Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Customer Experience, Retention & Loyalty › Service and Support

Clearing a Customer Support Backlog

How to clear a support backlog after a spike: slow what feeds it, tell everyone waiting, merge duplicates, take urgent cases first and never close early.
Tutorialintermediate15 min readUpdated

What a backlog costs the customers inside it

The guide to customer service metrics defines the backlog as the contacts still open together with the age of the oldest among them. It warns that a small backlog can still hold customers who have waited a long time, that contacts parked in a pending status hide work, and that the weekly report should list the oldest open contacts. The metrics tell a team that a backlog exists. They do not say what to do about the customers already inside it, and that is the job here.

Every contact beyond the standard is a person who has not heard back within the time the business set. They do not know whether their message arrived or when anyone will read it, so some write again, some try another channel and some give up. The guide to reducing customer effort counts chasing an update as one of the forms of effort customers are made to spend, and a backlog produces chasing on a large scale. Each chaser is another contact in the queue, so the backlog feeds itself. Inside the same queue, sorted by arrival, a customer with a safety worry or money taken in error waits behind a customer asking about a colour.

The theory of constraints explains where the effort should go. A system's output is set by the step that limits it, work piles up in front of that step, and time saved anywhere else only delivers work faster into the pile. In a support queue the step to check first is the time people have to read and answer. Every duplicate answered twice, every chaser answered as though it were new and every contact that a single message to a group could have resolved spends that time without bringing the oldest customer closer to a reply. The framework's order applies: find the limit and make full use of it before deciding to add capacity.

This tutorial is for the support lead or service manager whose queue has grown beyond its own standards after a campaign, a disruption, a system fault, a peak or a staff absence. It does not cover staffing, overtime, temporary or outsourced capacity, or forecasting how many people a queue needs; those are decisions for the business. It describes no help desk features such as bulk actions or automation rules, gives no reply-time standard, backlog size or clearance rate, and states no legal deadline for complaints; a business in a regulated sector should check the complaint rules that apply to it with a qualified adviser.

Before you start

  • Export the open contacts. Every open contact with its arrival time, channel, status, reason if it has one, the customer it belongs to and how many messages the customer has sent. Include contacts in pending statuses.
  • Write down the standards and the definitions. The reply standard for each channel, and the definitions the metrics guide asks a team to settle once: what counts as a reply, how merged and reopened contacts are counted and whether pending time stops the clock.
  • Find out what happened. What caused the spike, whether it is still running, and who owns it: the campaign owner, operations, the web team or whoever runs the system that failed.
  • Name a lead for the clear-up. One person decides the order of work, approves the acknowledgement and the group messages, and reports progress while the clear-up runs.
  • Have the support material to hand. The library of saved replies, the complaints procedure and the guidance on customers in vulnerable circumstances, so agents work from what is already approved.

The steps

  1. Count the backlog by age, reason and channel

    Take a snapshot of the open contacts and count them three ways. By age: how many are still inside the standard, how many are beyond it and how long the oldest has waited, with the oldest listed individually. By Contact Reason: what the customers are asking about. By channel: where they wrote in. Count the customers as well as the contacts, because a customer who wrote several times appears once for each message in the raw count, and report the contacts in pending statuses on a separate line.

    If contacts arrived untagged, tag a sample now with the reason list the reason-tagging tutorial builds from real contacts, so the counts show where the spike came from. Keep this snapshot: it is the baseline the clear-up is measured against, and the record the review starts from.

  2. Stop or slow what is feeding it

    Before changing how the queue is worked, look at what is still adding to it. The diagnosis page for rising support ticket volume ranks the reasons behind a rising contact rate and gives each a fix and an owner; use it to name the reason that is growing and send it to the team that can remove it, rather than repeating that work here.

