How-To Guide13 min read
How to Tell Which Support Ticket Pain Points Are Trending
Last updated: · By Matt Teixeira, Founder and CEO
On this page
If you get hundreds of support tickets and can't tell which pain points are trending, sort every ticket into a fixed list of problem themes and count unique accounts per theme each week, as a share of all accounts that wrote in. A theme is trending when that share beats its own eight-week baseline, not when its raw count goes up. Your help desk's AI covers one inbox, a spreadsheet and an LLM cover a few hundred tickets a month, and a feedback platform gives tickets, calls, and Slack one trend line.
Key Takeaways
- Freeze your theme list before you measure. Every regrouping breaks your trend lines.
- Count unique accounts, not tickets, as a share of everyone who wrote in. Raw counts rise with sign-ups, launches, and chatbot changes.
- Flag spikes (2x the eight-week average, at least 5 accounts) apart from steady climbs (share up in 3 of the last 4 weeks). Spikes go to engineering today; climbs go to the roadmap.
- A bug theme still above baseline two weeks after the fix is a product gap.
- Split by segment only after a theme flags. A theme that is flat overall can be doubling among new or enterprise accounts.
- Native tools see one inbox. Intercom Trends can't filter by segment, and Zendesk's intelligent triage reads only the subject and first public comment.
Why Can't You See Trends in Your Ticket Queue?
Because tickets are labeled for routing, not by problem, and nobody compares this week with a baseline. A field like "Technical issue" tells an agent where to send a ticket. It can hold six unrelated problems, so a real rise in one gets lost in the other five.
Step 1: Freeze Your Theme List Before You Measure Anything
Trend detection is a counting problem first and an AI problem second. If the groups shift every time you rerun the analysis, each chart compares different things.
- Name 12 to 20 themes after specific problems. That is more than the 8 to 15 product-area themes a taxonomy starts with, because a trend needs one problem per line. "Scheduled reports arrive empty" can trend. "Reporting" almost never does, because it averages a dozen problems together.
- Relabel history when you split or merge. If "Exports" becomes "CSV export timeouts" and "Missing export fields," re-sort the last eight weeks so both new themes start with a baseline.
- Keep an "Other" bucket and watch it. A rising Other is a trend in its own right: a problem that has no name yet.
For naming rules and safe ways to change the list, see the taxonomy guide.
Step 2: Label Every Ticket by Problem and Type
Give each ticket a theme and one of four types, because bugs and gaps trend differently and go to different owners.
- Bug: something that should work is broken.
- Product gap: the product can't do what the customer needs.
- How-to: the capability exists, but the customer couldn't find or use it.
- Account admin: password resets, invoices, and seat changes. Keep these out of product trend counts.
Read the whole thread before you label. A ticket that opens with "export is broken" often ends with "so it can't export more than 10,000 rows," which is a gap, not a bug. Zendesk's intelligent triage classifies from "a ticket's subject and the text of the first public comment" (Zendesk), so it tracks how tickets open, not what they turn out to be.
When a ticket raises two problems, label both. For a two-pass AI prompt that labels at scale, see how to use AI to analyze support tickets and sales calls.
Step 3: Count Accounts per Week, Then Divide by Volume
Your core number is unique accounts per theme per week, as a share of all accounts that filed a ticket that week. A customer who files nine tickets during one outage is one data point, not nine. The share strips out the rise that comes from simply having more customers.
| Metric | Use it for | It misleads you when |
|---|---|---|
| Raw ticket count | Staffing and queue planning | Volume rises for any reason, including one noisy account |
| Unique accounts per theme | Your default trend number | The theme is so small that 2 to 4 accounts swing by chance |
| Share of accounts that wrote in | Comparing one week with the next | A bot or help-center change alters who writes in at all |
| Accounts per 1,000 active accounts | Fast-growing customer bases | "Active" gets redefined from one month to the next |
Use Monday-to-Sunday weeks. Below about 1,000 tickets a week, daily buckets are noise except for incident watching.
Step 4: Separate Spikes from Steady Climbs
Spikes and climbs have different causes and different owners, so flag them with different rules. The thresholds below are rules of thumb for a few hundred to a few thousand tickets a week. Tune them after a month.
| Pattern | Flag it when | It usually means | Owner | Act by |
|---|---|---|---|---|
| Spike | Accounts this week are at least 2x the eight-week weekly average, and 5 or more | A regression, outage, or third-party change | Engineering and the support lead | Same day |
| Steady climb | Share rose in 3 of the last 4 weeks, and the 4-week share is at least 1.3x the 4 weeks before | A product gap, or a growing segment hitting a limit | Product manager | Next planning cycle |
| New theme | 3 or more accounts within 2 weeks of first appearing | A new use case, segment, or competitor move | Product manager | This week's review |
| Fading | Share fell 4 weeks in a row | A fix landed, or customers stopped bothering | Product manager | After checking usage and churn |
The floor matters more than the multiplier. Without a minimum of five accounts, a theme that goes from one account to three reads as "up 200%," and the team learns to ignore every alert.
