First, read contacts per customer by reason over time and rank the reasons behind the rise
Start with the rate, not the count. The guide to customer service metrics calls contact rate the metric most teams lack and most need, because it connects support to the rest of the business: contacts in the period divided by orders, active customers or bookings in the same period, with the base counted on the same dates as the contacts. A rise in volume during a busy month is not a rise in problems. Plot the rate over enough periods to see the seasonal shape, and read it beside the base, so a queue that grew with a successful quarter is separated from a queue that grew because something broke.
Then split the rate by reason. The tutorial on tagging support contacts by reason calls the reason list the cheapest research a business has: a monthly count of what customers struggle with, in their own words, routed to the teams that can remove the cause. Contacts per customer for each reason, period by period, shows which reasons rose and which stayed flat, and the service and support hub already says where each kind leads: a spike in order-tracking contacts is a delivery or communication problem, a cluster of "I thought it included" contacts is a promise problem marketing created, and a run of "how do I" questions is a product or help centre gap. If contacts have never been tagged, tag a sample now, spread across days, agents and channels, before reading anything else; the first row of the table is that case.
Finally, rank. The Pareto principle is the observation that value is often unevenly spread, so that a few items on a list carry most of it, and the framework's instruction is to rank the list by the right measure and check the concentration rather than assume a proportion. Rank the reasons by their contribution to the rise, contacts per customer now against the earlier period, and the few at the top are the ones to take to their owners this month. The rest are the useful many, read again next month. This page sits beside the tagging tutorial, which builds the reason list, the support channels and self-service guide, which owns the help centre and chat, and the diagnosis for order and account emails not arriving, which handles one common source of tickets on its own; it starts from a rate that has risen and ranks the reasons before deciding which team owns the fix.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Volume is up but nobody can say which questions grew; the largest tag is "general" or "other" | Whether contacts carry one main reason tagged at close, and what share sits in "other" | Contacts counted but never tagged by reason, so the growing cause stays invisible | Draft a short reason list from real contacts, tag at close, report monthly against a base with owners |
| The same few questions are answered by hand all day, and help searches return nothing or lead to a contact | Help searches with no result, articles followed by a contact, articles rated unhelpful, and what the customer tried first | No self-service, or a help centre that is unfindable or out of date, so the same questions are answered by hand | Run the help centre as a product: answer the top questions where they arise, fix findability, give articles owners and review dates |
| A wave of questions about an offer, a code or a claim began the day a campaign launched, and agents learned of it from customers | The campaign calendar against the queue, and whether a support brief existed before launch | A campaign or offer that reached the support queue before the team was briefed | Write the one-page support brief before every campaign, with saved replies, allowed actions and an escalation route |
| Contacts asking "where is" or "any update" cluster around one warehouse, route, service or date | The affected group in the order, booking or account system, and whether anyone told them before they asked | A delay or disruption customers hear about only when they chase | Tell affected customers first, with what changes for them, their choices and the next update, and brief support before sending |
| Customers come back about the same problem within days; fast first replies, rising repeats | Repeat contacts and reopened contacts by reason within the window, read beside first-exchange resolution | Repeat contacts about the same issue because the first answer did not resolve it | Pair every speed metric with repeats, fix the answers and processes behind the reasons that come back, and review conversations for quality |
| Questions from people with no order sit in the general queue behind complaints, answered slowly or by guesswork | Contacts from people with no order history, by channel, product and question, and who answered them | Pre-purchase questions sitting in the general queue with nobody assigned | Tag them, decide who answers each type, give agents product sheets with approved wording, and move recurring answers onto the pages |
Fix, verify and prevent each cause
Contacts counted but never tagged
A queue without reasons is a number without a cause. The tagging tutorial describes how most tagging fails: lists grow by accretion until agents pick the first plausible tag, reasons mix what the customer asked with what the agent suspects, "other" swallows a large share and nobody reads it, and every change to the list breaks the comparison with last month, so the report is neither trusted nor acted on. When volume is rising and the largest tag is "general", the growing cause is somewhere inside it.