    Where a campaign is driving the questions, that diagnosis says to write the support brief for the running campaign now rather than waiting for the review, and the campaign briefing tutorial sets out what the brief contains. Where the queue has shown up a mistake in a live ad or landing page, the SOP for live ad and page errors contains it, corrects it and briefs support first. Where the problem has become an incident, the marketing incident management playbook names an incident lead and makes the customer service lead one of the people told before customers start asking. During a sale or seasonal peak, the peak season trading playbook has the customer service lead report what customers are asking at the daily stand-up; bring the backlog count to that meeting too.

    Then protect the time that answers customers. Hold internal work that can wait, and keep agents' attention on the queue rather than on meetings about it. Whether to add people is a separate decision for the business, best taken once the feed has slowed and the avoidable work below has been removed.

  3. Tell everyone waiting what to expect

    Send a single acknowledgement to every customer in the backlog. Say that their message arrived, that replies are taking longer than usual, the time by which they can honestly expect a person to answer, taken from the clear-up plan rather than a hope, and what to do if the problem is urgent. Add links to the self-service answers that cover the largest reasons in the backlog. The effort guide, citing Nielsen Norman Group's work on how people read online, notes that readers scan help content for their answer instead of reading it through, so put those links near the top, labelled in the customers' words. Say plainly that writing again will not move them up the queue and is not needed.

    Keep it a service message. The proactive messages guide linked below relies on the UK regulator's guidance on direct marketing, through one of the site's email guides, for the point that a promotion inside a service message makes it read as marketing, so the acknowledgement carries no offer. Change the stated wait on the contact page and in the automatic replies at the same time, following the tutorial on designing the contact page, so a customer who checks there sees the same time. The acknowledgement is not a reply: each contact stays open until a person has answered it.

  4. Merge duplicates and chasers into one conversation per customer

    Find the customers who wrote more than once, wrote in two channels or sent a chaser asking for an update. Merge their messages into a single conversation that keeps the earliest arrival time, so the customer keeps their place, and every message they sent, so the reply covers everything. The metrics guide counts merged contacts once and treats a contact reopened by the customer as a Repeat Contact rather than a new one; apply both rules during the clear-up so the counts stay honest.

    Answer each customer once. Two agents replying separately to the same customer's email and chat waste the time the backlog is short of and can send two different answers. Chasers that arrive after the merge join the same conversation rather than starting a new one.

  5. Take urgent and sensitive cases out first

    Before working the rest, pull out the contacts that cannot wait in line: anything about safety, money taken in error or twice, damaged goods, a booking or delivery due before the queue would reach it, and access to an account someone else may have entered. The service and support hub warns against a queue sorted by arrival time, and routes order problems ahead of general questions; a backlog is when that routing counts for the customer. Each of these goes to a person, and the ones beyond an agent's authority go up the Escalation Route to whoever decides.

    Complaints keep their own process. The tutorial on writing a complaints procedure logs every complaint from every channel and gives each the business's acknowledgement and response promises; a backlog does not suspend them, and complaints found in the queue go into the log with their original arrival date. An agent unsure whether a message in the queue is a complaint applies the definition in the complaints procedure and, if still in doubt, treats it as one. Customers whose messages show that they need the process adapted, such as a bereavement, illness or sudden money difficulty, stay with a person and never receive a batch reply. The guide on customers in vulnerable circumstances sets out what agents may change and when to pass a case to a named person.

  6. Answer the rest in batches by reason, oldest first

    Group what is left by reason, setting aside any contacts about a single event, which the next step answers with one message before the batches start, and work each group with the oldest contacts first. An agent answering a run of contacts about the same thing has the same facts in front of them throughout, instead of switching subject with every message. Choose the order of the batches by the age of their oldest contact, so no reason's oldest customer is left behind while another reason is cleared.

    Start each batch from a Saved Reply in the library of saved replies. Check its facts with its owner before the batch begins, because a wrong fact in a reply sent to a whole batch reaches every customer in it. Fill in the parts every agent must change, respond to what the customer actually wrote, and respect the reply's "do not use for" rule: a contact that turns out to be a complaint or a different question leaves the batch and goes to the right place.

