Slack team health # See Who Went Quiet in Slack, By Name. TeamPredict counts how many messages each person posts per week, compares it to their own usual level, and shows the drop next to their name. Message text is never read. You tell your team and confirm you have done it, with no opt-out for them, and only then is up to the last 12 months of their counts read. [Start 30-Day Free Trial](https://app.teampredict.ai/signup)[See how it works](https://www.teampredict.ai/slack-team-health#how-it-works) No charge today · Cancel anytime Slack team healthPer person · sample - Priya R.EngineeringDeclining 34% below their usual level, two weeks running - Marcus T.SalesSteady In line with their 8-week baseline - Dana K.SupportWatch 17% below their usual level, one week so far Disclosure confirmed by the employerRecorded, not verified Their admin confirmed the team was told. Nothing is read, past or present, before that. TeamPredict messages nobody. number per person per week 1 months of history, read after you confirm 12 months, then counts are deleted 18 opt-out for the people counted No What is measured ## Volume Only. Nothing About What Was Said. TeamPredict stores one integer per person per week and builds everything from it. Here is exactly what that number carries and what it does not. Recorded ### A weekly total, attached to a person - One number per person per week: how many messages they posted in the selected channels - Weeks going back up to 12 months, read only after your admin confirms your team was told - Their name, and their department when it matches a person on your roster - Their usual level, taken as the median of their trailing 8 weeks - How far below that level they are now, and for how many weeks running Not recorded ### Nothing that says what anyone was doing - The text of any message, thread, or reaction, and no permalink back to one - Which channel a message was posted in - Who a message was addressed to or who replied to it - The time of day, so there is no picture of anyone’s working hours Who sees it ### Your admins, by name. Each counted person’s weekly count, usual level and trend is shown to admins in your TeamPredict organization, next to their name, including any admin who was not part of the decision to switch this on. The people being counted are shown nothing by TeamPredict. They get no message from us, no account here, and no notification of any kind, so what they know about this they know because you told them. How far back it goes ### You confirm first. Then the last 12 months are read. Counting does not begin on the day you switch this on. Once an admin has confirmed your workforce was told, TeamPredict reads the selected channels backwards and turns up to the last 12 months of their messages into the same weekly counts, which is why there is a baseline to judge this week against once that read finishes rather than a month and a half later. It is not instant: the read runs in the background and works through your channels in batches, so a workspace with a lot of them takes several passes before its history is complete. Nothing about anyone is read before that confirmation is on record. That order is the whole argument for doing it at all: disclosure first and then history is measurement people were told about, the other way round is a year of quiet monitoring with a note attached afterwards. Whether the disclosure actually happened is on you. We record the confirmation, we do not check it, and we have no way to check it. How much of that year exists is not up to us. A free Slack workspace serves roughly 90 days of channel history whatever is asked of it, paid plans serve more, and there is no way to know which you have until the read is finished. TeamPredict records the oldest week it actually reached and shows you that number, so three months of history reads as three months. The weeks Slack never served stay empty rather than being filled with zeros. How it works ## Four Steps, and One of Them Is You Telling People. Slack team health is off until you turn it on. Nothing about a person is read, this week or last year, until an admin confirms your workforce has been told. ### Connect Slack A workspace admin connects Slack. Public channels come through the app’s own token. Private channels need a second authorization with that admin’s personal Slack account, which reaches only the rooms they are already in. ### Pick the channels Your admin chooses the counted channels one at a time. A newly found internal channel arrives already ticked, so anything you do not want counted has to be switched off. Channels shared with another company arrive unticked. ### You tell your team, then confirm it TeamPredict messages nobody and cannot. You disclose the measurement through your own channels, then an admin ticks a box confirming it was done. We record who ticked it, when, and the exact wording. We do not verify it. ### Then their history is read Nothing is touched before that confirmation is on record. After it, TeamPredict reads the selected channels backwards, up to the last 12 months, and folds it into the same weekly counts. Their usual level is the median of their trailing 8 weeks. After the history lands, TeamPredict keeps counting each completed week. How far back the history really goes depends on your Slack plan: a free workspace serves roughly 90 days of channel history whatever is asked of it, and the product reports the oldest week it actually reached rather than the year it asked for. A person is flagged only when they sit at or below 30% under their own baseline for 2 consecutive weeks. A single quiet week is a holiday, a launch, or a conference, not a signal. What it never touches ## Five Limits That Do Not Have a Setting. Each one is a permission the product does not request or a field it does not store. There is no plan, no toggle, and no support request that turns any of them on. ### It never reads message text Not the message, not the thread, not the reactions, not a permalink. One integer per person per week is stored, and nothing else. ### It never reads or sends direct messages TeamPredict holds no DM permission at all, and no permission to post anywhere in your workspace. It cannot read a DM and it cannot write one. ### It never uses Slack’s Discovery API The export API that can pull a whole workspace, including rooms nobody in it has joined, is not requested and not used. ### It never records where or when A count does not carry the channel it came from, the person it was aimed at, or the hour it happened. It is a weekly total and that is all. ### It never keeps a count past 18 months A weekly row is deleted 18 months after the week it covers, so the table stays a rolling 18 months rather than a permanent behavioural record. Switching the feature off deletes them all. What it does do, stated once: it reports named individuals to your admins, it reads the private channels your authorizing admin is already in and has selected, it reaches back up to a year into what those people already posted, and it gives the people it counts no choice and no message. Telling them is yours to do, and your confirmation that you did is recorded rather than checked. You remain the data controller. Built for ## For Leaders Who Lost the Room and Want It Back. In an office you could feel someone go quiet. Distributed teams took that away and replaced it with a channel that scrolls a little slower each week. ### People teams at distributed companies Where the office used to tell you someone had gone quiet, and remote work took that away. This puts