Alert Enrichment: Tickets That Arrive with Context
A technician opens an auto-created ticket. The title says "sw-acc-14 unreachable for 10 minutes." Before they can do anything useful they need to answer a string of questions. What is sw-acc-14? Where is it? What plugs into it, and are those things down too? Who owns it? Did anyone change something on it recently? Has this happened before, and what fixed it last time? Where is the graph? Is there a procedure for this? Every one of those answers lives in some system somewhere. The technician spends the first twenty minutes of the incident being a search engine. I have been that search engine. Nobody trains you for it.
Ticket enrichment is doing that search automatically, at creation time, so the ticket arrives with the answers already in it. It is the difference between a ticket that says "something is down" and a ticket that says "the access switch in the second-floor east closet, owned by the network team, feeding twelve desks and one printer, last seen down three months ago when a failing power supply was replaced, graph here, runbook here, first step: check the closet PDU." (A PDU, or power distribution unit, is the rack's power strip.) One of those tickets gets worked; the other gets triaged.
This guide explains what to add, where each piece comes from, and how to lay it out. It explains how to build it up in stages so the first version takes an afternoon rather than a quarter. It also draws a clear line between enriching the alert and enriching the ticket, because they are different jobs with different budgets.
In this guide
- Alert-level versus ticket-level enrichment
- What to add: the enrichment inventory
- Where each field comes from
- The enriched-ticket template
- A before-and-after walkthrough
- Automated versus manual: start with three fields
- Keeping enrichment truthful
- Where to go next
Alert-level versus ticket-level enrichment
Enrichment can happen at two points, and this guide covers one of them. Alert-level enrichment adds context to the notification itself — the email or the page. Someone reading it on a phone at 3 a.m. can then decide in the first minute whether to get out of bed. That context has to be short, because the reader has seconds. It is limited to what the monitoring system knows or can attach cheaply. The guide on event enrichment covers that first-minute problem in depth.
Ticket-level enrichment is a different job. The ticket is where a person will spend the next hour, or the next three days. It is read on a full screen and it persists. And it can be filled in a little later than the alert, because the technician is not going to open it in the first thirty seconds anyway. That changes the budget completely. A ticket can carry a paragraph of device detail, a list of affected downstream devices, and three links. It can carry an excerpt from the last change record and a summary of the previous incident. And all of that is welcome rather than overwhelming. It can also be assembled from more sources than the monitoring system alone: the inventory, the change log, the ticket archive.
A useful way to hold the distinction: the alert answers "should I care right now?" and the ticket answers "what do I need to know to fix this?" The same fact — the device's location, say — may appear in both. But the alert gets one line and the ticket gets the floor, the room, the rack, and who has the key.
What to add: the enrichment inventory
This is the full list of what a well-enriched ticket carries. Not every ticket needs every item, and the section after next shows how to start with three. But this is the target.
Device details
Include the role (what the device is for: access switch, file server, branch firewall). Include the model class (a stacked switch, a virtual machine, a rack server — class, not the exact part number, unless the inventory has it). Add the physical location down to the closet or rack, and the owner. The owner is the person or group who can answer questions about it and approve changes to it. That is not always the group that gets the ticket. A device with no owner in the inventory is itself a finding worth a note.
The service it affects
This field states what the business loses while this device is down, in words a non-technical person would recognise. Examples are "second-floor east desks," "branch VPN," "the print server for the warehouse." This is the field that lets the help desk answer the phone intelligently before the technician has even opened the ticket. (The phone is already ringing.)
The dependency picture
Two lists. Downstream: what else is unreachable because this device is down, which the monitoring system knows if it has dependency rules configured. Upstream: what this device depends on, so the technician can check whether the real fault is one hop back. Suppose the dependency picture says "twelve devices behind this switch, all currently unreachable." The technician knows in one glance that the scope is the switch, not a single desk. How dependency relationships are discovered and modelled is a subject of its own, covered in dependency thinking.
Recent changes on or near the device
Any change record touching this device, its upstream neighbour, or its site within the last several days. The single most common cause of an outage is a change somebody made. And the single most common reason it takes an hour to find is that nobody thought to look. If your change log is a ticket category, this is a query against the ticket archive. If it is a spreadsheet, it is a lookup by device name. If it is nothing, this field is the argument for starting one.
The last time this happened and what fixed it
This is a search of the ticket archive for the same dedup key (the device-plus-check string that identifies one incident). It returns the most recent resolved ticket: when it was, how long it lasted, and its closing note. This is the field technicians love most, because it turns "I have no idea" into "last time it was the power supply." It only works if closing notes are written. That is one reason the queue-hygiene rules in the sibling guide on keeping the queue honest insist on them.
A graph link and a dashboard link
Include two deep links. One goes to the monitoring system's view of the specific device and check, showing the time around the failure. The other goes to the dashboard or map for the site or service it belongs to. The first shows what happened; the second shows what else is happening. Both should open the exact view, not a login page followed by a search.
