Build your own dashboard gauges with AI

Dashboards is its own page, pinned in the left navigation directly under Portfolio. Open it and you get a board you build yourself, by asking. Describe the number you care about in plain words and Current draws it as a gauge, a leaderboard, a trend line, or whichever shape reads best. It used to be the first tab on Reports; if that is where you look for it, the old link still brings you here. Reports is its own page, grouped Delivery, Team, Finance, Service and Sales, and it opens on Portfolio overview.

It is a staff surface. Anyone on your team who can open Dashboards can build one for themselves. Partners never see it, so nothing you build here can reach the portal.

You do not start from nothing. Every workspace is given one board called MSP Scorecard, and it is there the first time you open Dashboards. It holds the numbers that tell you how the business is actually running, in four groups: Service health (open tickets, opened and closed this week, kill rate, first response, SLA compliance, and how old the backlog is), The MSP money numbers (monthly recurring revenue, managed services margin, effective hourly rate, net revenue retention, reactive ticket hours per endpoint, and endpoints per partner), Sales pulse (open pipeline, forecast to close this quarter, deals won, quotes won), and Risk radar (backups failing, open EDR alerts, phishing caught, devices compliant).

It is built from what your workspace has actually connected, so nothing on it sits empty. A gauge that needs a product you have not connected is not placed at all, rather than added and left reading zero. The same goes for the permission walls: a gauge no role in your workspace can see is left off too.

A short line under the toolbar tells you what the scorecard is still missing, and it is worked out fresh every time you open the board rather than written down once. It separates the two cases, because they need different things from you: gauges that can be added now, which are ready to build this minute, and gauges that arrive when you connect the products they read. It counts only what you personally can see, so it never points you at numbers your access does not include. Add a tile, connect an integration, and the line shrinks on its own; complete the board and it goes away. There is nothing to dismiss, because it is the state of the board rather than an announcement.

It is shared with your team, and from the moment it arrives it is an ordinary board. Rename it, rearrange it, resize the tiles, add your own, drop the groups you do not use, or delete the whole thing. Nothing about it is protected, and a tile you remove is not put back. There has been one exception, for HaloPSA, and the note below says exactly what it was.

1. 1
   Open Dashboards

   It is pinned in the left navigation, directly under Portfolio. Quick search finds it too: press the search shortcut and type dashboard, or wallboard. A link you saved to the old Dashboard tab on Reports brings you straight here.
2. 2
   Describe the number you want

   Click Add a tile and type it the way you would say it out loud. Tickets we closed this week. Won revenue this quarter by rep. Hours logged this week by person. On an empty dashboard the sample prompts sit right there as chips, filtered to the integrations you actually have connected, to the PSA you run, and to what your role can see, so anything you click is something Current will build. Clicking a chip lands you straight in Edit tile with a live preview, no questions asked. Prefer a whole board at once? Start from a set offers ready-made boards for the service desk, sales and leadership, trimmed to what your workspace can actually draw, and tells you when a set arrives smaller than advertised and why. A set already on your board says Already on this board rather than adding a second copy of every tile, and one that is partly there adds only what is missing and tells you how many it left alone.
3. 3
   Answer the three questions

   Type your own request and Current always asks three short questions about it, composed to fit what you asked: which product it should pull from, which ticket queues it counts, what window it covers, how it breaks down. Each one offers three suggested answers to click plus a box to type your own. Answering is optional on every one, so you can answer none of them and carry on. Start over clears the round and puts you back at the box. The starter chips on an empty dashboard skip the questions entirely: those are already worked out, so Current builds straight from them.
4. 4
   Pick one of the tiles Current drew

   Next you get up to three finished tiles side by side, each drawn with your own numbers rather than a picture of what one might look like. They are read by your browser under your own permissions, so what you see is what you would get. Click the one you want, or move between them with the arrow keys. Start over goes back to the box.
5. 5
   Finish the tile, then add it

   The last step is the tile itself with a live preview beside it: its name, how it is drawn, which ticket queues it counts, whether monitoring alert tickets count, which partner it covers, how it breaks down, its target and its colour bands, already filled in from the industry benchmark. Any shape the tile's data cannot fill is greyed out with the reason, so you never add a tile that would sit empty. Add to dashboard writes it, and the board lets you move tiles and resize them in both directions. Close the step with changes you have not added and Current asks first.

