Systems and records
IT asset identity reconciliation: one device's history across its IDs
By SourceX Editorial · Updated
10 sections
Short answer
IT asset identity reconciliation links every identifier one device has carried, such as serial number, firmware UUID, management and directory device IDs, hostname and asset tag, into one dated history. Anchor on hardware identifiers, treat names and addresses as time-bound aliases, give each attribute one authoritative source, and log every merge with its evidence so a second reviewer can rebuild the same timeline without live secrets.
Record timeline
IT asset identity
Key takeaways
- Anchor each device on its serial number and a real firmware UUID; every software-assigned ID is a dated alias.
- Reimages, re-enrollments and renames routinely give one laptop several directory and management records.
- Give each attribute one authoritative source and one fallback, and write down why.
- Weak signals such as hostname and MAC address may confirm a match but never create one.
- Remove recovery keys, local admin passwords and pasted secrets at the source, before any export.
How do you reconstruct an IT asset history across several identifiers?
Pick one physical device as the entity, anchor it on its hardware serial number and firmware UUID, and treat every other identifier as an alias that was true for a dated interval. Hostnames, directory objects, management enrollments and asset tags all get attached to the device with a start date, an end date and the record that proves the link.
The reason one laptop ends up with three or more IDs is ordinary operations, not bad data entry. Microsoft's Entra ID guidance lists duplicate device objects after an OS reinstall, reimage or recovery as a normal source of stale entries, so a single rebuild can split one machine into two directory records. A re-enrollment in the management tool can produce a new managed device record with its own ID, and a rename gives the help desk a new hostname to type into tickets.
IT asset identity reconciliation is the work of merging those aliases back into one timeline and deciding, attribute by attribute, which system gets to say what is true. The output is a device history a second reviewer can rebuild from the same exports, with every merge and every refusal to merge written down.
Which identifiers are stable enough to anchor a device?
Hardware identifiers anchor; everything assigned by software is an alias. The table ranks the identifiers you will usually find, where each one comes from and the event that breaks it.
Two details matter in practice. The SMBIOS specification says an all-FFh UUID means the ID is not currently present but can be set, and all-00h means it is not present, so those values must never be used to join records. And the Microsoft Graph managed device resource carries both its own id and the azureADDeviceId of the directory device, which gives you a documented bridge between the management record and the directory record without guessing.
| Identifier | Typical source | Survives | Breaks when | Role in the history |
|---|---|---|---|---|
| Hardware serial number | SMBIOS System Information, read by osquery as hardware_serial or by the MDM as serialNumber | Reimage, rename, re-enrollment, user change | Board replacement on some models, blank or placeholder values | Primary anchor, paired with manufacturer |
| SMBIOS system UUID | Firmware, read by osquery system_info.uuid | Reimage, rename, re-enrollment | All-FF or all-00 values, board replacement, cloned virtual machines | Secondary anchor, only when the value is real |
| MDM managed device ID | Intune managedDevice id | Rename, user change | Re-enrollment or device wipe and re-registration | Alias with enrollment dates |
| Directory device ID | Entra device object, exposed on the MDM record as azureADDeviceId | Rename | Reinstall, reimage, recovery, change of join type | Alias, bridge to sign-in evidence |
| Hostname | Directory, MDM, endpoint agent, ticket text | Little; it is meant to change | Rename, reimage with a new naming standard | Time-bound alias for linking tickets and alerts |
| MAC address | Network interface, Wi-Fi and Ethernet fields on the MDM record | Reimage | Docks, USB adapters, randomized private addresses, NIC swaps | Supporting evidence only |
| Asset tag or CMDB record ID | Asset register or CMDB | Most operational events | Retagging, duplicate CMDB records from a second import | Business key for cost center, owner and disposal |
| IMEI | MDM record for phones and cellular tablets | Reimage, re-enrollment | Device swap under a support exchange | Primary anchor for mobile devices |
Which system is the authority for each attribute?
Give every attribute exactly one authoritative source and one fallback, and write the reason next to it. Without that rule, the last import wins, and the last import is usually the least careful one.