The runbook link
A link to the written procedure for this kind of failure, if one exists. The technician should not have to remember where runbooks live at 3 a.m. What a runbook should contain so that it is actually usable under stress is the subject of runbook anatomy. The ticket's job is simply to point at it.
A suggested first step
One sentence, drawn from the runbook or from the last resolution: "Check whether the closet PDU has power before driving out." This is the hardest field to keep accurate and the one with the highest payoff per word. That is because it converts the whole ticket from information into an instruction.
Where each field comes from
Enrichment is only a series of lookups, and each lookup has a source. Knowing the source tells you two things: how hard the field is to automate, and who to blame when it is wrong.
| Field | Source | Example value | Lookup by |
|---|---|---|---|
| Role, model class | Device inventory (or the monitoring system's discovered device list) | Access switch, stacked | Device name |
| Location, rack/closet | Device inventory | Building B, floor 2, east closet, rack 1 | Device name |
| Owner | Device inventory | Network team (contact: on-call) | Device name |
| Affected service | Service catalogue or a tag in the inventory | Floor 2 east desks; warehouse print | Device name or group |
| Downstream devices | Monitoring dependency map | 12 devices, all unreachable | Device name |
| Upstream device | Monitoring dependency map | core-sw-1 (up) | Device name |
| Recent changes | Change records (ticket archive, change log) | Firmware update on core-sw-1, two days ago | Device, neighbour, site; last N days |
| Last occurrence and fix | Ticket archive | Ticket 3902, three months ago: PSU replaced | Dedup key |
| Graph link | Monitoring system | Device and check view, last 4 hours | Device and check, URL template |
| Dashboard link | Monitoring system | Site map for Building B | Site tag, URL template |
| Runbook link | Documentation store | Access switch unreachable | Check type or device class |
| Suggested first step | Runbook, or the last closing note | Check closet PDU power first | Check type |
Two patterns jump out. Most fields are keyed on the device name. That means the device name in the alert must match the device name in the inventory exactly. The first week of any enrichment project is usually spent discovering that they do not. And the two most valuable fields — last fix and suggested first step — come from your own past tickets and your own runbooks. That means enrichment gets better precisely as your team's discipline gets better. It is a compound-interest feature.
The enriched-ticket template
Lay the ticket out so that the most decision-relevant information is at the top and the reference material is at the bottom. A technician reading on a laptop should get the scope and the first step without scrolling. A template that has worked well in practice:
sw-acc-14 unreachable for 10 minutes
WHAT
Check: ping, 0 of 10 replies since 07:42 (threshold: 3 misses)
Severity: critical -> priority P2 (important tier)
Dedup key: sw-acc-14/unreachable
WHERE AND WHO
Role: access switch (stacked), important tier
Location: Building B, floor 2, east closet, rack 1
Owner: network team
Affects: floor 2 east desks (12), warehouse print server
SCOPE
Upstream: core-sw-1 -- UP
Downstream: 12 devices, 12 unreachable (list: ws-2e-01..ws-2e-11, prn-wh-01)
RECENT CHANGES (last 7 days)
Change 4410: firmware update on core-sw-1, 2 days ago, completed
LAST TIME
Ticket 3902: same key, 3 months ago, 2h 10m. Closing note:
"Power supply failed; replaced from spares. Second PSU on order."
FIRST STEP
Check the east closet PDU for power before dispatching. If the switch has
power and no link lights, see runbook step 3.
LINKS
Graph: https://monitor.example.internal/device/sw-acc-14/ping?range=4h
Dashboard: https://monitor.example.internal/map/building-b
Runbook: https://docs.example.internal/runbooks/access-switch-unreachable
[monitoring] created 07:52 after a 10-minute confirmation delay.
Notice what the template does not include: raw SNMP (Simple Network Management Protocol) output, the full configuration, every metric the device has ever reported. Those are one click away behind the graph link. The ticket is a briefing, not a data dump, and a briefing that runs to three screens stops being read.
A before-and-after walkthrough
Here is the same incident twice, to make the difference concrete. First, the ticket as a bare bridge would create it:
Title: ALERT sw-acc-14 Description: Host unreachable. Severity: critical. Category: (none) Assigned to: (unassigned)
The technician who picks this up must look up what sw-acc-14 is, find it on a floor plan, and guess whether it matters. They must check the monitoring console for scope, ask in the team chat whether anyone changed anything, and search the archive for previous incidents. Twenty minutes, if the inventory is good and the colleague who knows the closet is at their desk. Then they can begin.
Now the enriched version above. The technician reads the scope line and knows it is the switch, not a desk. They read the "last time" block and know the power supply history. They read the first step and, rather than driving to Building B, ask the receptionist there to look at the PDU. That phone call takes two minutes and, in the most common version of this story, ends the incident. In that version, the cleaner unplugged the PDU to run a vacuum. The enriched ticket did not fix the switch. It removed twenty minutes of searching and one unnecessary drive from the path to fixing it, and it did so for every incident, every time.