Every tile keeps a small menu in its corner with two edit doors and nothing else that edits: Refine with AI (say what to change, in words, such as show me the last six months) and Edit tile, which opens everything about the tile in one panel. Beside them sit Move to section, Refresh now, Duplicate and Delete, plus See the rows where the tile can open its records and Mark verified on a tile Current worked out for itself. Duplicating a tile Current worked out for itself always hands you an unverified copy, so the sign-off belongs to a number a person checked rather than to a row that got copied.

If you are looking at a dashboard somebody shared with you and cannot edit it, every tile carries a What this tile counts button instead. It is read only and it shows the same facts an editor sees: the queues, the monitoring setting, the partner scope, the breakdown, the window, the benchmark and when the number was last read. When it could not read something, it says so rather than guessing.

A number you cannot open is a number you have to take on faith. So most tiles open: click the number, a bar, or See the rows on the tile menu, and a panel lists the records the tile counted, with sortable columns and a CSV export. Time and task tiles list their entries for anyone who can see them in the app. Ticket rows follow the CRM wall: if your role does not include CRM access you still get every count, but the row list stays closed. The panel tells you when a list is capped or measures something slightly different from the tile, so the two never disagree without saying why.

Words like first response, reactive, endpoint, billable and won deal mean slightly different things at different shops, and in most reporting tools they end up meaning slightly different things on two tiles of the same board. In Current each of those words is defined once, in one place, and every tile draws its wording from there. So the sentence under a number on your laptop, the same sentence in Present mode, the same sentence on a lobby TV and the explanation in What this tile counts are all the same sentence, because they are the same sentence.

You can read the definitions behind any tile without being able to edit it. Open What this tile counts from the tile menu and there is a section called How Current defines this, one plain line per term the number is built on: which day the workspace counts, which clock a response time was measured on, which queues count as project work, what makes a ticket reactive, what an endpoint is and which RMM it came from. Alongside it, as before, are the queues, the partner scope, the window, the colours and the freshness for that particular tile.

Open About this number on any tile and the first thing it says is the explanation: How it's calculated states the arithmetic in plain words, and How to read it says what a high or low number means and what to check before acting. Both are set as headings in the normal reading size, because they are the answer rather than fine print. Below them, This tile's scope groups the qualifications: the time zone the days were counted in, who is on the board, and anything somebody adjusted by hand. The same two explanations appear in What this tile counts, the read-only panel in the tile menu. Each definition also records what it deliberately leaves out and why. Reactive hours leave out tickets your monitoring raised by itself and the queues an admin marked as project work. Hours logged leave out time a manager rejected and time your PSA voided before invoicing. Endpoints leave out network gear, and a machine two of your tools both manage is counted once, with the tile saying how many were counted in more than one system. An exclusion with no reason written next to it is indistinguishable from a bug, so every one of them carries its reason.

A number on its own asks you to remember what good looks like. So tiles can carry a target and colour bands: green when a number is where you want it, amber when it drifts, red when it crosses the line. For service desk staples Current applies the industry benchmark on its own, and names the source right on the tile: SLA compliance at 95 percent, CSAT at 97, first response under an hour. Open Edit tile on the tile menu to set your own numbers, or switch bands off. A value with no band stays neutral. Colours never guess.

The AI also groups tiles into named sections as your dashboard grows, and the grouping is yours to change. Rename a section or dissolve it from the section's own menu, move a section up or down there, and send a tile to another section from the tile's menu. To add one yourself, open the dashboard menu beside the board name and choose New section. A section you make this way is saved straight away, even before you put anything in it. One that ends up empty because its last tile was moved or deleted is tidied away instead, so a heading over nothing does not linger. Two sections on one board cannot share a name: Current tells you before you add it, because the AI files new tiles by section name and a second heading with the same words would never receive one.