Keep a dated log of releases, pricing changes, campaigns, outages, and holidays next to the chart. Most spikes have a calendar explanation, so check it before you call anything a trend.
In a spreadsheet, both rules take a few formulas. This layout uses Google Sheets syntax; in Excel, swap COUNTUNIQUE(...) for COUNTA(UNIQUE(...)).
Sheet "Tickets": one row per ticket and theme
A week_start | B theme | C account_id | D type | E segment
Sheet "Accounts": themes down column A, week starts across row 1 from D1
D2: =IFERROR(COUNTUNIQUE(FILTER(Tickets!$C:$C, Tickets!$A:$A=D$1, Tickets!$B:$B=$A2)), 0)
Sheet "Share": same layout, divided by every account that wrote in that week
D2: =Accounts!D2 / COUNTUNIQUE(FILTER(Tickets!$C:$C, Tickets!$A:$A=D$1))
Spike flag (this week in column L, the eight weeks before it in D:K)
=AND(Accounts!L2 >= 2*AVERAGE(Accounts!D2:K2), Accounts!L2 >= 5)
Steady-climb flag (on the Share sheet)
=AND(SUMPRODUCT(--(I2:L2 > H2:K2)) >= 3, AVERAGE(I2:L2) >= 1.3*AVERAGE(E2:H2))
Step 5: Tell Bugs Apart from Product Gaps
The shape of the line tells you which one you have. Bugs spike on a date and fall away once fixed. Product gaps climb slowly with no start date, because nothing changed except how many customers reached the limit.
Then run a two-week test: after a fix ships, a bug theme should be back near baseline within about two weeks. If it isn't, the bug was hiding a gap. The export timeout gets fixed, and the tickets keep coming as "export is too slow for an account our size."
Send bugs to your issue tracker with ticket links, and gaps to roadmap review with the list of affected accounts. To confirm a fix reduced complaints, see how to measure whether a release reduced customer complaints.
Step 6: Break Flagged Themes Down by Segment
Averages hide trends, so split every flagged theme by the segments that would change what you build. A theme that is flat across all customers can be doubling among accounts in their first 30 days. That points to onboarding, not the feature.
The cuts that most often change the answer are plan tier or ARR band, account age (first 90 days versus tenured), platform or app version, seat count, and region. Pull them from Zendesk custom fields or Intercom attributes, or join them from your CRM by account ID.
Split after a theme flags, not before. Cut 15 themes by 6 segments and you have 90 lines to watch, and a few will spike by chance every week. To check whether a flagged theme is also turning more negative, see how to track customer sentiment over time.
Step 7: Set Alerts and a Weekly Review
Match the cadence to the pattern. Spikes need a daily check or an alert, climbs and new themes need a 20-minute review each Monday, and the theme list needs a monthly cleanup.
An alert should name the theme, this week's accounts against baseline, the segment where it's concentrated, three quotes with ticket links, and an owner. If it fires more than about once a week per owner, raise its floor; if it hasn't fired in a quarter, lower it or delete it.
For the Monday review, the weekly signal brief playbook covers what's new, what's growing, and what's turning negative, with one quote per item. The weekly customer feedback report template compares each week with a shorter four-week average on purpose: it picks what to discuss this week, while the eight-week baseline here decides what counts as a trend.
Worked Example: One Week, Three Themes
This example is illustrative, with made-up numbers. A B2B SaaS company averages about 400 accounts filing tickets a week. This week, after a sign-up campaign, 480 did.
| Theme | 8-week average | This week | Share, average vs. this week | Verdict |
|---|---|---|---|---|
| CSV export errors | 8 accounts | 26 accounts | 2.0% vs. 5.4% | Spike. It started the day after Tuesday's release. |
| Role and permission setup | 9 accounts | 16 accounts | 2.3% vs. 3.3% | Steady climb. Share rose four weeks running. |
| Login and SSO | 20 accounts | 24 accounts | 5.0% vs. 5.0% | Not trending. The count rose 20%, but so did everyone writing in. |
Engineering rolls back the export change, and two weeks later that theme is back to 9 accounts: a bug. Eleven of the 16 permission accounts have 25 or more seats, so that theme goes to roadmap review as a gap for larger customers.
Login and SSO, the biggest theme by raw volume, gets no action. That is the point of the method: the largest theme is rarely the one that's trending.
