Teams permanently deletes eligible messages after a SubstrateHolds hold and a timer job run; retention expiry timing typically ranges from 1-7 days.
Forecast calls leave a trail. Someone names the deals, says which ones are committed, and explains why a deal slipped. For a small B2B firm, much of that trail lives in Microsoft Teams: a channel post on the first day of the quarter, a thread that argues about a close date, a reply that says a deal moved to next quarter. When the quarter closes, the test is simple. Can a person who was not on the call find the record and follow it?
Teams is not a filing cabinet, and its messages do not stay put forever. Retention guidance for Teams describes how a message moves toward permanent deletion, and it breaks that path into two stages. Knowing those stages tells you what chat can promise a successor, what it cannot, and where the real forecast record should live.
This post walks through both stages with one example team. Then it shows how to keep the record somewhere a colleague can follow after the quarter is over.
Start with one example message
Our example is a small sales team. Each quarter, the team lead posts a commit list in a Teams channel on day 1. The post names the deals the team expects to close, the rep who owns each one, and a short reason for each call. Nobody treats that post as a record. It is simply how the team talks.
The retention path starts in the same place. On day 1, a user creates a chat or channel message. On day 5, the message is copied to the SubstrateHolds folder, where it can still be searched with eDiscovery tools for a minimum of 7 years from day 1 (the retention period).
Your firm may or may not have a retention policy on its Teams chats. Whether one applies, and what it says, depends on how your Teams setup was configured. Usually your admin is the person who knows. The example is useful because it shows the clock, not because it describes your own chats. Treat the 7-year figure as an illustration of a retention period, not as a promise about your data.
The clock matters because a forecast record is often read long after it was written. A deal that looked safe in week two of the quarter may be the thing a new hire asks about a year later. If the only copy of the reasoning sits in a channel, the question is not only whether the message exists today. It is whether it will still exist when someone needs it.
Stage one: the SubstrateHolds hold
The first stage is a hold. Messages remain in the SubstrateHolds folder for at least 1 day. The timeline sets a minimum time in the holding folder, not a fixed one, so a team should not plan around an exact hour or day.
For the example team, the commit list is a normal channel post on day 1. The day 5 copy starts the holding stage. The holding and deletion timings do not establish whether the post remains visible in the channel. Either way, scrolling back through a channel is not a plan for finding a decision months later.
Notice what the hold does and does not promise. Searchability with eDiscovery tools is different from a colleague opening the channel and finding the commit list where they expected it. A forecast record needs to be easy for that colleague to find and follow.
Stage two: the timer job
The second stage is the timer job. Once a message has sat in SubstrateHolds for at least 1 day and is eligible for deletion, the job permanently deletes it the next time it runs, typically between 1-7 days.
The same logic applies when a retention period ends. If the retention policy has an end date and that date passes, the message is permanently deleted the next time the timer job runs, typically between 1-7 days after expiry.
The gap between expiry and deletion matters when planning what to save. A date on the calendar marks expiry; the deletion timing is a typical window rather than an exact deadline. Save the decision record before relying on that window.
Apply that to the example. Suppose the admin set the 7-year retention period from day 1 described above. The commit list stays searchable with eDiscovery tools for at least that period. After expiry, the deletion step still depends on the timer job. Use the typical window to understand the sequence, rather than treating it as a guaranteed deletion date.
Put the whole sequence side by side for the example.
| Step | Timing in the example | What it means for the commit list |
|---|---|---|
| Post in the channel | Day 1 | Illustrative: the team posts its commit list |
| Copy to SubstrateHolds | Day 5 | Kept and searchable with eDiscovery tools |
| Hold for the retention period | At least 7 years from day 1 | Illustrative retention period for this message |
| Timer job deletes | Typically 1-7 days after expiry | Permanent deletion when the timer job runs |
The team can choose where to keep the forecast decision separately from the channel post.
Nothing in the sequence waits for the end of a deal, the end of a quarter, or the day someone asks a question about it.
Why the two stages make chat a weak record
The retention guidance describes keeping messages for a period and then removing them. Chat was built for conversation. A forecast record has a different job.
A forecast record must answer questions the chat was never designed to answer. What did we commit to on a given date? Which deals moved, and when? Who made the call? A channel holds those answers scattered across posts, replies, edits, and threads. A successor who arrives after the quarter closes has to rebuild the story from fragments. If the timer job has already run, some of those fragments are gone, and the successor may not know what is missing. That last part is the hard one. A gap you can see is easy to ask about. A gap you cannot see just looks like a quiet quarter.
Search does not solve this either. A keyword search for a deal name returns the messages that contain it, not the reasoning that connected them. A successor may find the post that says "commit" and miss the reply that changed it.
