Notification controls: a customer wants to mute a topic, not disappear from the system
Separate notification topics, channels and timing. Design preferences, temporary muting and a return to work without promising delivery.
Article language: English

In this article
Ewa mutes a project because someone is correcting captions on forty photographs. She does not want forty messages. She does want to know when a publication decision is waiting for her. If the setting reads "Notifications: on / off", she must choose between peace and knowing about her own task. That is a poor negotiation form with an application.
The example is fictional: we are designing a portal where several companies prepare material for a shared catalogue. Ewa approves selected collections, sometimes comments on photographs and merely follows some projects. Settings should reflect those differences. We are not designing the legal rules for marketing contact or guaranteeing delivery here. We are establishing what a user can choose knowingly and how they can understand the effect.
The single label "notification" can cover a new comment, a deadline reminder, an email with an action result and a system alert on a phone. Until the team separates those things, a convenient screen will merely be an elegant remote control for an unknown device.
A topic is different from where it is delivered
First name the events that actually matter to Ewa. A comment added to a followed collection carries a different weight from a decision assigned to her. A small photograph update need not interrupt work immediately. For each event, establish the recipient, the reason to inform them and a route to the relevant part of the project.
Only then choose channels. The same event may appear in a dashboard, an email and on a device. Three selections should not mean three different obligations. They are three ways of reaching the same matter. If reading one message changes the others, the product should state that clearly; if it does not, do not call all copies one synchronised inbox.
GitHub's notification settings guide distinguishes the types of activity watched from the ways updates are received. Certain repository events can be selected; participation in a conversation or a mention is a separate reason for subscription.
A similar distinction helps name settings in our portal without copying GitHub's whole model. Ewa might want email about tasks assigned directly to her and view updates to followed collections only in the dashboard. A "Turn off email" control should change the channel, not delete the assigned task or remove her from the project.
A preference map can start with four specific statements from the user:
- "I want to know when someone asks me for a decision." That chooses an event type.
- "I do not also need this on my phone." That chooses a channel.
- "I will read small updates together in the afternoon." That chooses a frequency, if the product supports summaries.
- "For one hour I do not want any interruptions from this collection." That is a temporary mute with a defined scope.
Turning those four statements into one switch removes the meaning the user needs. There is no need to show the whole table of combinations at the first visit, however. Start with understandable groups and reveal detail where there is an actual decision.
"Mute" needs an object
If Ewa mutes a particular collection, she expects the effect to concern that collection. Turning off notifications for her entire account would be a surprise. Organisation settings should also name the company. Someone working across several spaces (in Polish) may want different rules for each one.
Name the result before confirmation: "Mute Autumn collection updates until tomorrow". Show whether this covers every channel and what happens to direct requests for a decision. If those will still arrive, Ewa should learn that in the same place. A note hidden in long help text does not explain the button's effect at the moment of use.
Slack describes temporarily pausing notifications and setting a quiet schedule. Its documentation also identifies an exception: a direct-message sender can notify someone about an urgent message once per day despite the pause. The scope of a mute therefore matters together with its exceptions.
We can adopt different rules for the catalogue. If Ewa mutes updates only, say "Mute updates" rather than "Do not disturb". If an exception exists for an assigned task, show it as part of the rule. Do not call every message urgent merely because its author selected a field. The product owner must establish who may use that exception and for what purpose.
A temporary setting needs a visible end. "Until tomorrow" can be a convenient choice label, but confirmation should explain the resumption time using the time shown to Ewa. Turning something off permanently is a different decision. Do not restore notifications after an arbitrary period if the screen promised a mute without an end time.
For a recurring rule such as evening quiet hours, check changes of day and time zone too. This need not mean designing a full calendar, but it does require a clear answer about which clock the schedule uses. The same issue appears in recurring deadlines (in Polish), where a rule must work across later dates. "Every day 18:00 to 08:00" alone does not explain whether it still means Ewa's evening and morning after a time-zone change.
A mute should be findable and reversible. A short message by the collection can show its scope, end and a resume button. If the only trace is a transient toast after a click, Ewa will not know the next day whether notifications failed to arrive because of her choice or a fault.
If a collection has its own setting, show how it relates to the account rule. For example: "Uses account setting: assigned tasks only" or "Custom choice for this collection: no emails". Returning to the account rule is then a separate, recognisable action. An empty switch alone does not say whether it means off, no choice or use of the parent setting.