Automated versus manual: start with three fields
Full enrichment is a project. The good news is that the value is front-loaded: a handful of fields deliver most of the benefit, and they are the easy ones. Start with three.
- Location and owner. A straight lookup from the inventory by device name. This is the field that stops "where is it?" from being the first question.
- The graph link. A URL built from a template with the device and check name substituted in. Almost no logic, and it saves a login-and-search on every ticket.
- The last occurrence and its closing note. A search of the ticket archive by dedup key. Slightly more work, and the field technicians thank you for.
If you have no automation at all yet, the template still earns its keep as a manual checklist. Paste it into the ticket tool as the default description for the monitoring category. Ask whoever picks up an auto-ticket to fill in the blanks during the first five minutes. That sounds like extra work, and it is, but it front-loads the searching into a fixed routine instead of scattering it through the incident. And it produces a set of filled-in tickets that show you exactly which fields were worth the effort. Automate those first.
At Ashgrove Housing the runbooks lived in a folder that everyone knew about and nobody could find at three in the morning. Once the basic three fields were in place, the next one the team automated was the runbook link. They chose it because it took ten minutes: a URL template keyed on check type. The following month's review showed that the time from ticket creation to first note had roughly halved. That was mostly because the technician no longer opened the ticket, closed it, and went looking for the folder. Nobody had planned to measure that. The folder is still there. It has not been opened since.
With the first three in place, add the rest in roughly this order. Start with the affected service (a tag in the inventory), then the runbook link (a lookup by check type). Next comes the dependency picture (if your monitoring system exposes it in notifications or through an API, its interface for other programs). Then add recent changes (a query against your change records), and finally the suggested first step. That last one is the one field that may need to stay manual for a while, curated by whoever maintains the runbooks.
A design choice that makes all of this easier is two-pass enrichment. The first pass runs at creation and includes only what is in the alert plus the cheap lookups, so the ticket exists immediately. A second pass runs a minute or two later. It uses a scheduled job that finds recently created monitoring tickets and fills in the slower fields by appending a note. Because a ticket is persistent, nobody minds that the change-record excerpt arrived three minutes after the ticket did. The bridge itself stays fast and simple. This is one of the concrete advantages a ticket has over an alert: it can be improved after the fact.
Tip: when a lookup fails, write "unknown" or "not found in inventory" in the field rather than leaving it blank or guessing. A blank looks like an oversight; "unknown" is a finding. Over a month, a count of "unknown" values by field is a precise to-do list for whoever maintains the inventory.
Keeping enrichment truthful
Enrichment inherits the accuracy of its sources, and a confidently wrong ticket is worse than a bare one. Picture a technician who drives to the east closet because the ticket said so. They find the switch was moved to the west closet a year ago and learn to distrust every enriched field. Three habits protect you.
First, date-stamp what you can. "Location per inventory, last updated four months ago" tells the reader how much to trust it. Second, let tickets correct the sources: when a technician discovers the inventory is wrong, the closing note should say so. In that case, someone should fix the inventory, not just the ticket. A monthly review of closing notes that contain the word "actually" is a surprisingly effective audit. Third, prefer live sources to copied ones. A location pulled from the inventory at ticket-creation time is at least as fresh as the inventory. A location copied into the monitoring system's device notes a year ago is frozen there. Wherever a field can be looked up rather than stored, look it up.
And accept that some fields will be wrong sometimes. The goal is not a perfect ticket. It is a ticket that is right far more often than a technician's memory, and that says "unknown" when it does not know.
Where to go next
An enriched ticket is the same incident with the first twenty minutes of searching already done, so nobody has to be the search engine. Build it from the inventory, the monitoring system's links and dependency picture, your change records, and your own archive of past tickets. Start with location, a graph link, and the last fix. Add the rest in passes, and keep the sources honest so the ticket stays trusted. A ticket that is confidently wrong once is a ticket that is ignored from then on. Enrichment is also where a ticket bridge starts feeding on its own output. That is because the closing notes written this month become the "last time" field next month. That is why the final guide in Connecting Monitoring to Ticketing and ITSM is about keeping the queue, and its notes, worth reading. If your team wants the same context pushed into a chat channel during incidents, the ChatOps for Monitoring guides cover that path.
Put this into practice with SysPeep Monitor
Most of the fields in this guide are things a monitoring system already knows. SysPeep Monitor auto-discovers devices and interfaces into an inventory and lets you define the dependencies that become a ticket's scope line. It keeps live dashboards, network maps, and reports in a single web console that a ticket's graph and dashboard links can point straight into. It does not write the ticket; it holds the context. Check the product documentation for its notification options, and try the free trial on any Windows machine.
Download Free TrialLearn About SysPeep