Best Practices
This page sets out a sensible order for standing the platform up, and the habits worth keeping once it is running.
You can do any of this in any order, but doing it in this order avoids most of the rework.
Setting up for the first time
- Add the customer first. Every device, ticket and discovery is filed under a customer, so create the customer before you enrol anything.
- Enrol one test device before the rest. Enrol a machine you control, confirm it appears and reports in, and confirm you can reach it, before you roll enrolment out to a customer’s estate.
- Decide the device groups before the devices arrive. Groups are what policies and actions are applied to. Settling on a structure early — by site, by role, by operating system, whatever matches how you work — saves moving devices later.
- Apply a baseline policy to the test device. Start from a predefined profile rather than a blank one, apply it to the test device alone, and confirm it behaves as you expect.
- Add monitors and alerts to that baseline. Begin with the few conditions you would always want to hear about — a disk nearly full, a device offline, the security client not reporting.
- Only then roll out. Apply the baseline to a group, watch it for a few days, and widen from there.
Working with procedures
- Clone a predefined procedure rather than writing one from scratch where one nearly fits. Predefined procedures cannot be edited, so the clone is the working copy.
- Run a new procedure against a single device before you run it against a group, and read its execution log afterwards.
- Have procedures approved deliberately. A procedure you have just edited drops back to the Edited status and will not run until it is reviewed again.
- Prefer a schedule in a policy over repeated manual runs for anything you do more than once.
Working with monitors and alerts
- Give a condition a sensible duration. A processor condition that fires on a one-second spike will bury the real alerts.
- Keep a device-status condition in a monitor of its own: the platform says plainly that it should not be mixed with other conditions, and that auto remediation does not work with it.
- Point auto remediation at a procedure you have already tested by hand.
- Check the number of policies using a monitor before you change it, and before you delete it.
Working with patches
- Decide deliberately whether newly published patches are approved without review, and set the automatic-approval toggle to match that decision.
- Install to a pilot group before the whole estate, especially for anything whose reboot column is not a plain No.
- Use a patch procedure on a schedule inside a maintenance window rather than installing on demand during working hours.
Maintenance windows
- Define a window for every profile that carries disruptive tasks, and set the eight on-demand task rules in it deliberately — warn where you want a prompt, forbid where you never want it to happen.
- Read the maintenance-window counts in the Run Procedure dialog before running anything broad; they tell you how many devices are blocked and how many will merely warn.
Day to day
- Filter the console to one customer when you are working inside that customer, so an action cannot reach the wrong estate.
- Use the saved filters on each list rather than scrolling. Every list in the console has a filter panel behind the funnel icon.
- Work from the dashboard each morning: the Ticket Health cards show what is overdue, unassigned and expiring, and the AI Insights panel explains what changed.
Note: the single habit that prevents the most damage: when a control acts on live endpoints, confirm which devices your selection actually resolves to before you click. Several actions in the console submit immediately, without asking you to confirm.