Fix: follow the tutorial. Read a sample of recent contacts from every channel with identifying details removed, draft reasons in the customer's own terms, keep the list to two levels, separate the reason the customer raised from the root cause behind it, write a definition and an example for each reason, tag at close with one main reason per contact, and give "other" a rule: a note from the agent, read monthly by the list's owner. Then count the reasons against a base the business already tracks and rank them.
Verify: a second person re-tags a random sample without seeing the original tags and agreement is tracked as a quality measure for the list; the share in "other" falls and the top reasons each name a team.
Prevent: give the list one owner with a change log, change it on a fixed cycle with old reasons mapped to new so the history can be restated, and put the monthly count, with owners and agreed actions, on the calendar of the marketing and product leads as well as the support lead's.
The same questions answered by hand because self-service fails
The support channels guide treats the help centre as the cheapest channel and usually the most neglected, and asks that it be run like a product, with users, a purpose, measures of success and a backlog. Its content comes from support: the questions customers ask most often, in the words they use, each article answering one question and starting with the answer, because, as the guide takes from Nielsen Norman Group's research on how people read online, people looking for help scan for their answer and leave if they cannot see it. Findability is the other half: customers reach articles through the site's own search, through a link at the point of need such as an order page linking to the returns article, and through search engines, and an article none of those reach is answered by hand every time. An out-of-date article is worse than none, because it creates contacts and damages trust.
Fix: take the top reasons from the ranking to the help centre's own measures, searches that return nothing, articles followed by a contact and articles rated unhelpful, and read a few contacts for what the customer tried first; then write or rewrite the article in the customer's words, link it from the point where the question arises, and move the answer onto the product page or into the order email where that is where the question is really asked. Where a chat assistant answers first, keep it to what it reliably answers well and hand over to a person with the conversation attached. Every channel must be usable by customers with disabilities, and the guide names the Web Content Accessibility Guidelines as the reference for the digital channels, so a help centre, a form or a chat widget that cannot be used with a keyboard or a screen reader shuts some customers out at the moment they need help.
Verify: contacts per customer for the reason fall over a comparable period while orders hold, and the help centre's no-result searches and article-then-contact counts fall with them.
Prevent: give each article an owner and a review date, tie help-centre maintenance to the release process so a change to the product or a policy changes the help, and keep the help centre's failures as the backlog that the monthly reason review feeds.
A campaign that reached the queue before the brief
The tutorial on briefing customer service before a campaign says why this cause is so common: a campaign brief is written for the people who make the campaign, not for the people who will answer customers about it, yet the first evidence of how a campaign is landing arrives in the support queue, as questions about whether a product is included, codes that did not work and customers who bought the day before asking for the discount. An agent who learns about the offer from the customer answers every question as a guess, and an agent who improvises the terms, or adds a promise the campaign never made, creates the problem the campaign's own checks were meant to avoid.
Fix: for the campaign now running, write the brief today rather than at the review: the exact terms and exclusions, the claims in their approved wording, the questions agents have already heard, saved replies for them from the library of saved replies, what agents may and may not offer, and a named escalation route for a wrong price, a broken code or a misleading claim; where the queue has exposed an error in a live ad or page, the live ad and page error response SOP owns the containment and briefs support first.
Verify: the daily note from support to the campaign owner shows the top questions falling after the brief and the creative fix, and the review after the campaign records which questions the brief missed.
Prevent: make the support brief an item in the campaign process beside the readiness checks, sent early enough for agents to read and question before the first customer does, with the end date and switch-off agreed in advance so codes, banners and campaign replies do not outlive the offer.
A delay customers discover by chasing
The guide to proactive service messages describes the three ways customers usually discover a service problem: they wait for something that does not happen, they try to use something that does not work, or they hear about it from someone else, and each moment produces a contact that starts with a customer who already feels let down and has done the work of finding out. A cluster of "where is" and "any update" contacts around one warehouse, route, service or date is that problem arriving in the queue.
Fix: find exactly who is affected from the system that holds the commitment, the order, booking or account records, checked against a second source so the filter is right; send the message once the business can say in one sentence what has happened, who is affected, what changes for them and when they will hear next, leading with what it means for the reader and their choices, free of offers; put the same facts on a status line, the relevant help article and the phone greeting; and brief the support team with the same facts before the message goes out, not after the first calls arrive. The guide to reducing customer effort names telling customers first as the main fix for chasing.