Activity dates need the most care. Entra ID only replaces a device's activity timestamp when the new value is more than 14 days (plus or minus 5 days) later than the stored one, and Microsoft states that the attribute is not an audit and points to sign-in logs for finer detail. Because of that rule, the stored timestamp can trail the device's real last activity by up to about 19 days (14 days plus the 5-day variance), so a history that treats it as a precise last-seen date will be wrong. Use it to bound an interval, and use sign-in or agent check-in records when a decision depends on the exact day.
| Attribute | Authoritative source | Fallback | Why |
|---|---|---|---|
| Serial number, model, manufacturer | Endpoint agent or firmware read | MDM hardware inventory | Read from the device itself, not typed by a person |
| Ownership (company or personal) | MDM ownership field | Purchase record in the asset register | Decides whether the device belongs in scope at all |
| Assigned user | Asset register assignment with dates | MDM primary user | Assignment history needs start and end dates the MDM field may not keep |
| Hostname at a point in time | Directory and agent history | Ticket and alert text | Tickets quote the name that was valid when they were opened |
| Last activity | Sign-in or agent check-in logs | Entra activity timestamp | The timestamp is updated in steps of about two weeks and is not an audit |
| Compliance and OS version | MDM compliance record | Endpoint agent | Policy state is defined by the management tool |
| Lifecycle state (in use, repair, retired, disposed) | Asset register | Disposal certificate or vendor RMA record | Finance and disposal records outlast the device objects |
| Cost center and department | Asset register or ERP | Directory department of the assigned user | Ownership for reporting sits with finance, not IT tools |
What should the entity resolution and source authority task brief include?
The task brief is the document you hand to the person or model doing the reconciliation, so it has to state the inputs, the rules and the expected output precisely enough that two people would produce the same timeline. The template below is the asset; copy its rows and fill the middle column for your own estate.
| Brief section | What it contains | Illustrative entry | Reviewer check |
|---|---|---|---|
| Task statement | The entity, the time window and the question the history must answer | One timeline per company-owned laptop and desktop, 2021 to 2025 | Is the entity physical hardware, not a user or a login? |
| Source inventory | Every export with system name, export date, row count and field list | MDM export, directory export, agent export, CMDB export, ticket export | Does each source have an export date and an owner? |
| Identifier crosswalk | Which field in each source maps to serial, UUID, device IDs, hostname, tag | MDM serialNumber maps to agent hardware_serial | Are normalization steps (case, whitespace, leading zeros) written down? |
| Match rules in order | Ordered rules with the evidence each one requires | Rule 1 exact serial plus manufacturer; rule 2 MDM azureADDeviceId to directory device ID | Can a rule fire on a placeholder or sentinel value? |
| Authority matrix | One authoritative source and one fallback per attribute | Assigned user from asset register, fallback MDM primary user | Does any attribute have two authorities? |
| Conflict log | Every disagreement between sources and how it was settled | Two serials for one hostname in the same week | Is each settlement tied to a rule or a named reviewer? |
| Merge and split decisions | Records joined or kept apart, with evidence IDs | Directory objects A and B merged after a reimage ticket | Can the decision be reversed from the log alone? |
| Exclusions | Fields and records removed and why | Recovery keys, local admin passwords, personal devices | Were secrets removed before export, not after? |
| Output schema | Device timeline rows with interval, alias, source and evidence | device_key, alias_type, alias_value, valid_from, valid_to, evidence_ref | Does every row carry an evidence reference? |
| Acceptance checks | Tests the finished history must pass | No device holds two active hostnames on the same day | Were the checks run and their results kept? |
How should match rules be ordered, and when should records stay apart?
Order the rules from the identifier that is hardest to fake or reuse down to the weakest supporting signal, and stop at the first rule that fires. A weaker rule should only ever confirm a match a stronger rule proposed, never create one on its own.
At a glance
How should match rules be ordered, and when should records stay apart
Splitting is the decision teams skip. Two records that share a hostname a year apart are usually two devices that inherited a naming slot, and merging them would hand one laptop another laptop's repair history. Keep records apart when the hardware anchors disagree, even if every software alias agrees.
