GoldenDane Process X-ray · demo on public data← Back to goldenda.com
Process X-ray · IT help desk

Tickets take 40 days to close. The fix takes 6.

We ran our Process X-ray on 4,580 real support tickets (21,348 logged steps, 2010–2014) from the IT help desk of an Italian software company. We turned the log into a process graph and a handover network between 22 staff accounts. Here's where the time goes, and what we'd fix first.

39.9 days
Median time from ticket opened to closed
6.0 days
Median time until the issue is actually fixed
29.9 days
Fix time for the slowest 10% of tickets
226
Different paths through what is meant to be a 4-step process
Exhibit 1 · The process as it really runs

The process map, rebuilt from the log

Each box is a step; each arrow is a hand-off from one step to the next. Only 52% of tickets follow the textbook path (assign → take in charge → resolve → close). Hover any arrow for its numbers.

Normal transitionSlow transition: median wait 5 days or moreLine width = number of tickets · label = tickets · median wait
G start Ticket opened n0 Assign seriousness 4,938 events start->n0 4,384 n3 Insert ticket 118 events start->n3 118 n6 Take in charge ticket 5,060 events start->n6 74 end End n0->n0 439 · median 1 h n5 Resolve ticket 4,983 events n0->n5 258 · median 2 h n0->n6 4,160 · median 6 h n7 Wait 1,463 events n0->n7 69 · median 4 h n1 Closed 4,574 events n1->end n2 Create SW anomaly 67 events n2->n5 42 · median 16 d n3->n0 111 · median <1 h n4 Require upgrade 119 events n4->n5 82 · median 15 d n5->n1 4,558 · median 30 d n5->n5 254 · median <1 h n5->n6 154 · median 1.3 d n6->n2 53 · median <1 h n6->n4 72 · median 3.1 d n6->n5 3,563 · median 3 h n6->n6 99 · median 18 h n6->n7 1,265 · median 21 h n7->n5 773 · median 9.0 d n7->n6 558 · median 5.0 d n7->n7 113 · median 24 h
Exhibit 2 · Where the time goes

72% of all ticket time is a waiting window after the fix

After a ticket is resolved, it stays open for a fixed confirmation window before it closes. That window was about 30 days until 2012; the data shows it dropped to 15 days in 2013. It isn't work, but it makes "time to close" look more than six times longer than "time to fix" in every report.

Share of all elapsed ticket time, by what the ticket was waiting for4,580 tickets
Closing window after the fix: 72% of all elapsed ticket timeClosing window after the fix72%Hands-on handling: 11% of all elapsed ticket timeHands-on handling11%Queue before someone picks it up: 9% of all elapsed ticket timeQueue before someone picks it up9%Parked in 'Wait' status: 8% of all elapsed ticket timeParked in 'Wait' status8%Escalated to development: 1% of all elapsed ticket timeEscalated to development1%
Exhibit 3 · The real problem is the long tail

36% of tickets take more than two weeks to fix, and they hold 82% of all fix-days

Most tickets are fast: 29% are fixed within a day. The average fix time (11.1 days) is dragged up by a minority that gets stuck. Those are the ones customers remember.

Days from ticket opened to first fixUnder 14 days14 days or more
02004006008001,0001,2001,4001,6001,8000–2 days: 1,651 tickets2–4 days: 347 tickets4–6 days: 292 tickets6–8 days: 253 tickets8–10 days: 121 tickets10–12 days: 100 tickets12–14 days: 139 tickets14–16 days: 190 tickets16–18 days: 257 tickets18–20 days: 172 tickets20–22 days: 172 tickets22–24 days: 113 tickets24–26 days: 111 tickets26–28 days: 106 tickets28–30 days: 101 tickets30–32 days: 60 tickets32–34 days: 74 tickets34–36 days: 75 tickets36–38 days: 53 tickets38–40 days: 49 tickets40–42 days: 61 tickets42–44 days: 43 tickets44–46 days: 23 tickets46–48 days: 1 tickets48–50 days: 0 tickets50–52 days: 0 tickets52–54 days: 1 tickets54–56 days: 1 tickets56–58 days: 0 tickets58–60 days: 3 tickets0102030405060 daysmedian 6 days
Exhibit 4 · What makes a ticket slow

Three patterns separate fast tickets from stuck ones

Median days to fix, comparing tickets with and without each pattern. These are associations in the data, not proof of cause, but they show where to look first.

Waiting to be picked up

0510152025≤ 1 day (2,478 tickets): median 1.2 days to fix1.2≤ 1 day2,478 tickets> 3 days (1,221 tickets): median 18.0 days to fix18.0> 3 days1,221 tickets

Median pickup is 6 hours, but 29% of tickets wait over 3 days, and those hold 92% of all queue time.

