A quiet week in the support queue feels like a healthy customer base. It usually isn't. A queue sorts the people who raised a hand, so it can only ever show you the customers still bothering to complain. The account that argued with you every week and has said nothing in 90 days never shows up. That drop is the signal, and here is how to pull the list.
TLDR: A support queue is a sorting system for the people who already raised a hand. It ranks them by urgency and tells you nothing about the customers who stopped bothering. So a quiet week reads as a healthy base when it is often a fade. The number to watch is not how many tickets you have, it is which accounts used to file them and have gone silent. Pull the accounts with at least two tickets in days 91 to 180 and zero in the last 90, and you have a call list your dashboard will never generate for you.
Why does a quiet support queue feel like good news?
Because a queue is built on battlefield triage, and triage is very good at one job: deciding who gets treated first among the people who made it to the tent. Priority levels, first response time, CSAT, backlog age. All of it operates on a population that has already self-selected by contacting you.
That is fine for running a support team. It is useless for reading retention, because the interesting customers never walk into the tent. This is survivorship bias wearing a helpdesk badge. You are studying the wounded who reached help and drawing conclusions about the whole field.
The scale of the gap is not subtle. The most-quoted figure in customer service, originally from Lee Resources and repeated in every stats roundup including Help Scout's, is that for every customer who complains, around 26 stay silent. Esteban Kolsky's research at ThinkJar landed in the same place: roughly 1 in 26 unhappy customers actually complains, and the rest just leave. Whichever number you trust, your ticket queue is a rounding error against your unhappy base.
The orange band is the entire point of this article. No helpdesk report produces it by default.
What is silent churn, and why does the queue miss it by design?
Silent churn is the ordinary way SaaS customers leave. Usage tapers, tickets stop, emails go unanswered, and then the renewal quietly does not happen. Nobody sent an angry message. Nobody asked for a discount. There was no save opportunity because there was no conversation.
The queue misses it because a ticket is a request, and requests come from people who still believe you will act. When somebody stops filing tickets, the most common explanation is not that their problems disappeared. It is that they stopped expecting you to fix them. That is a much later stage of the relationship than an open P2.
A customer arguing with you is spending energy on you. That energy is the cheapest renewal signal you will ever get, and it disappears before the cancellation does.
Harvard Business Review made the adjacent point back in 2010 in Stop Trying to Delight Your Customers: what drives disloyalty is the effort a customer has to spend to get something resolved. High-effort interactions push people toward the exit. The bit teams skip is what happens next. Once a customer decides the effort is not worth it, they stop making it, and your queue records that surrender as a reduced backlog.
The three kinds of silence, and only one of them should scare you
Not all quiet accounts are the same, and treating them as one bucket is how good outreach lists turn into spam. Here is the split I use on every teardown.
Account pattern
What the queue shows
What it actually means
What to do
Still talking
Open tickets, recent history
Still invested. Still expects you to fix things
Resolve well and fast. This is your safest cohort
Always quiet
Nothing, ever
No baseline, so no information. Could be healthy self-serve or a dead seat
Ignore the silence. Read product usage and seat activity
Went quiet
Nothing, and a shrinking backlog you read as a win
They stopped expecting you to fix it
Call them. This week, before renewal season
Complaining loudly
P1s and a bad CSAT score
Annoyed, engaged, and usually renewable
Fix it properly and tell them you did. Do not panic
The row that surprises people is the last one. The customer calling your onboarding a mess is not your renewal risk this quarter. They are still arguing, which means they still care what you do next. I would take ten of those over one account that used to write paragraphs and now writes nothing.
Get weekly churn reduction tips.
No spam, unsubscribe anytime.
You're in.
Check your inbox shortly.
Panels one and three both end quiet. Only one of them is a problem.
What is your drop-off list actually worth?
Before you spend an afternoon building the list, it helps to know the size of the prize. Drop in last month's numbers and this runs the same split the video describes, then puts revenue against the group nobody is looking at.
The drop-off list calculator
Four inputs. It separates the accounts your queue renders from the ones it hides.
Still talking
98
your queue shows these
Went quiet
42
nobody is looking
Never talked
260
judge on usage
42 accounts that used to file tickets have filed nothing in 90 days. Call them this week, before the renewal conversation calls you.
Revenue sitting on the drop-off list: $10,500 / month, which is $126,000 a year. If one in three of them leaves without ever telling you why, that is $41,580 of annual revenue that never generated a single support ticket on its way out.