a measurable version of it back, per person. ### Engineering and product leads For the person who used to run the channel and has stopped posting in it, while their tickets still close on time. ### Employers who will say they are doing it The measurement is named, so it only works if you are willing to tell people yourself. TeamPredict will not do it for you, and it will not start counting until an admin confirms you have. What a decline actually tells you ### It Is a Prompt, Not a Verdict. A declining status means one thing: this person posted fewer messages in the selected channels than they usually do. Message volume is a weak proxy for engagement, and it is worth knowing how weak before you buy this. It falls for annual leave, for illness, for parental leave, for a move into deep work, and for someone who simply prefers talking to writing. - A drop can mean leave, illness, or a stretch of public holidays - It can mean a crunch where someone is heads-down, or a week that turned into calls - It can mean the conversation moved into a meeting, a DM, or a channel you did not select - It can also mean the thing you are worried about, which is why you ask rather than assume The honest use is to ask the person what changed, in the next one-on-one. If you want early warning from outside the company instead, that comes from public LinkedIn signals: [see how the risk score works](https://www.teampredict.ai/retention-software). FAQ ## The Questions Your Team Will Ask Per-person visibility, who does the telling, how far back it reads, private channels, GDPR, and how much a message count is really worth ### Can I see one person’s numbers? Yes. That is the main feature. Every counted person appears by name with their weekly message counts, their usual level, how far below it they are now, and a status. The list is sorted steepest fall first. Anyone with an admin role in your TeamPredict organization can see it, including an admin who was not part of the decision to switch it on. ### Do employees know they are being counted? Only if you have told them. TeamPredict does not tell them and cannot: it holds no permission to post anything in your workspace and sends no message to anyone. Telling your workforce is your job, done through your own channels. Before a single week is counted, an admin in your TeamPredict organization has to confirm in the product that your company has already told its people what is being recorded, that it is shown to admins by name, that up to the last 12 months of it may be read, and that they cannot opt out. We record who confirmed it, when, and the exact wording they agreed to. We do not check whether it happened, and we have no way to: if you confirm and have told nobody, the counting still runs and that is on you, not on us. There is also no opt-out. Nobody can decline and nobody can make the counting stop, so if someone objects, the people who can act on it are at their employer. ### How far back does it read? Up to the last 12 months, and only after your admin has confirmed your team was told. The order is the point: nobody’s past is touched until your company has stated it disclosed the measurement. Once an admin asks for the history, TeamPredict reads the selected channels backwards and folds what it finds into the same one number per person per week, so the baseline is there when that read finishes rather than a month and a half later. The read is not instant: it runs in the background and works through your channels in batches, so a workspace with a lot of them takes several passes before its history is complete. How far it actually reaches depends on your Slack plan. A free workspace serves roughly 90 days of channel history whatever is asked of it, paid plans serve more, and there is no way to know which you will get until the read is done. TeamPredict records the oldest week it really reached and shows that, so a workspace that returned three months says three months instead of implying a year. ### Does it read DMs or private channels? It never reads direct messages. TeamPredict holds no DM permission of any kind, not to read one and not to open one: it cannot post in your workspace at all. Private channels are different: they are read when the admin who authorized the connection is already a member of them and has selected them. That access comes from that admin’s own Slack account, so it reaches exactly what that person could already read and nothing more. Slack’s Discovery API is not requested and not used. ### Which channels are counted? The ones your admin selects, one at a time. Public channels come from the app’s own Slack token. Private channels appear in the list only if private channels are switched on and the authorizing admin is a member of them. A newly found internal channel arrives already ticked, so a channel you do not want counted has to be switched off. Slack Connect channels, the ones shared with another company, arrive unticked instead, because the people posting in them work for someone else and your confirmation does not cover them. They stay in the list and you can tick them if you mean to. ### Is this legal? We make no claim either way, and we are not the ones who disclose it. Nothing is read until a workspace admin connects Slack and an admin confirms your company has told its workforce, nothing at all is read before that confirmation is recorded, only message counts are stored, and counts are deleted 18 months after the week they cover. But the confirmation is your statement, not our verification: we do not check it and cannot. And this is per-person monitoring of named employees with no opt-out, reaching back up to a year into what they already posted, and the rules for that differ by country, state, and role, with works councils and employee-representative consultation on top of it in parts of Europe. You are the data controller. Having a lawful basis, and giving any notice or consultation you owe, is your job and not ours. ### What about GDPR? We claim no certification and no compliance verdict. What we can describe is the design: only integer message counts are stored, no message content is ever read or logged, nothing about anyone is read until an admin at your company confirms your workforce was told, past weeks included, weekly counts are pruned 18 months after the week they cover, and switching the feature off deletes them. What we cannot describe is the disclosure itself, because we do not make it and do not verify it: we hold a record that your admin said your people were told, and that record is a claim rather than evidence. There is no consent step and no opt-out anywhere in the product, so if consent is the lawful basis you need, this feature does not give it to you. You stay the data controller and decide the basis for your own workforce. ### How reliable is a message count? On its own, weak. It measures how much a person types in the channels you selected, and nothing else. It falls for annual leave, illness, parental leave, a week of heads-down work, a week full of calls, and for a conversation that moved into a meeting or into a channel you did not select. A sustained drop is worth a conversation. It is not evidence that someone is unhappy, underperforming, or leaving. ## Tell Your Team. Then Watch the Trend. Slack team health is included with monitoring, off until you enable it, and empty until you confirm your team was told. Then the last 12 months are read, as far back as your Slack plan still holds. [Start 30-Day Free Trial](https://app.teampredict.ai/signup) No charge today · Cancel anytime