  7. Resolve whole groups with one proactive message

    Where a large part of the backlog is about the same event, such as a delay, a fault or a cancelled service, answer the whole affected group at once rather than contact by contact. The guide to proactive service messages finds the affected customers from the system that holds the commitment, checked against a second source where possible, and writes the message to lead with what changes for the reader, their choices and when they will hear next, with the same facts on a dated status line. Send it to everyone affected, including customers who have not written yet, because they may be the next contacts in the queue.

    Then work the matching contacts in the backlog. Where the message answers everything a customer asked, close the contact with a short note linking to it. Where the customer raised something the message does not cover, give them a personal reply. Give agents those facts ahead of sending, so the replies that follow agree with it.

  8. Keep sampling replies for quality

    A team under pressure to clear a queue has every reason to rush, and a rushed reply that does not solve the problem comes back. The tutorial for reviewing support conversations reads a weekly sample of real conversations against a short scorecard built from the business's standards. Keep that review running during the clear-up rather than pausing it, and include replies from the batches and the group messages in the sample.

    When the sample finds a mistake in a saved reply or a group message, correct the reply first, then send a short correction to every customer who received the wrong version. Share what the sample finds with the agents while the batches are still running.

  9. Never close or park contacts to stop the clock

    A backlog count invites gaming. Goodhart's law warns that a number people are rewarded for moving loses its value as a measure, since effort goes to whatever shifts it at the lowest cost. The cheapest ways to move a backlog are closing contacts that have not been answered, moving them into "waiting on customer", and sending a reply that says nothing. Each makes the number fall while the customer is still waiting, and leaves that customer to write again.

    Set a Guardrail Metric beside the backlog count and read them together: the age of the oldest open contacts, the time spent in pending statuses, contacts reopened by customers and repeat contacts by reason. Write down what closing a contact means during the clear-up, an answer that solves the problem or states the next step, and do not let the people judged on the count change that rule.

  10. Record the cause and the plan for next time

    When the oldest contact is back inside the standard, write a short review while the details are fresh: what fed the backlog, when the team knew, when customers were told, what each step above achieved, and the counts from the first snapshot beside the counts at the end. Make sure every contact closed during the clear-up carries its reason, tagged at close, so the next monthly reading of reasons includes the spike.

    Send the cause to its owner: an unbriefed campaign to marketing, a system fault to whoever runs the system, an unannounced delay to operations. Then write the plan for the next spike: who decides to send the acknowledgement, a draft of it ready to adapt, the order of triage from the table below, and any staffing decision the business took, recorded as its own decision with the evidence behind it.

Triage order: a table

The table sets out the order in which contacts are taken: duplicates and chasers are merged first, then the rest run from the ones that cannot wait to the ones worked in batches.

ContactsWhy they come at this pointHow they are answeredOwner
Duplicates and chasersAnswering them separately wastes time and risks conflicting repliesMerged into the customer's earliest contact and answered onceClear-up lead
Anything about safety or with a deadline the queue would missWaiting causes harm or a missed momentA person, with the escalation route for anything beyond their authoritySupport lead
Money taken in error, double charges and damaged goodsThe customer is out of pocket while they waitA person who can put it right, within their authoritySupport lead
ComplaintsThey have their own procedure and promisesLogged with the original arrival date and handled under the procedureComplaints owner
Customers who need the process adaptedA batch reply would add to their difficultyA person, with the adaptations agents may makeNamed person for adapted support
A whole group about the same eventA single message answers them all, including those who have not writtenA proactive message, then a short note or a personal reply on each contactOperations, with the support lead
Everything elseThe bulk of the work, best done with the same facts in viewBatches by reason, oldest first, from reviewed saved repliesAgents, with the clear-up lead

An archery shop clears a backlog of emails (illustrative)

Illustrative example — not a real result

Jostriqua Archery clears the queue a stock fault left behind

Jostriqua Archery, an invented online shop selling bows, arrows, targets and accessories, has 3 people answering email and chat and promises an email reply within 1 working day. On a Thursday a fault in its stock system stops order confirmations until the following Tuesday, and 610 orders go unconfirmed. By the Tuesday the email queue holds 742 open contacts from 498 customers, and the oldest has waited since the Thursday morning. 318 of those customers are asking about unconfirmed orders.