Common Mistakes to Avoid
- Comparing with last week only. One quiet week makes the next look like a surge. Compare with the eight-week average.
- Trending on raw counts. A sign-up push lifts every theme at once and looks like a crisis.
- Re-theming every month. No theme keeps enough history to trend.
- Calling a fade a win. Tickets also stop when customers give up, so check usage or churn for those accounts first.
Tools That Help
Start with your help desk's own AI if all support lives in one inbox and you already pay for that tier. Move to a dedicated tool when feedback also comes from calls, Slack, or surveys, or when you need the segment split. For a ranked comparison, see the best tools to find recurring customer pain points.
| Option | Trend detection | Main limit | Cost |
|---|---|---|---|
| Zendesk intelligent triage and Explore | Prebuilt triage dashboard by topic and sentiment; custom weekly reports need Explore Professional (source) | Topics come from the opening message; Explore threshold alerts only recolor a live widget (source) | Suite or Support Professional and up (source) |
| Intercom Topics and Trends | Top 10 weekly shifts, each subtopic against its previous 12 weeks (source) | Needs Fin active and skips migrated conversations (source); no segment filter | Pro add-on, from $100/month (source) |
| Spreadsheet plus an LLM | The formulas in Step 4 | You maintain the export and rerun it | Free to low, plus your time |
| SentiSum | Anomaly and fixed-count alerts on a tag's volume, daily or hourly, by email and Slack (source) | Support and CX focus | Annual contract, price not published; free scoped audit (source) |
| Enterpret | Trend + Anomalies chart, default z-score cutoff of 1.5 (source) | Built for high-volume enterprise teams | $120,000/year AWS listing (source) |
| Deck | Themes Surges view; subtheme Trending and Sudden Changes sorts, rebuilt weekly | No alerts or custom thresholds; counts insights, not accounts | Free plan; Zendesk and Intercom from $99/month on Starter |
If you're on Intercom with Fin live, turn on Trends before you buy anything; it already compares every subtopic with 12 weeks of history. What it can't do is the segment split, because you "cannot control which conversations Trends looks at" (Intercom). On Zendesk, triage topics reflect the opening message, so pair them with your own ticket fields or an outside tool. The Intercom and Zendesk guides cover exporting full threads for either path.
Where Deck Fits
Deck does the sorting, typing, trend math, and segment splits across your tickets and other feedback sources, so you don't maintain the spreadsheet.
- Trends: the Themes Surges view compares each theme's recent insights with the 14 days before (Deck MCP docs). Subthemes, on Starter and up, sort by Trending and Sudden Changes and rebuild every Sunday evening UTC (Deck subthemes docs).
- Type and segment: each insight keeps its quote, source link, sentiment, and a category such as Bugs & Errors or Feature Requests (Deck themes docs). Segment rules read Zendesk and Intercom tags and custom fields, and the Segment filter re-counts every view (Deck segments docs).
- Baseline: Zendesk and Intercom pull the last 90 days on first connect, then sync once a day (Deck Zendesk docs). Surges dates each ticket by when it was created, but subtheme trends date it by when Deck processed it, so a backfill reads as recent there for two weeks.
- Cost: Zendesk and Intercom need Starter or above. At Deck's planning estimate of about 3 credits a ticket, 300 tickets a week is roughly 3,900 credits a month: about $169 on Starter ($99 with 2,500 credits, plus overage) or $249 on Business ($199 billed annually).
- Limitations: Deck doesn't send spike alerts or let you set thresholds; the closest is a weekly Slack summary of new feedback (Deck Slack docs). Trend numbers count insights, not unique accounts, and aren't divided by ticket volume, so keep the Step 3 share alongside them. Syncs run once a day, so Deck won't replace incident monitoring.
To try the Surges and Trending views on your own tickets, compare Deck's plans.
Frequently Asked Questions
Sources
- Zendesk Help: Automatically detecting customer intent, sentiment, and language
- Zendesk Help: Overview of the Intelligent triage dashboard
- Zendesk Help: About the Zendesk Explore plan types
- Zendesk Help: Adding alert thresholds to Explore live data components
- Intercom Help: How to use Trends to spot shifts in your support data
- Intercom Help: Use the Topics Explorer to see what's driving volume
- Intercom pricing
- SentiSum Help: How smart alerting works
- SentiSum pricing
- Enterpret Help Center: Anomaly
- Enterpret on AWS Marketplace
- Deck MCP docs
- Deck subthemes docs
- Deck themes docs
- Deck customer segments docs
- Deck Zendesk integration docs
- Deck Slack insights delivery docs
- Deck pricing
Sources checked November 2, 2026.