Table of contents
Research context
This protocol measures how often a remote contractor laptop fails a required device posture check. The unit is one device check for one contractor on one scheduled review date. A failure means a blocking setting is off or unreadable at the check time. The study describes the recorded posture, not the security of the person using the device.
Define the baseline from the client policy, including identity enrollment, disk encryption, supported operating system, screen lock, and the required endpoint agent. Record the policy version and the date it became effective. If the baseline changes during the window, split the sample at that date. A new check item changes the denominator.
Define a pass as a recorded value that meets the baseline for every blocking item. Define a failure as a blocking item that is off, and define unknown as a blocking item the tool could not read. Keep failure and unknown separate in the report. The two states need different responses.
Set the scope to one client, one contractor group, and one inventory source. Record the device models, the operating systems, and the check tool version. If a new tool is introduced during the window, analyze the earlier and later cohorts separately. A new tool can change what is visible.
Set the population as every scheduled device check in the window, including devices that were offline or replaced. The unit is the check, not the contractor, because one person can have more than one device. A replacement device creates a new unit linked to the same contractor. Test the definition on two devices before collection begins.
Record the device identifier, the contractor group, the check time, the baseline version, the result for each blocking item, the repair action, and the retest time. Add the person who verified the repair. Keep the check linked to the inventory record and the policy. A second reviewer should be able to reconstruct the result.
Practical application
Separate a failure found before access was granted from one found during an existing session. The first is a gate, while the second is a detection event. Report the two groups separately. Combining them hides whether the control stops access or only observes it.
Check the devices that were never checked because they were offline or not enrolled. Record them as unknown rather than excluding them. A device outside the inventory is a different risk from a device that failed a setting. Report the count with the reason.
Report the number of checks, the pass rate, the failure rate, and the unknown rate. Show the counts by check item, device model, and contractor group. Present the median and longest repair time. Use the same denominator for every rate and show the checks with incomplete records.
Separate a failure caused by the device from one caused by the inventory tool. A tool that cannot read an encrypted disk is different from a disk that is not encrypted. Report the two categories separately so the client can choose the right fix. A new tool will not solve a policy that the device cannot meet.
Inspect the failures that repeat on the same device or model. A repeated encryption failure may point to a provisioning image, while a repeated update failure may point to a maintenance window. Compare the pattern with the repair path. Report the device models without naming the contractors.
Operating controls
Do not treat a pass as proof that the device stayed compliant after the check. A contractor can change a setting or install an unapproved tool later. Report the time since the last check and the sampling rule. The study describes a point-in-time result, not continuous compliance.
The record cannot show whether a failed device was used for production work before the check or whether the contractor intended to bypass the rule. It can show only what the inventory reported at the check time. A failure is not proof of a breach, and a pass is not proof of safe use. The protocol also cannot capture a personal device that was never enrolled.
A useful additional view is the share of failures that were repaired within the stated window. A fast repair may indicate a healthy support path, while a slow repair may indicate a missing administrator. Record the repair time and the owner. That field turns the failure count into a service measure.
Give the result to the client security owner and the offshore lead with one observable change. The change might be a pre-access device check, a standard provisioning image, a named repair administrator, or a quarterly manual sample. Record the owner, the effective date, and the measure that will show whether the change worked. Repeat the protocol with the same definitions after the change.
Close the review with one change and a date to re-measure. The change should target the check item with the most failures or the longest repair time, not the item that is easiest to automate. Record the expected effect and the evidence that will prove it. A device baseline that is never sampled is a claim, not a control.
Sources
Frequently asked questions
What should happen first?
Begin with a bounded ticket, approved access, explicit acceptance evidence, and a named reviewer.
Who approves production changes?
The client's technical owner keeps final merge and release authority.
Related research
What share of outsourced postmortems close their action items?