The web team fixes the confirmation emails first and sends the missing confirmations that afternoon. The owner sends a single acknowledgement to all 498 customers: replies within 3 working days, a link to the order-status page, chat for anything urgent, and no need to write again. Merging the customers who wrote more than once or in both channels leaves 498 conversations. 23 damaged-equipment cases and 11 double charges are taken out first and answered by the owner the same day.

A proactive message to all 610 affected orders confirms each order, gives its dispatch date and says the delay was the shop's fault. It answers 287 of the 318 unconfirmed-order conversations; the other 31 had a second question and get a personal reply. The rest are worked by reason with 6 saved replies the owner has checked. On the Wednesday a sample of 40 replies finds one saved reply giving an old returns address; it is corrected and the 17 customers who received it get a correction. Nothing is closed without an answer, and time in the waiting status is reported on its own line.

By the following Monday the oldest waiting contact is inside 1 working day again. Repeat contacts over the next 2 weeks run at 6% of conversations, the same as the month before the fault.

Common mistakes

  • Working the queue in arrival order. A safety worry or a double charge waits behind a question that could have waited.
  • Answering while the queue is still filling. Agents clear contacts from the front while the cause adds them at the back.
  • Saying nothing to the customers waiting. Silence leaves each waiting customer to chase, and every chaser is another contact.
  • Putting an offer in the acknowledgement. A promotion in a service message makes it read as marketing at the moment the customer wants an answer.
  • Answering the same customer twice. Separate replies to an email and a chat waste time and can contradict each other.
  • Batch replies to complaints and sensitive cases. A saved reply sent to someone who is bereaved or angry adds to the harm.
  • Replying contact by contact to a group problem. A separate answer to every contact about the same event takes far longer than a single message to everyone affected.
  • Pausing quality review to go faster. A mistake in a batch reply reaches every customer in the batch before anyone notices it.
  • Closing or parking contacts to make the count fall. The number improves and the customers are still waiting.
  • Clearing the backlog and moving on. Without the cause sent to its owner and a plan written, the next spike starts from the same place.

The service and support hub places this clear-up within the response standards and routing it restores.

Frequently asked questions

Is it acceptable to close old support tickets to clear a backlog?

No. Close a contact when the customer has had an answer that solves the problem or states the next step, not when it has waited too long. Closing unanswered contacts, or moving them into a waiting status, makes the backlog look smaller while the customers inside it are still waiting, and those customers come back as repeat contacts or complaints. Report the time contacts spend in waiting statuses separately so nobody is tempted.

Should every customer in the backlog get an acknowledgement?

Yes, a single message to everyone waiting, saying that their message arrived, when they can honestly expect a reply, what to do if the problem is urgent, where answers to the common questions are, and that they do not need to write again. It is a service message, so it carries no offers. It does not count as a reply, and the contact stays open until a person has answered it.

Is hiring or overtime the way to clear a support backlog?

Staffing, overtime and outside help are decisions for the business, and this tutorial does not cover them. Whatever is decided, first stop or slow what is feeding the queue and remove the work that does not need doing: duplicates, chasers and contacts a single message to an affected group can answer. Extra hands working a queue that is still filling, or answering the same customer twice, spend their time on work those steps would have removed.

Terms used on this page

Frameworks

Sources

  1. Direct marketing guidance — Information Commissioner's Office (UK)
  2. How Users Read on the Web — Nielsen Norman Group

Related