Best practices and common mistakes

Best practices

• Start targeted, survey with discovery. Keep the fleet in targeted mode as steady state; use discovery as a scheduled survey (e.g. quarterly), not a permanent posture.
• Enroll with tokens, drive with central config. Ship agents with an enrollment token (or use the preconfigured Windows wizard) so each device gets its own revocable credential, and keep use_central_config = true. It is the only credential a Scout accepts from v0.6.0 anyway. Manage what they track centrally in Groups/Rulesets and the collection policy, never by editing TOML per machine.
• Migrate legacy configurations before you upgrade them. A Scout still configured with api_key, username or password will not start once it reaches v0.6.x. Sweep Inventory → Agents for workstations that never enrolled and re-deploy those first; agents on v0.5.0 and earlier keep working in the meantime, so this is a queue to work through, not a fire.
• Name groups after org units. Marketing, Finance, Servers — not Adobe-users. Membership should survive tooling changes.
• Keep security-relevant rulesets on every group. Apply Remote access and Security tools universally; those categories are risk-relevant regardless of department.
• Run workstation agents elevated and interactive. Elevation unlocks bandwidth; an interactive console session unlocks foreground focus (level 3). A highest-privilege logon task gets both.
• Track adoption over time, not just a snapshot. Pick a window that fits the question (30 days for recent behaviour, 6–12 months for seasonal tools) and watch the trend column — a tool that is fading is as actionable as one that is heavily used.
• Hunt installed-but-unused tools. The "Installed, unused" filter is your fastest reclamation list — tools present on devices that nobody actually opened in the window.
• Re-survey to catch drift. Adoption changes; re-run discovery periodically and re-check the trends so your view of what is used stays honest.
• Close the Review queue. Discovery only pays off if someone classifies what it finds. Treat the queue as a work queue with an owner, not a backlog.
• Stand up a Software Estate dashboard early. Give stakeholders a working usage/adoption/trends view; evidence beats anecdote in every tooling conversation.

Common mistakes

Mistake	Consequence	Fix
Leaving a large fleet in discovery permanently	High report volume; noisy, un-curated estate	Survey, curate, return to targeted
Expecting bandwidth on unelevated Windows	Bandwidth reads 0	Run elevated (the installed service is) — and remember bandwidth is a TCP-only lower bound
Running the agent in a session-0 / headless context	No foreground → never level 3; usage understated	Use an interactive, highest-privilege logon task
Using the Machine Agent connection key for inventory auth	The agent can't authenticate inventory	Give the inventory module its own enrollment token (on the Machine Agent an ingest api_key also works)
Writing patterns too broadly (e.g. office)	False-positive product matches	Prefer specific substrings; edit and trim the Default rulesets
Reading a single day as the whole story	A quiet day looks like an unused tool	Judge adoption over a window (30/90 days) and the trend, not one snapshot
Never re-surveying after the first rollout	Adoption drifts and the estate goes stale	Re-run discovery and re-check trends periodically

A discover → curate → target rollout (worked example)

A newly acquired office of 120 workstations comes online; nobody knows what is installed.

1. All 120 agents enroll into a new NewOffice group placed in discovery for two weeks.
2. New titles fill the Review queue; IT classifies products, sets vendors, marks paid/managed, and assigns policy.
3. The keepers' patterns are promoted into rulesets (a Design suite here, Developer tools there).
4. Those rulesets are attached to the group, and the group is switched to targeted.

In two weeks the office went from a black box to a curated, targeted, low-volume estate — and the same loop is scheduled to re-run quarterly to catch drift.

Tip: Adoption in practice: Camtasia showed 31 devices with it installed but only 6 actually using it across 90 days, and a declining trend. The "Installed, unused" filter named the 25 idle devices to reassign or clean up — a defensible call backed by measured usage rather than a hunch.