Verify: count how many of the affected customers still had to get in touch, and compare with an earlier disruption that was not announced, in the business's own records; the guide calls that a more honest measure of proactive messages than any open rate.
Prevent: name an owner and a deputy for disruption messages with a short rule for when to send, promise and keep the next update in every message, close the problem with a final message, and send the cause to the root-cause review.
Repeat contacts because the first answer did not resolve it
The service metrics guide's warning is that a fast reply that does not solve the problem starts the clock again, and that repeat contactsA customer contacting the business again about the same problem, the clearest sign that the first answer did not end the customer's work. are the clearest sign of a channel answering quickly but badly. Every repeat is a second contact in the count, so a queue whose first replies got faster can grow for exactly that reason, and the guide's definition decisions matter here: a contact reopened by the customer is a repeat, not a new contact, and the repeat window is set from how long problems in the business normally take to show they are fixed.
Fix: read repeat contacts and resolution in the first exchange by reason, beside time to first reply, and take the reasons with rising repeats to the people who own the answer or the process, because a reason with high repeats points to an answer or a process that does not work rather than to an agent; keep the original reason on the second contact so repeats can be counted by reason, as the tagging tutorial asks; and sample real conversations for whether the reply was complete, since numbers cannot say whether an answer was right.
Verify: repeats for the affected reasons fall within the window while first reply time barely moves, which is the sign that speed was no longer being bought with incompleteness.
Prevent: pair every speed metric with the repeat contacts that would expose it, keep the repeat window and the clock rules out of the hands of whoever is judged on the number, and end the weekly service report with decisions and owners.
Pre-purchase questions nobody owns
The guide to answering pre-purchase questions points out that not every contact comes from a customer, and that questions from people with no order and no lead record fall between teams: support is organised around orders and accounts, sales around qualified leads, so a buyer's question waits behind service problems or is answered by someone who does not know the product well. The guide's rule is that pre-purchase questions should not sit in the same queue as complaints without a way to tell them apart, because both lose when they compete for the same agents.
Fix: tag contacts from people with no order history by channel, product and question for a few weeks; decide who answers which kind, support agents for factual questions about fit, delivery, returns and how things work, specialists for deeper technical fit, sales for tailored pricing, finance and business buyers' requirements; give agents a short product sheet per product with claims taken from the approved wording and a rule never to add a promise the page does not make; and set a standard for how quickly these questions are answered in each channel from the business's own evidence of what buyers expect.
Verify: record every pre-purchase question with its product, type and outcome, and watch the most-asked questions fall after their answers are moved onto the product, pricing, delivery or checkout page where the question arises, with no fall in sales.
Prevent: take the most-asked questions from the record to the pages each month, keep counting after each move, and hand questions to sales with the question and contact details attached so the buyer is never asked twice.
Prevention: every contact tagged, a support brief before every campaign and a monthly reason review with owners
The queue rises quietly because each cause above is owned by a team that never sees it: the web team does not read help searches with no result, the campaign owner does not read the first morning's chats, operations does not count the customers who chased, and nobody owns the buyer's question. The routine that prevents the next rise makes each of them visible to its owner every month. The customers already waiting in a queue the rise has swollen are a separate job, which the tutorial on clearing a support backlog takes in order, from telling everyone waiting what to expect to closing nothing that has not been answered.
The first habit is that every contact carries a reason. The reason list stays short, in the customer's terms, with definitions and examples beside the tag field, tagged at close with one main reason, audited monthly for agreement and changed only on a fixed cycle with a change log, so the trend survives. Contacts per customer by reason, against orders, active customers or bookings, is the demand section at the top of the weekly service report, and the reasons that moved most are highlighted for the people who can act. The service metrics guide puts contact rate per customer and complaint volume by cause on the retention dashboard, as the churn and retention metrics guide asks, read by customer group so they act as a leading indicatorA measure that moves before the business outcome it is expected to predict, paired with a lagging measure that confirms the outcome later. rather than a support figure; the contact reason entry is the reference for what the tag records.
