DO reports: what Microsoft’s Delivery Optimization report shows, and what it can’t.

The Delivery Optimization tab opens with tiles: total bandwidth savings, the saved bandwidth split between peer-to-peer and Microsoft Connected Cache, device counts, and the total downloaded. Below them, Device Configuration counts devices with and without peering, by download mode; Content Distribution shows volumes by source for each content type; and Efficiency By Group breaks savings down by Group ID, city, country, and ISP.²

How bandwidth savings is worked out

Bandwidth savings is the share of the bytes downloaded that came from peers or Connected Cache:¹

(BytesFromPeers + BytesFromGroupPeers + BytesFromCache) ÷ (BytesFromPeers + BytesFromGroupPeers + BytesFromCDN + BytesFromCache)

Savings at or below 60% shows a warning icon, and under 10% an error icon.¹ The formula has no term for BytesFromIntPeers, although both tables carry that column,³⁴ so bytes from Internet peers don’t count as savings. In Group mode a device looks for LAN peers first, so bytes from a peer on the same LAN count as LAN bytes, not Group bytes.¹

Peering on, peering off

Peering is On for download modes LAN (1), Group (2), and Internet (3), and Off for HTTP only (0), Simple (99), and Bypass (100). Unknown means the download mode is invalid or empty.¹

Content types, and the window

Content Distribution sorts downloads into categories: Quality Updates, Feature Updates, Driver Updates, Office, Edge Updates, Teams Updates, Apps (Store apps), Intune Apps (Intune Win32 apps), and Other. Since 22 July 2025, non-Windows, non-Store software updates such as Visual Studio, Exchange, and SQL Server count under Feature Updates, not Other.¹ How Win32 apps use Delivery Optimization has a guide of its own.

The figures cover the last 28 days, and a device must have done something with Delivery Optimization in that window to be counted at all.¹ UCDOStatus also carries each device’s own 7-day and 28-day figure, BWOptPercent7Days and BWOptPercent28Days.³

• Why a device didn’t peer

  UCDOStatus counts the peers a device connected to, the ones it couldn’t connect to, and those with an unknown relation (PeersSuccessCount, PeersCannotConnectCount, PeersUnknownCount).³

  DoctorDO’s reading No column says why. A blocked TCP 7680, a VPN, a file under the minimum size, and a proxy that breaks the cloud service all look the same in it.

• In the office, or away from it

  Microsoft’s FAQ says the report doesn’t distinguish data by where it was downloaded, such as in the office or out of it. City and country are where a device downloaded the most data,¹ approximated from its IP address.³

• Which office a Group ID is

  Efficiency By Group shows each Group ID as an encoded, case-sensitive SHA256 hash. Microsoft gives a PowerShell sample that hashes a Group ID you know, and the Delivery Optimization verbose log records both values.¹

  DoctorDO’s reading So the report can rank your groups but not name them. How Group IDs decide who peers is on the Learn page.

• What your internet provider’s cache served

  Where an internet provider runs a Connected Cache server, BytesFromCache leaves out the bytes it delivered.¹

• Devices outside its scope

  Windows Server, Surface Hub, and IoT editions aren’t covered, nor are devices that are only Microsoft Entra registered. Microsoft also lists a known issue: no data for Enterprise multi-session devices.⁵

DoctorDO’s reading A low figure tells you where to look. It doesn’t tell you what to change.

Microsoft’s FAQ puts it plainly: the Delivery Optimization PowerShell cmdlets use the cache on the device, while the report is calculated from Delivery Optimization telemetry events.¹ The cmdlets are documented on Monitor Delivery Optimization.⁷

Files also leave the device’s cache: unless you change DOMaxCacheAge, each is kept for three days.⁸

DoctorDO’s reading Expect the two to agree in direction, not to the decimal. A cmdlet run on one device shows its jobs now; the report covers every device over 28 days. Where Internet peers carry real traffic, a device’s peer share runs higher than the report’s savings.

DoctorDO isn’t Microsoft’s report and isn’t part of it. It’s an independent community project for Windows devices managed with Microsoft Intune: a collector on each device reads what Windows records about Delivery Optimization, through the same cmdlets, and sends it to a portal in your own Azure subscription. What it adds to the report:

• Every device’s bytes by their five sources (the HTTP source, Connected Cache, and LAN, Group, and Internet peers), and every site compared, so a low figure has a place.
• Findings that name a cause, such as TLS inspection of the Delivery Optimization service, Wi-Fi client isolation, or a minimum file size that keeps most content off peers, each with what to change and the Microsoft page it rests on.
• What each download was: its content type, the app that asked for it, and the host it came from.

Connect your Windows Update for Business reports workspace, and DoctorDO’s Reporting page sets Microsoft’s figures beside its own: bandwidth savings next to DoctorDO’s delivery efficiency, both device counts, both windows (30 days for DoctorDO, a rolling 28 for Microsoft), Microsoft’s content types with their savings, and security update status. The comparison says that DoctorDO’s efficiency counts Internet peers and Microsoft’s savings doesn’t.

DoctorDO reads the workspace with its own managed identity, given the Log Analytics Data Reader role on that one workspace, in your own tenant. Its queries are fixed in its code, and none of them returns device names, cities, ISPs, or device IDs.

DoctorDO’s reading Close figures confirm each other. A wide gap points at devices without the collector, or at a site the collector doesn’t see.