Painéis (BI)

Each row is a dashboard built here. The list shows which groups see each one, how many components it has, who created it and who last changed it — click the name to open and edit.

Monte painéis com cards, tabelas e gráficos: escolha a fonte, configure, posicione, filtre, pré-visualize e lance. — este é o papel da tela "Construtor de painéis" dentro da plataforma.

A dashboard is a set of components on your own data: cards with a number, tables, bar, column, donut and pie charts. You choose the source, what to show and how to slice it — without asking IT for anything and without waiting for a new system version.

CARD shows a single number — the count of records or the sum of a field, calculated in the database over the whole set. TABLE lists row by row. BARS and COLUMNS show the same count grouped by a dimension; bars read better with long labels, columns read better in a time series. DONUT and PIE show COMPOSITION: how much each category represents of the total.

The screen is divided into 12 columns. A component of width 12 takes the whole row; 6 takes half; 3 takes a quarter. On the phone everything becomes a single column, so a dashboard built on the computer stays readable on the phone without you building it again.

Grab the component by its handle (the lines icon on the left) and drop it wherever you want: onto another component, to go before it; onto the row's area, to go to its end. The arrows still work for those who prefer to click — ←→ move within the row, ↑↓ between rows.

In date filters, prefer relative periods (last 3 months, start of current month) to typing a date. A fixed date ages: the dashboard keeps showing August in December, looking current. A relative period is recalculated every time someone opens it.

The preview button runs with REAL data, through the same path as the published dashboard — what you see is what will come out. You can open the whole dashboard in a new tab even as a draft, and nobody else sees it until you launch it.

In Dashboard data, choose how often the open dashboard reloads itself (30 seconds to 15 minutes) — made for a dashboard on a TV or wall, where nobody clicks. TV mode enlarges the numbers for reading from a distance, removes the export bar and gains the full-screen button. Combine the two.

In grouped components, each numeric field has its own function: sum, average, median, maximum, minimum or count. The column header says which calculation you are seeing. Use average or median for lead time, percentage and score — summing those quantities gives a number that exists and means nothing; the median ignores the extreme value that distorts the average.

When filtering by a text field, choose among the values that EXIST in that field, with search. Typing still works ("contains" accepts part of a text), and period accepts the format 08/2026. A field with too many values goes back to free text.

Ask Tenkai "revenue by customer in the last few months" or "how many tickets by priority": it shows a preview of the numbers and an "Add to dashboard" link. The link opens this builder with the component ready — source, metric, axis, chart type chosen by the shape of the question (a series over time becomes a line; one dimension becomes columns; a share becomes a donut; a single number becomes a card). You choose the dashboard, adjust whatever you want and save. Nothing is saved until you save, and the component records which question it came from.

"Suggest components" looks at the sources already used in this dashboard (or all of them, in an empty dashboard) and proposes cards and charts that don't exist in it yet — each suggestion has a button to add it. Inside a component's popup, "Improve with Tenkai" asks the assistant for a better title, the most suitable axis and the period comparison; the suggestion only goes into the component if you accept and save. The manual way stays the same: no button was removed.

The panel form follows the standard: name and description on top, and on the Data tab also who sees it, the e-mail delivery and TV mode; the components on the Layout tab (drag, arrows and the pencil that opens the settings), the panel filters on the Filters tab and the history on its own tab. Save, Save and close, Duplicate and Ctrl+S sit in the header, next to Launch and the preview.

When you ask Tenkai for "a dashboard of X by Y", the Build the dashboard link opens this builder with the proposed components (total, trend, ranking, share). Check the name and save: the components go in right away, each marked as created by Tenkai. To adjust a chart without repeating the question, tell Tenkai "switch to donut", "only the South region" or "last 6 months".

On the component, choose the Child dashboard and the Argument (the child filter key): clicking a value opens the child already filtered. Export JSON takes the whole dashboard (components, filters and profiles by name); Import creates a draft dashboard and warns about what does not exist here. On each filter, Remember last decides whether the dashboard opens with each person's last value.

DRAFT is yours: only the builder can open it, through the preview. PUBLISHED appears in the menu of those with the chosen profiles. You can return a dashboard to draft at any time — it leaves the menu without being deleted.

If the list flags "published without any group", the dashboard is live and nobody sees it. It is the quietest state on the screen: it throws no error, it does not disappear, it simply does not show up for anyone. Open the dashboard and choose at least one profile in "Who sees this dashboard".