Parked in "Wait" status

0510152025Never parked (3,342 tickets): median 2.8 days to fix2.8Never parked3,342 ticketsParked (1,227 tickets): median 18.9 days to fix18.9Parked1,227 tickets

27% of tickets get parked. They take 18.9 days to fix against 2.8 for the rest.

Passed between people

05101520250 handovers (1,570 tickets): median 1.0 days to fix1.00 handovers1,570 tickets1 handover (2,219 tickets): median 6.3 days to fix6.31 handover2,219 tickets2+ handovers (780 tickets): median 18.4 days to fix18.42+ handovers780 tickets

Every handover before the fix multiplies the time. Tickets solved by the first person take 1 day.

Also: the 3% of tickets escalated to development (upgrade or software bug) take 26.2 days, against 5.7 for the rest.

Exhibit 5 · The handover network

One person sits in the middle of everything

This is the social graph of the help desk: who passes tickets to whom. Agent 2 handles 20% of all steps, is involved in 49% of the handovers between agents, and sits on 2.3× more of the shortest paths in the network than anyone else (betweenness centrality). That is a bottleneck in busy weeks and a risk when Agent 2 is away.

Most central personAgent (size = steps handled)Closing desk (automatic closing accounts)Arrow width = number of handovers
H Closing desk Closing desk Value 2 Agent 2 Value 2->Closing desk Value 1 Agent 1 Value 2->Value 1 Value 9 Agent 9 Value 2->Value 9 Value 4 Agent 4 Value 2->Value 4 Value 14 Agent 14 Value 2->Value 14 Value 12 Agent 12 Value 2->Value 12 Value 7 Agent 7 Value 2->Value 7 Value 1->Closing desk Value 1->Value 2 Value 1->Value 9 Value 13 Agent 13 Value 1->Value 13 Value 1->Value 4 Value 6 Agent 6 Value 1->Value 6 Value 9->Closing desk Value 9->Value 2 Value 13->Closing desk Value 13->Value 2 Value 8 Agent 8 Value 8->Closing desk Value 8->Value 2 Value 8->Value 9 Value 8->Value 13 Value 4->Closing desk Value 4->Value 2 Value 6->Closing desk Value 14->Closing desk Value 14->Value 2 Value 12->Closing desk Value 16 Agent 16 Value 16->Closing desk Value 7->Closing desk Value 15 Agent 15 Value 15->Closing desk Value 11 Agent 11 Value 11->Closing desk
What we'd do

Five changes, in order of effort

Estimates come straight from the log and assume the team has the capacity to make the change; a pilot would test them.

1. Measure time-to-fix, not time-to-close

Report the fix date as the headline KPI and shorten the confirmation window to 7 days, with one-click reopening for customers.

Median time-to-close on paper: 39.9 → 13.7 days, with no change in the work.

2. A one-working-day pickup rule

Flag every ticket not taken in charge within one working day, and give someone ownership of the queue each day.

Average fix time: 11.1 → ~8.2 days if no ticket waited more than a day for pickup.

3. Give "Wait" a reason and a deadline

Require a reason code when parking a ticket, and sweep every parked ticket weekly. 27% of tickets pass through this status today.

Targets the group that takes 18.9 days to fix instead of 2.8.

4. Route tickets right the first time

Use the ticket's product, customer and description to suggest the right agent at intake. This is where an AI classifier pays off.

Fewer handovers: tickets solved by the first agent take 1 day, not 6–18.

5. Reduce the dependency on one person

Cross-train two agents on the ticket types that only Agent 2 handles today, and route those types to them first.

Less queueing in peak weeks, and no single point of failure.
Read this with care

Limits of this demo

How a Process X-ray works

Two weeks, one export, a clear answer

Step 1 · Day 1

You export a log

Four columns from your ERP, ticketing or order system: case ID, step, timestamp, person or team. No system access needed.

Step 2 · Days 2–8

We rebuild the process

Process map, waiting times, rework loops, handover network and the paths that cost the most.

Step 3 · Days 9–10

You get the X-ray

A report like this one on your own data, the top fixes with estimated impact, and a walkthrough with your team.

GoldenDane · Applied AI for operations teams

Want to see this on your own process?

Orders, invoices, service tickets, quality deviations: if a system logs it, we can X-ray it. Send us a note and we'll tell you within a day what an export would need to contain.

hello@goldenda.com

Data: the public research dataset "Dataset belonging to the help desk log of an Italian Company" by Mirko Polato (2017), published on 4TU.ResearchData under the 4TU General Terms of Use. Analysis and graphs by GoldenDane ApS with Python, pandas, NetworkX and Graphviz. Staff accounts are anonymised in the source data; "Agent" numbers follow the dataset's own codes.

© 2026 GoldenDane ApS · CVR 46351789 · Copenhagen · goldenda.com · Privacy