Skip to content
Interface language: English
Design / Signal

A beautiful dashboard, yesterday's data: showing freshness without pretending to be real time

Distinguish the time data describes, its arrival and the result of refresh so a dashboard can show delay, gaps and errors honestly.

Article language: English

A person tends a white illuminated petal near a resting cat in a garden at night.
In this article

Daria presses "Refresh". In a fictional parcel collection dashboard, the words "Data from moments ago" appear, but the parcel count still comes from the morning's report. The time the view was fetched has changed. The clock on the screen is accurate; that alone has not made the information needed to plan collection any younger.

A dashboard design should distinguish the time data describes, the time it arrived and the state of the latest refresh attempt. Not every detail must occupy the heading permanently. The user should, however, know what state they are looking at and whether they can use it for the next step. An animated arrow does not confirm a new measurement.

One number has more than one time

Daria coordinates collection of ready parcels from several workshops. In our fictional process, one sends a report describing the state at nine o'clock. The system receives it twelve minutes later, and Daria opens the dashboard at 9:40. Each time describes a different event. An "Updated 9:40" label does not say how many parcels were ready at that moment.

A basic value description could say: "7 parcels ready as of 9:00". The details say that the report was received at 9:12. If we read the same version again at 9:41, we can show "Last checked 9:41; no newer report". Daria can see what succeeded and what remains unconfirmed. This is a proposed interface behaviour, not a description of an existing integration.

First establish with the person responsible for the data which times are actually available. Do not design an "as of" label from a value that means only the import time. If the report preparation time is unknown, show that limit. You can describe the received file and its delivery date, but do not invent an unknown observation time.

Differences between stages also occur in existing tools. Microsoft describes a Power BI model in Import mode as a copy of data from a specific point in time. Report queries use imported tables, and fetching changes from the source requires a model refresh. Reading a view and refreshing the model are therefore different actions.

There is no need to transfer that architecture to every dashboard. Your implementation team should explain the data path, and the designer should translate it into an honest promise on screen. If a button only rereads the available report, its message cannot claim to have checked the entire source. If it starts a longer process to prepare a new dataset, the user should be able to recognise that work is in progress.

In a Grafana guide to integration dashboards, refresh repeats queries while data appearance depends on collection. Those are two stages. The design implication for Daria is simple: asking more often for the same available state need not produce a newer observation. The interface should be able to show a successful read without a new value.

An unchanged number does not itself mean refresh failed, either. The workshop may genuinely still have seven parcels. The operation state and report identifier are needed, not a guess based on a difference in numbers. When new data has the same value, its preparation time may change in accordance with the actual event. When an old copy is reread, do not change its data time.

Design the period name separately from the freshness time. "Today" may mean the total number of parcels prepared since midnight, rather than the number currently waiting for collection. Collecting some parcels then need not reduce the measure. Daria needs specific information for route planning. If the dashboard shows a different measure, change the description or choose the right result instead of polishing the date alone.

Designing charts in a brand's visual language (in Polish) likewise involves defining the measure and labels. Here the underlying state can also change. The time label should belong to the same component rule as the number, unit and status. It cannot be decoration updated whenever the screen redraws.

Missing, delayed and failed need different responses

Suppose the workshop is expected to send a report every half hour, but at 10:15 the dashboard still shows the state from nine. This rule belongs only to our example. An older value alone does not say whether nothing changed, delivery is delayed or the system failed to read it. The design should preserve known facts and indicate the appropriate state.

If the source confirms a new report with zero parcels, show zero with its time. If no report arrived, do not substitute zero. If fetching failed, the message should describe a read error. You may retain the previous value as the last known state, but with a visible warning and the correct time. A number without that context could lead Daria to send a vehicle to the wrong address.

Distinguish a first opening from a later failed fetch. At the first opening we have no value to show. On a later attempt we may have an earlier confirmed state. The same empty rectangle does not explain both situations. The component design should cover each separately, with an available action: retrying the read, identifying the source of the problem or contacting someone according to the actual process.

Do not remove a correct number from a neighbouring card because another workshop failed to provide data. At the same time, do not present a total as complete when it covers only some workshops. The heading can identify the scope, while an outdated card shows its state. Daria should recognise which part of the route she can plan and which stop needs confirmation by another route.

Partial data is a special case. The system received reports from two of three workshops but does not know the third result. "14 parcels" may be a correct sum of received entries, yet not a complete total for the whole route. "14 parcels; data from 2 of 3 workshops" explains more than a small information icon by the heading. Users should not discover the omission only after beginning an action.