Where these numbers come from: the middle two sliders are just a cohort split, so the only assumption baked in is the last one. The one-in-three figure applied to the drop-off list is deliberately conservative. It is well below the churn rate you would expect from a fully dormant account and well above your base rate, because a drop-off cohort is a mix of accounts already gone, accounts drifting, and accounts whose champion left and whose replacement never learned to ask you for anything. Swap in your own number once you have tracked one list through a renewal cycle. The point of the calculator is not the third decimal place, it is that the group with the largest revenue attached to it is the only group your support tooling does not report on.
How do I pull my drop-off list this week?
This is a one-hour job the first time and a saved view after that. You need ticket history joined to accounts, and nothing else.
Export 180 days of tickets with the account attached. In Zendesk that is the search API or an Explore export. Intercom exposes it through list conversations, Freshdesk through the tickets endpoint, and Help Scout through the Mailbox API. If your tickets are keyed to individual people rather than companies, roll them up by email domain. It is imperfect and good enough.
Count tickets per account in two windows. Days 91 to 180 is the baseline. Days 1 to 90 is the recent window. Two counts per account, one pivot table.
Filter to the drop. Keep accounts with two or more tickets in the baseline window and zero or one in the recent window. Two is the threshold that matters. A single old ticket is not a habit, so it cannot break.
Join revenue. Pull current MRR per account from Stripe subscriptions or whatever bills them, and sort the list descending. You now have a call list ranked by what it costs you to be wrong.
Cross-check against product usage. Overlay weekly active seats from Mixpanel, Amplitude, or PostHog trends. Tickets down with usage steady is usually fine. Tickets down with usage down is the account you clear your afternoon for.
Look at the last ticket each one filed. Read the final three messages from every account on the list. This is the part people skip and it is where the actual answer lives, because the last thing they asked for is very often the thing that made them stop asking.
If you would rather have a system doing this on a schedule than a spreadsheet you rebuild each quarter, the health score monitoring playbook covers wiring engagement deltas into an alert, and the at-risk account triage agent post walks through automating the read of those last three messages. The general case is covered in how to predict churn without ML, which is mostly this same trick applied to four other signals.
What do you actually say to an account that went quiet?
Not a check-in. Not a survey. Definitely not "just circling back". A generic touch from a vendor you have already mentally written off is easy to ignore, and it confirms that nobody read anything.
Open with the specific thing they last raised, and ask whether it ever got sorted properly.
"You flagged in March that the CSV export was timing out on your bigger accounts. Did that ever get resolved on your side, or did you end up working around it?"
It works because it is answerable in one line and it proves somebody actually looked. The workaround branch is the reply you need most: a customer who built a workaround has already started routing around your product, which is the step before routing around your invoice. Nielsen Norman Group's guidance on interviewing users applies here: ask about the specific past behaviour, not about feelings in general.
What comes back tends to fall into a few buckets, and each one has a different next move. A champion left, so you need a new one and the usage-drop playbook is the wrong tool. They are trialling a competitor, which is what detecting competitive evaluation is for. Or their use case moved and you never noticed, which is a product conversation and belongs in your feedback loop rather than a save offer.
Where this framing breaks down
I want to be honest about the limits, because the drop-off list is not a churn oracle and treating it like one will burn your credibility internally.
It does not work for self-serve products with thousands of accounts and almost no support contact. If 95% of your base has never filed a ticket, your baseline population is too small to build a program on, and you should be reading activation and usage instead. Start with the customer health score approach and billing-data signals in that case.
It also produces false positives, and plenty of them. Some accounts go quiet because you genuinely fixed the thing that annoyed them, and calling those people to ask why they have not complained recently is an odd conversation. That is why the usage overlay in step five is not optional. Silence plus healthy usage is a win you should log rather than a risk you should chase.
And it will not catch the fastest churn, the accounts that never got far enough in to develop a support habit at all. Those leave through the front door in the first 60 days, which is a different problem with a different fix. That one is about time to value, covered in trial length and time to value and the front door problem.
Every serious retention practitioner writing today lands somewhere near this point, whether it is Lincoln Murphy on desired outcomes, Lenny's Newsletter on engagement depth, the Reforge material on retention as an output of activation and engagement, or the operator posts on Gainsight, ChurnZero, Vitally and Intercom. Support contact is one input among several. It just happens to be the one sitting in a system you already pay for, unqueried.
Why the queue keeps getting read as a health metric
Because it is the only customer-facing number that updates every hour and has a green arrow on it. Backlog down, first response time down, CSAT up. All genuinely good operational news, and all measured on the surviving population.