The second habit is the support brief before every campaign, and the proactive message before every disruption. Both put the same facts in agents' hands before customers have them: the terms, the claims in approved wording, the saved replies, what agents may offer, who to contact when something is wrong and when the next update comes. Agreed in advance and written down, the escalation route works like a service level agreementA published promise between teams, or with a supplier, about what each type of request needs and how quickly it will be handled, agreed rather than imposed. between marketing and support: each type of campaign problem has a named route and an owner on every day the campaign runs. The daily note from support during a campaign is campaign data that often explains a dip before the dashboard does. The review after each campaign records the questions the brief missed, so the next brief starts from them.
The third habit is the monthly reason review with owners. The top reasons, ranked by contacts per customer and by their contribution to any rise, go to the team that can remove the cause, with the root cause worked out for each, a few anonymised quotes in the customer's words and an agreed action with a date: promise problems to marketing, page and help gaps to whoever owns them, operational faults to operations, buyers' questions to the product pages. Treat the share of contacts in the top reasons as a KPIA metric tied directly to an objective. only where an owner and an action stand behind it. Read the Pareto ranking again each month rather than assuming last month's few are still the few, because the concentration moves as fixes land and campaigns change. When complaints rather than questions are what is rising, the diagnosis for customer complaints that are increasing reads the same tagged contacts for failures that were never recovered; when the queue is calm but a relationship score is falling, the diagnosis for a relationship survey score that is falling starts from the survey instead. The customer experience troubleshooting hub holds the other symptom pages of this collection. Once the rise is under control, the customer service audit scorecard rates these habits alongside channels, replies and complaint handling, so the next review starts from a ranked list rather than a symptom.
A school uniform shop finds a size guide behind a third of its tickets (illustrative)
Plonvutra Schoolwear
Plonvutra Schoolwear, an invented online school uniform shop with 5 agents, watches tickets rise through the summer from about 1,900 a month in May to 4,300 in August while orders rise from 6,200 to 9,100. The support lead reads the rate first: 31 contacts per 100 orders in May against 47 in August, so the queue has grown far faster than the customers. Contacts carry no reason at all, so the support lead tags a sample of 400 August contacts against a fresh list of 12 reasons drafted from them.
Three reasons carry most of the rise. Sizing questions are 34% of the sample; the help centre has a size guide, but it sits two levels down under "ordering", returns no result for "what size", and is linked from nothing a parent sees before checkout. Questions about a back-to-school offer are 18%, almost all from the first four days after launch, when agents learned of the offer from customers and could not say whether blazers were included. Repeat contacts about delayed embroidery are 15%: an embroidery machine failure added a week to personalised orders and no customer was told, so each affected parent wrote twice, once to ask and once to chase.
The shop moves the size guide onto every product page beside the size selector and into the order confirmation email, and rewrites it in the words parents use. The campaign owner writes a one-page support brief a week before the next offer, with the exclusions and 6 saved replies. Operations sends a proactive message to the 740 customers with personalised orders in the queue, with the new date and the choice of a plain garment now, and briefs the agents first. By October contacts run at 29 per 100 orders, sizing contacts are 11% of the sample, and the reason review sits on the marketing and operations leads' calendars with an owner beside each of the top five reasons.
Common mistakes
- Counting volume without a base. Contacts rising in a busy month is not a rise in problems; read contacts per order, customer or booking.
- Hiring before ranking. Agents absorb a cause that another team could have removed at its source, and the rate stays high.
- Tags inherited from the help desk. A list nobody drafted from real contacts hides the growing reason inside "general".
- A help centre written once. An article that is out of date or unfindable creates contacts instead of preventing them.
- Briefing support on launch morning. Customers learn about the offer before the people answering them, and every answer is a guess.
- Waiting for customers to ask about a delay. Each one who finds out alone writes once to ask and once to chase.
- Celebrating faster first replies. Speed bought with incomplete answers comes back as repeats and a larger queue.
- Leaving buyers in the general queue. Questions from people who have not bought wait behind complaints and lose the moment.
