mark the parts an agent must change, link help articles and carry an owner and a review date | Missing: copied from old emails / Partial: no owners, or articles pasted in / Solid: built from reasons, owned, dated |
| Pre-purchase questions owned | Questions from people deciding whether to buy are tagged, answered in approved wording by someone with product knowledge, handed to sales with the question intact and used to fix the pages | Missing: buyers wait in the general queue / Partial: answered but never recorded / Solid: owned, honest, fed back to the pages |
Standards follow the support channels guide, routing the service and support hub, and tone the tutorial on writing customer service tone guidelines, which builds on the Brand VoiceThe consistent personality a brand expresses in its words across channels, with tone adjusting to the situation and the audience. rather than inventing a second one. Each Saved ReplyA pre-written answer a support agent inserts and then edits for the customer in front of them, keeping facts and tone consistent across the team. is judged against the tutorial on building a library of saved replies, and buyers' questions against the guide to answering pre-purchase questions.
Complaints and remedies
| Criterion | What good looks like | Score guidance |
|---|---|---|
| One complaint log for every channel (gate) | What counts as a complaint is defined, every channel feeds one log, and each complaint is closed with the response sent and its cause recorded | Missing: no log / Partial: some channels only, or closed without a cause / Solid: defined, every channel, causes recorded |
| Recovery sequence followed | Complaints are acknowledged quickly, fixed, put right where due and followed up, and public complaints get a calm reply before moving to a private channel | Missing: remedy without acknowledgement, or arguments in public / Partial: some agents only / Solid: the same sequence every time |
| Written remedies and goodwill limits | Each common failure has a standard remedy, customers' rights are kept apart from goodwill, agents use an allowance without asking and every remedy given is recorded | Missing: every gesture needs a manager / Partial: rights and goodwill mixed, or nothing recorded / Solid: a table, an allowance used, a record |
| Customers in vulnerable circumstances | Agents notice signs without labelling people, adapt the process within set limits, record only the support agreed with the customer's agreement and can stop automated messages | Missing: one process for everyone / Partial: sympathy without authority, or too much recorded / Solid: limits set, minimum recorded, automation stopped |
The log follows the tutorial on writing a complaints procedure, the sequence the service recovery and complaints guide and remedies the tutorial on setting remedies and goodwill limits, with a written escalation routeA short written route naming who front-line staff contact, how and how fast, when a problem is beyond what they can decide alone. beyond the allowance. Those guides describe consumer rights from the Competition and Markets Authority's pages as general direction, never legal advice. The last row follows the guide to supporting customers in vulnerable circumstances, which keeps what is recorded to the support agreed, in line with the GDPR.
In step with marketing
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Support briefed before campaigns | Every campaign and promotion reaches support as a brief with the exact terms, the claims, likely questions, saved replies, what agents may offer and the end date, before customers see it | Missing: agents learn the offer from customers / Partial: briefed on launch day, or on the creative only / Solid: briefed before launch and for the end |
| Customers told first about disruption | Customers affected by a delay, fault or cancellation hear it from the business first, with what changes for them, a promised update and a closing message, and support has the same facts | Missing: customers find out by chasing / Partial: no update or close, or support not told / Solid: told first, updated, closed |
| Hours and languages stated plainly | Each channel states its hours in the customer's own time and the languages it answers in, and those promises are kept and reported by market | Missing: unstated / Partial: stated in the business's own time, or not kept / Solid: stated, kept, reported |
Briefings follow the tutorial on briefing customer service before a campaign, and disruption messages the guide to proactive service messages, which keeps a notice a service message rather than a promotion, in line with the ICO's direct marketing guidance. Hours and languages follow the guide to supporting customers in several languages and time zones, and a promise the experience cannot keep goes to the tutorial on checking marketing promises against the experience.