After changing the account rule, Ewa should know whether the collection exception still applies. For our catalogue, suppose it does: the exception remains visible until she removes it. Acceptance will therefore cover a collection without an exception and one with a custom choice. Only comparing them will show whether the description of the hierarchy agrees with the result.
Device permission does not replace a product preference
Ewa has enabled messages about decisions on her phone. The site does not, however, have browser permission for system notifications. These are two different settings. The screen may correctly remember her choice yet be unable to display an alert. An unexplained "On" status makes a promise she cannot verify.
MDN describes the results of a site's request for notification permission: granted, denied or an unknown decision. In the last case the application acts as if permission were denied. The request should follow a user action.
The design can therefore show the preference and channel availability separately: "Selected for tasks assigned to you; the browser is blocking notifications". The next button must lead to an action actually available. Do not offer a button promising another prompt if the browser setting needs to be changed. Nor should notifications for the entire account be labelled muted when the block concerns one device.
For camera, microphone and location permissions (in Polish), we also separate a choice in the product from permission in the environment. Here there is an additional question of message arrival. Even permission to display is not proof that a particular notification arrived and was noticed. Confirmation of a preference change should say that the setting was saved.
Also test failure to save preferences themselves. After a channel is switched off, a control may move immediately before the server confirms the choice. On error, Ewa needs to see whether the previous rule still applies. Do not leave a visually disabled channel beside "try again later" if messages will still be sent. Retain her other choices too, so retry does not require setting everything from the beginning.
Email likewise needs an identifiable recipient. If the account has several addresses, show which one serves the selected channel. Changing a company's contact address need not automatically change Ewa's personal notifications. The product defines that relationship and the screen must represent it honestly. For an unverified address, show the required step instead of assuring her that messages are already reaching the inbox.
It is useful to compare such rules with reducing interruptions in team work (in Polish). A settings panel alone cannot explain what a team considers urgent or where a matter goes without an immediate answer. In our example Ewa still has her own list of decisions; muting alerts does not remove her responsibility for the catalogue.
What remains after a mute
Muting updates need not erase information from the collection's history. Ewa may want to return after an hour and see what changed during the quiet period. If the product works that way, design a path to the outstanding events. If the history covers selected actions only, name that scope rather than promising a complete record of everything.
There is no need to send a separate message for every small change missed when notifications resume. A summary or new-event markers in the dashboard may be an option. First establish what the actual mechanism does: delay delivery, skip notifications or prepare a digest. These variants create different expectations at the end of the quiet period.
A summary has its own meaning. If information arrives at 16:00 about a request Ewa already answered at 15:00, it should not claim that her decision is still pending. The event can be shown as history or omitted under an agreed rule. A notification about a past event and the current state of a task need different wording.
Name the role of messages necessary to a particular process. Someone requesting a link to regain access should recognise that they are deliberately triggering the message. Do not hide that action under a general "news" category. At the same time, classifying a message as needed for a process does not determine its legal basis. This article organises the design; it does not replace a separate assessment of communications.
From transactional messages (in Polish), retain the principle that event, subject and next action should agree. Settings must make clear which part of that communication the person controls. Do not present a limited preference as a "Never contact me again" button if the product still provides for specified exceptions.
Acceptance with an event that should not have appeared
Choose a specific account and collection with test data. Mute updates, add a new comment from a second account, then assign a decision to Ewa. Compare both results with the setting's description. A saved switch on screen alone will not show whether the mechanism distinguishes events the way the design does.
Repeat the test for different channels. Also check the case where system permission is blocked even though the product preference was enabled. The person accepting the work should distinguish an unsaved choice from an unavailable device and from unconfirmed delivery. Not everything missing from a phone means the same fault.
When changing a rule, inspect messages already waiting. Ewa may turn off email shortly after an event occurs but before a digest is sent. Agree at which point the preference is applied and test the result. If a change covers only future events, the save message should explain that, especially when earlier messages may still arrive.
At Aeneia Design, an event map and the current settings screens can form material for the work. They help connect language to actual behaviour. Ewa wants to make the publication decision without forty separate notifications that someone has just corrected another comma in a caption.