When you save a view with the Views button, it stores four things: the search text, the applied filters, which columns appear (and in what order) and which column the list is sorted by. Applying the view restores everything at once — including resetting the sort and the page you were on, so the result is exactly what the view describes. Views saved before this version remain valid: whatever they didn't store simply doesn't change when applied.

Nas fichas, sair de um campo já dispara a conferência do que dá para conferir na hora: campo obrigatório deixado em branco e e-mail mal formado. O aviso aparece embaixo do campo e desaparece quando você começa a corrigir. Campo numérico não precisa de aviso: ele simplesmente não aceita letra enquanto você digita. Isso é conforto, não a validação final: ao salvar, o sistema confere tudo de novo do lado do servidor.

Uma métrica nomeada carrega dono, definição (o que mede e como se calcula) e validade. O Tenkai só responde "como está o MRR" quando a métrica tem definição e dono — sem contrato ele diz o que falta em vez de inventar um número. Vencida a validade, a resposta avisa que o dono precisa revisar a definição.

New dashboard → give it a name and choose WHO SEES IT (the profiles) → Save → add components → configure each one → Preview → Launch dashboard. Until you launch it, the dashboard is a draft and only you see it. Once launched, it appears in the menu of whoever has the chosen profiles.

Bars and columns answer "which is bigger"; donut and pie answer "how much each one represents of the whole". If your question isn't about the total, use bars — a pie with many categories becomes a striped wheel and can't be read. The donut draws the 7 largest and adds up the rest in "Others", saying how many were grouped, so you don't think they disappeared.

The COMPONENT's filter narrows only that card or table — it's what separates revenue from expenses within the same source. The DASHBOARD's filter appears at the top and applies to all of them at once, like the filter on list screens, and each person keeps the last value they used.

In the table component, create rules like "SLA less than 50 → yellow". The FIRST rule that matches wins: put the most restrictive one first ("less than 30 → red" above "less than 50 → yellow"), otherwise red never appears. The colors follow the light and dark themes on their own.

On the card, configure the condition (e.g. LESS than 50) and the recipient. The alert is evaluated when the dashboard renders — on a TV with auto-refresh, that means continuous monitoring — and fires at most ONCE a day, so it doesn't become noise people learn to ignore.

In the record's activity panel, each note has "Turn into request". The click creates a ticket with the note's text, who wrote it and the link to this record — nothing is retyped. Use it when the note is a request for action ("someone needs to review this"); the note stays where it was. The button only appears once the record has been saved; a read-only session (demo) doesn't create one.

"Who sees this dashboard" still decides who opens the dashboard. Inside it, each component can be restricted to some groups: in the configure-component popup, "Who sees this component" lists the groups; with none checked, the component inherits from the dashboard. A group that opens the dashboard but isn't in the component simply doesn't see that card or chart — no warning, no empty space. That's how margin stays with the board and volume with operations, on the same dashboard.

In the same popup, "Row rules for this source" creates MANDATORY filters per group (or for everyone) on the dataset: field, operator and value. The value can be an attribute of whoever is looking — {usuario.codusuario}, {usuario.email}, {usuario.nome}, {empresa.codempresa} — and then the same rule works for everyone: "owner = {usuario.email}" shows each person only what is theirs. The rule applies to the dataset across the whole company: dashboard, Explorer, API, MCP and Tenkai's answers obey the same row. A company administrator is not filtered. A rule with a field that doesn't exist in the dataset is ignored and noted in the log — it never becomes a wrong query.

Every named metric (Save as metric) starts having one point per day, recorded by the BI's daily snapshot (Parameters ▸ BI ▸ Snapshot time, default 06:00). The card that uses the metric shows below the number a "30-day average" and how many points already exist; Tenkai answers "how is MRR compared with last week" by reading the series, without recalculating anything. The series starts empty on the day the metric is born — the platform doesn't make up a past.

Besides "greater than X", the card alert accepts the Deviation operator: the value is how many standard deviations from the average of the last 30 points trigger the warning (2 = outside what happens on ~95% of days). It requires a named metric (it's the series that gives the average) and at least 7 points — before that the alert stays silent and the screen says why. It fires at most once a day, by email, like the others.

When you pick a named metric in a component, the Metric contract box appears. Write what it measures and how it is calculated, choose who answers for it and, if you like, until when the definition holds. Without a definition and an owner, Tenkai does not quote that metric in its answers; once the review date has passed, it quotes it with a review warning.

Open the record and look at the Timeline on the right: it shows each change as "from one value to another", with the person and the date. When there are many changes, use the two filters at the top of the timeline — one by field, the other by person. The search runs in the database, so old changes show up just the same. You can also ask Tenkai: "who changed contract 482?".