Measurement and insight
| Criterion | What good looks like | Score guidance |
|---|---|---|
| Contacts tagged by reason (gate) | Every contact is tagged on closing with a single main reason drawn from a short list worded as customers put it, agreement is checked on a sample and changes to the list are logged | Missing: no tags, or free text / Partial: used inconsistently, or changed without a record / Solid: one list, tagged at close, checked |
| Service metrics read in pairs | First reply, resolution, repeat contacts and contact rate are defined once, each read beside a guardrail and by reason and channel, in a weekly report that ends in decisions | Missing: one headline average / Partial: no definitions, guardrails or base / Solid: paired, by reason, ending in actions |
| Conversation quality reviewed | A fair sample of whole conversations, with every complaint and escalation added, is read against a short scorecard drawn from the written standards by calibrated reviewers | Missing: nobody reads replies / Partial: failures only, or no calibration / Solid: fair, traced to standards, calibrated |
| Causes sent to their owners | The largest contact reasons and complaint causes are reviewed monthly against a base, each goes to the team that can remove it, marketing included, and the change is recorded | Missing: counted and forgotten / Partial: no owners, or marketing absent / Solid: owners named, changes recorded |
Each Contact ReasonThe main reason a customer contacted support, tagged by the agent from a fixed, defined list so volume and other service metrics can be read by cause. is judged against the tutorial on tagging support contacts by reason. Metrics follow the customer service metrics guide, which reads Contact RateSupport contacts in a period divided by orders, active customers or bookings in the same period, showing how often customers need help. against orders or customers and treats each Repeat ContactA customer contacting the business again about the same problem, the clearest sign that the first answer did not end the customer's work. as an answer that was fast but incomplete. The last row is scored against the prevention routine in the diagnosis for rising customer complaints, and the five whys takes a recurring reason from symptom to a cause someone owns.
How to score and prioritise
Score the two gates first. If contacts carry no reason, the channel, help article and saved reply rows describe a guess: mark them provisional. If complaints have no single log, the recovery, remedy and cause rows rest on whichever complaints were remembered.
Then rank the remaining weaknesses in this order:
- Whatever leaves customers without an answer or a fair fix: no route to a person, an assistant that guesses, no recovery sequence and agents who cannot act.
- Whatever makes answers wrong or inconsistent: stale articles and saved replies, unwritten tone, unowned pre-purchase questions and campaigns that reach support before the brief.
- The loop that removes causes, with marketing in it, and disruption told to customers before they chase it.
- Measurement polish last, unless a missing quality review or a single headline metric is hiding a problem above.
Within each group, rank by the contacts the weakness touches: a fix behind the largest reason outranks the same fix behind a rare one.
Reading the result: what to do with each band
Red means a gate scored missing. Hold changes to channels or staffing, start tagging every contact and logging every complaint from every channel, and read a full cycle before deciding anything else. If contacts are also rising faster than customers, the diagnosis for rising support ticket volume ranks the reasons once tags exist. Rescore on the same evidence window.
Amber means the gates hold at least partially but several other rows are partial or missing, the usual outcome. Work a short ordered list, one area at a time, and read the reasons after each change.
Green means the gates are solid and most rows are too. Move effort into the causes the reasons point at, often on product pages, delivery promises and campaign terms, and rescore each quarter to catch drift.
From findings to a prioritised fix list
Write each partial or missing row as a finding in plain words: what you saw, the evidence and why it matters to customers. Then build one fix list with a line per fix: the finding, the fix, the owner, the guide that explains it and the rescore date, ordered by the priorities above rather than by ease.
Group the lines by who does the work: the reason list and the log with the support lead, articles and saved replies with the content owner, remedies with whoever holds their budget, campaign briefs with marketing and an assistant's sources with its named owner. Show the band, the gate findings and the first fixes to the owner of the customer experience, and keep the full scorecard for the rescore.
Illustrative example
An online wallpaper shop scores its support after a sale
Fenquibrath Wallpapers, an invented online wallpaper shop with a support team of three, runs the audit over the eight weeks around its spring sale, when contacts rose from about 600 to 1,450 a month. Contacts carry no reason, so the first gate scores missing; a hand-tagged sample of 200 shows 38% asking whether rolls from different orders will match and 21% chasing deliveries. The complaint log scores partial: email complaints are logged, but 40 that arrived by phone or in reviews are not. The sale reached support from customers rather than a brief, so that row scores missing, and the chat assistant scores missing because it quotes a returns window the shop changed in the spring; 9 of 26 saved replies repeat the same old window. The band is red. The fix list puts reason tagging and one complaint log first, the assistant's source articles and the stale replies second, a batch-matching article linked from every product page third and a support brief for the summer sale fourth. A month later both gates score partial or better, batch questions fall to 14% of contacts and the band moves to amber.
Common mistakes
- Scoring speed before reasons. A fast reply to a question nobody should have needed to ask is not good service. Score the gates first and read every metric by reason.
- Logging only the complaints that arrive by email. Calls, reviews and social comments are complaints too.
- Treating the assistant as a channel nobody reviews. Its answers are the business speaking; read a sample.
- Counting saved replies instead of owners. A big library with stale facts is worse than a short current one.
- Treating repeated failures as agent faults. Failures repeated across agents point at a reply, an article, a policy or a promise.
- Leaving marketing out of the loop. Many contacts start with a campaign's terms or a page's claim; score whether marketing hears the reasons, not only support.