- Rule 1: exact match on normalized serial number plus manufacturer, ignoring blank and placeholder values.
- Rule 2: exact match on the firmware UUID, only when it is not all-FF or all-00 and is not shared by more than one serial.
- Rule 3: management record to directory record through the documented device ID field, such as azureADDeviceId on an Intune managed device.
- Rule 4: IMEI for phones and cellular tablets.
- Supporting only: hostname within a dated validity window, MAC address, assigned user. These confirm or question a match; they never create one.
- Hold for review: one serial with two models, one UUID across several serials, or a device that appears active in two places on the same day.
Worked example: one laptop, three IDs
This example is illustrative and invented; the dates, names and counts describe a typical pattern, not a real estate. One laptop is issued, renamed, reimaged after a disk fault and reassigned, and the four exports disagree in exactly the ways the rules above are built for.
| When | Source | What it shows | Decision |
|---|---|---|---|
| Month 1 | Asset register and MDM | Tag A-1042, serial S1, managed device ID M1, hostname FIN-LT-07 | Create device key D1 anchored on serial S1 |
| Month 9 | Directory and ticket export | Rename to FIN-LT-22; tickets keep quoting the old name for 2 weeks | Add hostname alias with an overlap window; link tickets by date |
| Month 14 | Ticket export | Disk fault, reimage approved under a change record | Record the change ID as evidence for the next two rows |
| Month 14 | Directory export | New directory device object; the old one stops showing activity | Merge both directory objects into D1 on rule 3 and the change record |
| Month 14 | MDM export | Re-enrollment creates managed device ID M2, same serial S1 | Merge M2 into D1 on rule 1; close M1 interval |
| Month 20 | Asset register | Reassigned to a new user in another department | Close the first assignment interval; open the second with the register date |
| Month 31 | CMDB import | A second CMDB record for tag A-1042 from a late import | Merge on tag and serial; log the duplicate as a source defect |
| Exception | Agent export | A different laptop later reuses hostname FIN-LT-07 | Keep apart: serial differs, so the hostname alias is time-bound to D1 only until month 9 |
Why a closed ticket rarely explains a device's final state
A closed ticket says work happened; it rarely says why the device ended up in its final state. The decision lives in the linked records around the ticket, so the history should carry references to them rather than only the ticket status.
At a glance
Why a closed ticket rarely explains a device's final state
Link each ticket, alert and change to the device through the hostname or user that was valid on that date, not the current one. That single rule is what makes the reconstructed history useful: a disk replacement in month 14 belongs to D1 even though the ticket names a hostname D1 no longer uses.
- Incident or request ID, opened and resolved dates, and the hostname quoted in the ticket.
- Change or approval record for reimages, hardware swaps and disposals.
- Alert or detection ID when the work was triggered by monitoring, without the raw payload.
- Runbook or procedure version that was followed, so the steps can be compared over time.
- Vendor repair or return record when parts or the whole unit were exchanged.
- A bounded outcome: back in service, replaced, retired or disposed, with the record that shows it.
What must come out before a device history leaves the IT team?
Remove credentials and recovery material at the source, before any export leaves the tools, because device records sit next to them. Microsoft documents that BitLocker recovery keys for encrypted Windows devices can be stored in Entra ID and viewed from the device details, and that Entra can store and rotate local administrator passwords through LAPS. Neither belongs in a reconciliation file.
Ticket text is the second leak. People paste passwords, Wi-Fi keys and tokens into tickets, so scan free text for secrets and replace user names with stable pseudonyms that keep the assignment intervals intact. Internal IP plans and network diagrams that would map the environment should be generalized or dropped; the history needs the device and its events, not the attack surface.
- Drop: recovery keys, local admin passwords, certificates, tokens, Wi-Fi pre-shared keys.
- Pseudonymize: user names, email addresses and personal phone numbers, with one key per person.
- Generalize: internal IP addresses, building and desk locations below site level.
- Keep: hardware anchors, alias intervals, event dates, change and incident IDs, outcomes.
Which devices fall outside a defensible reconstruction?