It is the same failure mode as reading a single blended churn percentage instead of the retention curve underneath it, or trusting an NPS score built from the 12% of customers who answered the survey, which is what your NPS is lying to you is about. Response-based metrics measure the responders. Zendesk's CX Trends reporting, Qualtrics on NPS methodology, and the service data HubSpot and Paddle publish all carry the same caveat somewhere in the footnotes, and everyone reads the headline instead.
The fix is boring. Add one column to your weekly retention review: accounts whose ticket count dropped to zero. It takes a query. It will be the only column on the page that describes customers who are not currently talking to you.
Start with the diagnosis, not the outreach
If you only do one thing from this article, pull the list. If the list turns out to be large and expensive, the next question is which leak it belongs to, because a drop-off cohort caused by a broken onboarding needs a completely different response to one caused by a champion turnover problem.
The free Churn Health Check is 8 questions and about a minute, and it tells you which of those you are dealing with before you commit a quarter to the wrong fix. If you want to see how your overall number compares to companies your size first, Retention Rank puts you on a percentile, and the MRR Impact Simulator shows what the drop-off list is worth over 12 months at your scale. From there, the leverage audit and the priority finder rank what to do first, and the full experiments library has the playbook for whichever answer comes back. Diagnose first. The outreach is the easy part.
Free interactive tool
Score your retention setup in 60 seconds
8 questions. Get your tier (Critical to Best-in-Class), your weakest spots, and 3 specific things to fix next.
Answers to the questions I get most often about this topic.
Do support tickets predict churn?
Ticket volume on its own predicts almost nothing, because it only measures the customers willing to contact you. What does predict churn is the change in ticket volume for a single account. An account that filed six tickets a quarter for a year and has filed none in the last 90 days has changed its behaviour toward you, and that change is measurable. Absolute volume is noise. The delta per account is signal.
What is silent churn in SaaS?
Silent churn is when a customer disengages and eventually cancels without ever telling you what went wrong. They stop logging in, stop filing tickets, stop replying to your emails, and then the renewal simply does not happen. It is the default way most SaaS customers leave. Roughly one unhappy customer in 26 actually complains, which means the complaint queue is the smallest and least representative slice of your unhappy base.
Does a quiet support queue mean my customers are happy?
No, and this is the trap. A quiet queue means either your product got easier or your customers stopped expecting you to fix things. Those two states look identical on a support dashboard and mean opposite things for renewal. The only way to tell them apart is to check whether product usage held steady while tickets fell. Usage flat plus tickets down is a genuine win. Usage down plus tickets down is a fade.
How do I build a support ticket drop-off list?
Export tickets for the last 180 days with the account attached to each one. Count tickets per account for days 91 to 180, then for days 1 to 90. Keep every account with at least two tickets in the older window and zero or one in the recent window. That is your drop-off list. In Zendesk, Intercom, Freshdesk, or Help Scout this is one search plus a pivot table, and it usually takes under an hour the first time.
Is a customer who complains a lot more likely to churn?
Usually the opposite, as long as you resolve the complaints. A customer who files detailed tickets is still spending energy on you, still assumes you will fix it, and is still invested enough to be annoyed. The dangerous version is the customer whose complaints got progressively shorter and then stopped. Effort spent on you is a form of commitment, so a decline in that effort matters more than the tone of any single ticket.
What time window should I use to detect a support drop-off?
Ninety days against the previous ninety works for most B2B SaaS on monthly or quarterly billing. If your product is seasonal or your customers only touch it at month end, widen both windows to 180 days so a normal quiet stretch does not look like a fade. For high-frequency products where accounts contact you weekly, tighten to 30 days against 30. Match the window to the natural rhythm of the account, not to your reporting calendar.
What should I say when I reach out to an account that went quiet?
Do not open with a check-in or a survey. Reference the specific thing they last raised and ask whether it ever got resolved properly. Something like "you flagged the export timing out back in March, did that ever get sorted, or did you find a workaround?" It gives them a concrete thing to answer, it proves you were paying attention, and the workaround answer is the one that tells you they have already started routing around your product.
Should accounts that never contact support be on the drop-off list?
No, keep them separate. An account with no ticket history has no baseline, so silence there carries no information at all. Some of those accounts are perfectly healthy self-serve users and some are dormant seats nobody ever adopted. Judge that group on product usage and seat activity instead, and reserve the drop-off list for accounts whose behaviour toward you visibly changed.
Founder of ChurnTools. I spend my time studying how SaaS companies lose customers and building tools to help them stop. Previously worked in SaaS growth and retention across multiple B2B products. I also write about growth and answer-engine optimization (AEO) at growthpigeon.com.