Project Online retires on 30 September 2026. That is nineteen days from today. In August I argued that the tool choice falls out of one honest question about what the platform was actually doing for you. That advice stands. What changes at nineteen days is the order of work, because one step cannot be done after 1 October and everything else can.
The step is the archive. Do it first. Decide later.
This piece is procedural. The switch-off has not happened yet, so nobody has a war story about it, including me. That is the argument for doing the archive now rather than becoming the story.
1. What “retired” means on 1 October
Microsoft’s announcement is short on comfort. Its own FAQ answers the question directly: after Project Online is retired, “you will no longer be able to access your projects or any associated data within the service,” and it is “essential that you back up your data/projects and complete your transition” before the date. No read-only period is offered, and no path is described for getting the data back afterwards. End of sale for new customers was 1 October 2025, so the runway has been public for a year.
Unaffected, per the same notice: Project desktop, Project Server Subscription Edition, and Planner in either tier. The Project Web App sites, the enterprise resource pool, timesheets, workflows and the OData reporting feed under your Power BI reports are what go dark.
Working assumption for planning: anything still inside PWA on the evening of 30 September is gone.
2. Four things that must survive, in this order
Rank by consequence of loss, not by how much of it there is.
- Records with a retention obligation. Timesheet history that fed capitalisation or client billing. Approvals. Decision logs attached to projects. If an auditor, a tax authority or a client can ask for it, it goes out first, and the retention period is whatever your finance policy says it is, not what the tool defaulted to.
- Closed-project baselines and final schedules. The audit trail for anything that might be revisited: a contractual dispute, a regulatory review, a post-implementation benefits check.
- The reporting dataset, not the reports. Every Power BI and Excel report bound to the OData feed dies with the feed. The report file is worthless without the tables under it. Pull the tables: projects, tasks, assignments, resources, timesheet lines, custom fields. Then the report can be rebuilt against a file.
- Project site content. Documents, risk and issue lists, and deliverables stored on each project’s SharePoint site. These often outlive the schedule that spawned them.
Anything not in those four is a candidate for deliberate loss. Write down what you are choosing not to keep and why. That sentence is the cheapest insurance against the question in March.
3. Formats that open without the platform
The test for every archived item is whether someone can open it on 1 October with tools you will still have.
- Schedules: open each project from PWA in Project desktop and save it as
an
.mppfile. Also export a PDF of the baseline view for the closed projects. A PDF is dumb and permanent, which is what a baseline should be. - Tabular data: the OData feed can be read into Excel or Power BI and saved out as CSV or Excel. One file per table, dated. Include the custom field values; they are usually where the report logic lived.
- Timesheets and approvals: same feed, same treatment, plus a PDF snapshot of the approval state where a payment depended on it.
- Site content: standard SharePoint export for documents and lists. Keep the folder structure; the structure is the index.
Not acceptable: a proprietary backup blob nobody can read, or “it is in the migration tool’s staging area.” Staging is not an archive.
4. Someone opens it
An archive nobody has opened is a hope. Pick a person who did not build it and have them open, from the archive alone:
- one timesheet record from 2023 with its approval,
- one baseline from a closed project,
- one report re-run against the exported dataset for a past month-end.
Record who opened what and when. That record is the evidence that the archive exists. Without it you have a folder.
5. The bridge, if the decision is not made
If you have not chosen a destination, nineteen days is not enough to choose one well and migrate to it. Choose a bridge instead and let the real decision run in October at a normal pace.
A bridge carries five things for one quarter: active projects and owners, milestones where a date has been promised to someone outside the team, open decisions with owners and ages, the resource commitments that are contractual, and the month-end report rebuilt on the exported dataset. Planner basic or a disciplined spreadsheet register does this. It is not a PPM platform and does not need to be. It needs to be the system of record from 1 October, declared as such, so that nobody keeps a private copy of the truth.
Put a real schedule only where a date is promised to an outsider. Everything else is a list.
6. The nineteen-day calendar
Days 1 to 3: inventory and owners. List every project, every custom field in use, every report bound to the feed, every site with content. Name one owner per item in section 2. Decide what is deliberately not kept.
Days 4 to 10: export. Retention records first, then baselines, then the dataset tables, then site content. Date every file. Keep an index as you go; do not plan to write it afterwards.
Days 11 to 14: rebuild the month-end report on the exported dataset and run it for August. If it does not produce August, it will not produce October.
Days 15 to 17: open-test per section 4. Fix what does not open.
Day 18: freeze. No new entries in PWA. The bridge is live and declared.
Day 19: buffer. Hold it as slack. Something in the export will have failed silently, and this is the day you find it.
1 October, morning. The bridge register is the system of record. The archive index is in the governance record. Anyone who asks where something is gets a path, not a shrug.
7. What I would write down
Two lines, in the governance record, dated:
Project Online archive completed on date: retention records, closed-project baselines, reporting dataset, site content. Opened and verified by role on date. Items deliberately not retained: list.
From 1 October the system of record is bridge. The PPM destination decision is due date, and the August analysis stands.
If the archive is done and opened, a missed migration is an inconvenience. If it is not, a missed migration is a records problem with your name on it. The nineteen days are for the first kind.
Source: “Microsoft Project Online is retiring: What you need to know,” Microsoft Tech Community, Planner Blog, 5 September 2025, including its FAQ. Quotations above are from that post, verified on 11 September 2026. This article describes how to protect records; it is not advice on your retention obligations, which your finance and legal policies define.