There is also a human problem. People write chat messages quickly and casually. A rep who types "probably slips to next quarter" does not expect that sentence to become evidence. A record should capture the decision in plain words, once, in a place built to be read later. The chat thread can stay as the discussion, but the decision belongs somewhere else.
What a successor needs from the record
A successor does not need every message from the quarter. They need a few specific things. They need a dated snapshot of each call, showing the deal list as it stood that day. They need the owner of each deal, so they know who to ask. They need the reason for every move between categories, such as a deal that dropped from commit to best case. And they need to know where the snapshots live, so they do not have to search.
Each of those needs points away from chat. A snapshot is a picture of a moment, which a shared document holds better than a stream of posts. An owner is a field, not a sentence buried in a thread. A reason can be one sentence stored next to the deal. None of this requires new software. It requires a rule the team follows every time.
Order matters as well. A successor reads a forecast the way a story unfolds, so the record should run from the first call to the last, with each change dated next to the call that made it. A list sorted by deal name hides the sequence the successor needs.
The example team can meet all of this without much effort. At the end of each call, the team lead copies the final commit list into a shared sheet, adds the date, and fills in the reason for any change. The Teams post remains the working conversation. The sheet becomes the record. When the quarter closes, the successor opens the sheet and reads the story in order, without needing the channel at all. For a related check, see the forecast handoff guide.
Build the record outside the chat
Start by choosing one place for the forecast record that the whole team can open. A shared spreadsheet works for a small firm, and a CRM works if your team already uses one well. The important part is choosing once and writing down where the record lives, so the next person does not have to guess.
A new person should be able to tell from the file name alone which forecast they are looking at.
- At the end of each call, copy the final commit list into the record with the date, the owner of each deal, and a short reason for every change since the last call.
- Use the Teams post as context, not as the record. A link to the post can help a reader who wants more detail, but the record should make sense without it.
- At quarter end, save a dated copy of the record. The copy shows the forecast as it stood at close, so a later edit does not rewrite history.
- Ask your Microsoft admin, in writing, which retention policy covers your team's chats and whether that policy has an end date. Ask whether the answer is the same for every channel your team uses. A written answer is worth more than a guess from memory.
Asking is cheap. Assuming is expensive, especially when the answer changes what you need to save.
Where the example team could slip
The most likely failure is not a dramatic deletion. It is a quiet gap. The team saves the commit list each week, but the reason for a move lives only in a thread nobody copied. The record shows that a deal dropped from commit to best case and does not say why. A successor reads the sheet, sees the drop, and asks the team lead what happened. If the team lead has left, the answer may be gone too.
A second slip is trusting the channel as a backup. A team may assume that because the post is visible today, it will be visible next year. Visibility today is not a reason to leave the only copy of a decision in chat. Treat the channel as a working surface that can change under you.
A third slip is saving the record only once, at the end. A forecast changes every week. A record that captures only the final state loses the path that led there, and the path is often what a successor needs.
A fourth slip is letting one person hold the reasons. If the explanations live in one manager's direct messages, the record is only as complete as that manager's memory. Spread the job of writing reasons across the people who make the calls, so the knowledge does not sit with a single person.
A quarter-end check that takes minutes
Before each quarter closes, run a short check. Open the saved record and read down the list of deals that moved during the quarter. For each one, confirm the reason is written in the record itself, not only in a Teams thread. If a reason exists only in chat, copy it into the record now, while the thread is still there to read.
Check the saved decisions while the discussion is still available to compare. Quarter close is a useful review point, but it does not guarantee that every message remains available. If a thread is already missing, mark the decision's reason as missing rather than guessing.
Then look at the owners. If a rep who owns a deal has left the firm, make sure the record names a current person who can answer questions about that deal. A record with no owner is harder to follow than one with a clear name next to every line.
Finally, note the date of the check in the record. A short line such as "reviewed at close" alongside that date tells the successor that someone looked, and when. It also builds a habit the team can repeat next quarter, which is what keeps the record useful over time.
The test is what happens next quarter
The two stages in the retention guidance are a reason to move the record, not a reason to panic. Chat keeps the conversation for a while, and the timing ranges describe how the hold and deletion steps play out. The record keeps the decision. Those are different jobs, and a small firm does better when it stops asking one tool to do both.
Go back to the example team. The team lead posts the commit list on day 1. It becomes part of the holding and retention timeline described above. Meanwhile, the sheet holds the dated snapshot, the owners, and the reasons. When the quarter closes, the successor opens the sheet and can follow every move without searching a channel.
If you want a place to start on Monday, open the sheet, enter the current commit list with today's date, and add a reason for each change since the last call. The first entry is the hardest. After that, the habit does most of the work.
That is the test for any forecast record. Can someone who was not on the call read it, understand it, and act on it? If the answer depends on a Teams message surviving a timer job, the record is not finished. If the answer depends on a shared sheet and a short reason beside each change, it is.