If a model works on such input, checking alerts and missing data (in Polish) is a separate task. The dashboard designer should understand the possible states and give them a readable form. A warning colour alone does not confirm measurement quality. Scope and time must remain available without reliance on colour recognition.

Aeneia conceptual photography

For ageing data, establish a rule suited to the task. A daily statistic need not warn after one minute. Parcel status for urgent collection may need more frequent confirmation. The designer should not choose a limit based on how much room remains in a label. The process owner's expectations and the actual ability to deliver data are needed.

PatternFly's timestamp guidance covers local format, time zone and a more precise representation alongside relative time. For the dashboard, this suggests a rule: "15 minutes ago" gives a quick sense of age, while a specific date and time let someone relate the value to an event. The label should describe the right stage of the process, not an arbitrarily chosen date.

Relative time must age while the view remains open. "Just now" cannot stay that way throughout the morning. When the user returns to a tab after a break, they need a visible freshness state. Merely displaying the dashboard again does not justify resetting the data time to the present.

The exact time should also be available to someone who does not use a mouse. Do not hide the only date in a hover-only description. After a longer break, the hour needs a day because "9:00" alone does not distinguish today's report from yesterday's. When the process spans time zones, show which zone the label uses. Do not compare two local times as though they had a common reference point.

In an Aeneia Design dashboard project, these card states, messages and refresh behaviour can be designed together. The starting point is an explained data path and the reader's task. Mock-ups should also show incomplete and delayed results, because that is when labels stop being a small detail beneath a large number.

Aeneia Design

Show us the dashboard, the sources of its values and the decision it should support. Let us design the meaning of dates, incomplete-data states and refresh messages that reflect actual behaviour.

Discuss data freshness in your dashboard

Keep the meaning of the whole view during refresh

The button should show whether a read is in progress and what its result was. If another click does not make it faster, do not leave users thinking repeated presses are needed. For longer report preparation, use behaviour appropriate for a task someone can return to (in Polish). A short animation should not pretend indefinitely that the wait will be only a few seconds.

After a successful refresh, retain the user's selection if it still makes sense. Daria may be viewing one workshop or a particular route. Returning to all data under an unchanged heading introduces a different mistake from staleness. The filter, values and time label must describe the same scope, including in the operation summary.

Test what happens when a response returns after a quick workshop change. Daria first selects one workshop and moves to another before the fetch finishes. If the earlier response arrives later, it must not appear beneath the new selection's heading. On error, do not leave the old seven parcels as the new workshop's value either. Retaining a last known state makes sense only when the reader can see which object it belongs to.

When refresh changes item order, preserve a point of return. After cards move, the selected workshop should remain recognisable. A new rank must not silently change the object whose details the user is viewing. The user should recognise the data change rather than search again for the object they were working with. Preserving context may justify a stable order or a visible marker for the selected item.

An open detail and its summary card also need an agreed relationship. If the detail stays on an old version, show that state or refresh it according to the designed rule. Do not present seven beside a new item list totalling nine as though both views describe the same moment. The meaning of time should remain visible after moving to an interactive chart. The guide to interactive charts (in Polish) discusses its design.

Sharing a view requires a separate decision: does the recipient get the current dashboard or a saved state? A screenshot should contain the scope and correct date because it will not refresh with the source. A link, by contrast, may open a newer version. The two materials serve different purposes even if they share a title.

Data export (in Polish) works similarly. The downloaded file should match the selected scope and have an understandable state label. If the export is prepared later than the view, explain whether it captures the earlier dataset or fetches a new one. "Downloaded at 10:20" does not replace information about the time the figures describe.

For acceptance, prepare several different states of Daria's task: a new complete result, a successful read with no newer version and one missing workshop. A person going through the design should recognise the data state and the next step. Finding a button with a circular arrow is not enough. Also record how they interpret the time label, because its meaning is what we are testing.

In our scenario, Daria can now distinguish seven parcels confirmed in the morning from an unknown current state. She can plan part of the route and obtain the missing confirmation according to the support process. The dashboard did not speed up data delivery. It stopped presenting an uncertain state as a fresh fact, allowing the right work to begin.

Aeneia / Aeneia editorial

Editorial article. Source links are included in the text.

About Aeneia

Brands we have worked with

  • L’Oréal
  • Garnier
  • Netflix
  • Carrefour
  • Warner Bros. Discovery
  • Biedronka
  • HBO Max
  • Wolt
  • Żabka
  • COSRX
  • Embassy of China in Poland
  • Mixa
Aeneia / Initium

From knowledge to a decision

Tell us what you want to put in order. We will agree on where to start.

Contact Aeneia

We partner with and use solutions from:

  • Google
  • OpenAI
  • Microsoft
  • Stripe
  • Anthropic
  • Linear