Stop and narrow the scope when the record set includes devices the company does not own, devices tied to an open investigation, or gaps you cannot explain. The management record's ownership field, which the Graph resource lists as company, personal or unknown, is the first filter.
Personal devices enrolled under a bring-your-own-device policy carry the employee's own data and expectations; exclude them unless a privacy reviewer approves a narrower field set. Devices on legal hold or in a security incident file go to whoever controls the hold. Managed service providers should treat client devices as the client's records, not their own. And if a device's history has an unexplained gap longer than your stale-device window, mark it as a gap rather than bridging it with a guess.
Two external benchmarks help decide whether the source data is fit to start. CIS Safeguard 1.1 asks for an inventory with the network address if static, the hardware address, machine name, asset owner, department and network approval status, reviewed at least twice a year. NIST's NCCoE practice guide SP 1800-5 shows how commercially available tools can track the location and configuration of networked devices across an enterprise. If your estate falls far short of either, fix the inventory first and reconstruct history second.
When a reconciled device history becomes a licensable record set
A clean device timeline with linked tickets, changes and outcomes is the kind of operational record AI developers use to train and evaluate IT operations agents. Having access to the exports is not the same as having the right to license them: vendor terms, employee notices and client contracts each need their own check, and de-identification does not create a licensing right on its own.
An IT team's reconciled histories can be licensed through SourceX, which handles qualification, rights review, contracting, preparation and delivery. It begins with a short company assessment that needs no upload of device records, and nothing ships until the company has approved the scope and terms and signed the agreement.
Frequently asked questions
What is IT asset identity reconciliation?
It is the process of deciding which records in different IT systems describe the same physical device and merging them into one dated history. Each identifier, such as a hostname or directory device ID, is kept as an alias with the interval when it was valid. The result shows one device's life across renames, reimages and reassignments.
Is the serial number enough to identify a device?
Usually it is the best anchor, but not on its own. Blank, placeholder or duplicated serials occur, and some board replacements change hardware identifiers. Pair the serial with the manufacturer, confirm with the firmware UUID when it holds a real value, and hold any conflict for review.
Why does a reimaged laptop show up twice in Entra ID?
Reinstalling, reimaging or recovering a device can create a new directory device object while the old one stops reporting activity. Microsoft lists this as a common cause of duplicate and stale entries. Merge the two objects on hardware evidence and the change record rather than deleting the old one before you have read its history.
Should MAC addresses be used to match devices?
Only as supporting evidence. Docking stations and USB network adapters move between machines, and many operating systems randomize Wi-Fi addresses on some networks. A MAC address can confirm a match that the serial or UUID proposed, but it should never create one.
How often should the asset inventory be reviewed before reconstructing history?
CIS Safeguard 1.1 calls for reviewing and updating the enterprise asset inventory at least twice a year, or more often. If your inventory has not been reviewed recently, refresh it before you start, because stale owner and department fields will flow straight into every reconstructed timeline.
Sources
- Microsoft Entra ID documentation explains that duplicate device objects can follow an OS reinstall, reimage or recovery, and that the activity timestamp is only replaced after a delta of more than 14 days (plus or minus 5) and is not an audit. Source
- The Microsoft Graph Intune managedDevice resource lists id, serialNumber, azureADDeviceId, imei, wiFiMacAddress, ethernetMacAddress and managedDeviceOwnerType (unknown, company, personal). Source
- The osquery system_info table, described as system information for identification, exposes uuid and hardware_serial columns. Source
- The DMTF SMBIOS specification defines the System Information structure with a serial number and a 16-byte UUID; all-FFh means the ID is not present but can be set and all-00h means it is not present. Source
- Microsoft Entra ID can store BitLocker recovery keys for encrypted Windows devices, viewable from device details, and can store and rotate local administrator passwords through LAPS. Source
- CIS Safeguard 1.1 asks for an enterprise asset inventory recording network address if static, hardware address, machine name, asset owner, department and network approval, reviewed bi-annually or more often. Source
- NIST SP 1800-5, an NCCoE practice guide, demonstrates commercially available tools that track the location and configuration of networked devices and software across an enterprise. Source
See if your company qualifies
A short company assessment. No data uploads are needed.