An engineer in your network operations centre gives notice on a Thursday. By Monday you hear it was the pay. By the next quarter two more have left and nobody mentions pay at all. The pattern points elsewhere. Below are the reasons that recur in helpdesks, NOCs and cages, with ways to test each against your own queue data.
It covers managed service providers, support desks, cloud practices, data centre teams and network integrators. The wider hub is IT infrastructure and managed services. For the general retention approach, see employee retention and attrition management.
The pager is the first reason
On-call rotas look fair on a spreadsheet. In practice, the engineer who knows the firewall estate gets every midnight call, while others sleep. Alert noise adds to it. A monitoring tool that fires two hundred low-value alerts a night teaches people to stop trusting it, and then to stop caring.
- Count pages per engineer per month, not just rota slots.
- Separate real incidents from noise and track the noise ratio.
- Record sleep-hour interruptions and give time off the next day.
- Rotate the difficult clients, not only the shifts.
The ladder from L1 to L2 stalls
L1 desk staff watch a ticket queue, follow scripts and escalate. Many join hoping to reach L2 or L3 within a year. If nobody can say what separates the levels, they leave for a firm that can. A written ladder lists the tasks, tools and shadow hours needed for each step.
- 1
L1 to L2
Handles password resets and triage alone for a quarter, resolves a set of ticket types without escalation, runs a patch window under supervision.
- 2
L2 to L3
Owns a customer's change requests, writes a runbook, leads a post-incident review.
- 3
L3 to architect or team lead
Designs for a client, mentors, signs off on risk in a change advisory call.
Customers who treat engineers as their own staff
A client shouts at an on-site engineer on a Friday evening because a switch stack failed. The engineer reports to your firm, but sits in the client's building. He has no one to speak to. Field technicians and embedded engineers quit over this isolation more often than over rack-and-stack work itself.
Visit them. Pair them with a peer at your office. Give them a say when a contract renewal decides whether they stay at that site.
Skill premiums that nobody explains
Cloud administrators and security analysts fetch market premiums. A helpdesk technician sees the gap in pay and says nothing. When he later learns the cause, trust drops. Publish how pay bands move with skills, tools and billable rates. See the notes on role design for MSPs.
SLA penalties land on the person
A P1 ticket has a fifteen-minute response clock. A missed clock costs the firm service credits. The account manager passes the pressure to the shift lead, and the shift lead passes it to whoever picked up the ticket last. Engineers learn that the dashboard blames individuals, not queues. Watch for this in how breach reviews are run.
- Read breach reports aloud as queue problems first: staffing at 3 a.m., a missing runbook for a SAN controller fault, a stale escalation matrix.
- Ban naming a single engineer in client-facing breach summaries.
- Give the shift lead authority to pull in a second-line engineer without asking the account manager.
Backup restore tests, hypervisor patch windows and VLAN migrations all happen out of hours. People remember who stayed back and who got thanked. A simple weekly note of thanks, naming the change ticket and the job done, costs nothing and is rarely done.
Hearing the real reason
The form handed over on the last day tells you little, since leavers guard their reference. Read the queue instead.
- 1
Ticket and on-call data
Look at pages, reopen rates, handover gaps and shift swaps for leavers in their last three months.
- 2
Stay talks
Hold a short conversation at month eight and again at month sixteen with every engineer. Ask what they would change about the rota.
- 3
Leaver call after sixty days
A neutral person, not the manager, asks what drove the exit now that nothing is at stake.
- 4
Pattern review each quarter
Group the reasons by team, client and shift. Act on patterns, not single stories.
Skills sit alongside this. Read competency and talent management for ways to map who can do what.
Data centre and network teams differ
Data centre technicians work rack aisles, UPS rooms and cooling plant, often on rotating shifts with strict access control. Their complaints are about shift predictability and safety. Network integrators lose people to project fatigue: three cutovers in one weekend, with no comp-off after. Treat each group on its own terms: the helpdesk wants a ladder, the cage technician wants fixed hours and clean air handling, the integrator wants cutover weekends capped.