Every tile has two edit buttons in its menu, and they do different jobs. Refine with AI is the one you already know: describe the change and the AI rebuilds the tile. Edit tile is the other door. It shows you the tile's real inputs as ordinary controls with the tile drawn live beside them, and it never talks to a model unless you ask it to. It used to be three separate doors (Edit data, Edit thresholds and target, and Rename) with three separate saves. It is one panel now, everything is in it, the tile redraws as you change it, and one save writes the lot. Refine with AI is still its own menu entry, and it is also a box at the top of this panel, under the tile's name, so you can describe a change without leaving. It sits first because for most changes it is the quickest route: say what you want in a sentence, and everything below it is there for when you want to set something by hand.

Below that the panel reads as eight plain sections, in the order you would need them to work the figure out by hand. Where it comes from names the system the number is read out of. What it measures says what is being counted. How it splits is whether the tile shows one figure or breaks it up by person, partner or queue. What period is the tile's window. What it counts holds the queue list and the other narrowing choices. How it is judged is the target and the colour bands. How it is drawn is the shape. And Units and exclusions closes with the unit and what the tile deliberately leaves out, which lands last because a caveat before the claim is just noise.

Each section opens with one sentence saying what it controls, then the value in force on the tile today, then the control to change it, with the tile drawn live beside you. A More link at the end of each section opens the rest: the fuller explanation, what Current chose and why so you can see which parts of the tile were the AI's judgement and which were somebody's own, and the exact tables and queries the number is read from. On a tile that counts tickets, What it counts holds your own ticket queues as a checkbox list, with each queue's name and how many tickets are in it, plus the checkbox for whether tickets your monitoring raised automatically should count. Every ticket tile has counted them from the start and said so on its face; this is the switch that note promised. A tile built before this arrived has no recorded reasoning to show, so those sections say the choice was Current's default rather than inventing a reason after the fact.

What it counts also names any narrowing somebody has already put on the tile. A tile pointed at a single partner, a single project or a single person says so there and on the tile's own face, so a number for one person is never read as a number for the team. Where a gauge can be widened back to all of them, the section carries the checkbox that does it. To point it somewhere else instead, describe that at the top of the panel. Ticket gauges are the one place a person or project narrowing cannot be applied at all, because your PSA's tickets carry its own assignee and no link to a Current project. A ticket gauge carrying one of those says so plainly and shows the reason instead of a figure, and the same checkbox takes the narrowing off so the gauge draws again.

A tile Current composed itself, because no gauge in its own library fitted what you asked for, is editable here too. The panel offers what it works out and from which column, which date the dashboard's timeframe applies to, how it breaks down, how many rows it draws, and the conditions it counts, all built from the tables dashboards are allowed to read. Reading something else entirely is still a rebuild, so describe that at the top of the panel.

If a gauge is ever retired, the tiles built on it keep their name, colours and shape, and all three still save. What the tile counts cannot be changed at that point, so the panel says so and points you at the box that rebuilds it.

Nothing that affects a decision hides behind More. What a change will do before you save it, the reason a control is unavailable on this gauge, a list that is longer than the panel can show, and any read that failed all stay in plain view whether the More links are open or shut.

Before you save, the panel tells you in one plain sentence what your change will do to the number and which way it moves. Untick three queues and it says the number gets smaller, names how many tickets stop being counted, and says all of that before anything is written. A control that only changes how the tile looks says so outright rather than leaving you to wonder: switching the shape, the colour bands or the target never moves the figure, and adding a breakdown splits the same total across rows without changing the total.

A number tile or a gauge asked over a timeframe also answers what the same question gave for the period immediately before it, and says so under the number: up 12 percent vs the month before, down 4 vs the week before. A figure that has not changed draws no arrow, because no movement is not news. The comparison is worked out from the same records the tile is drawn from, so it is there on the first day you connect. Nothing has to be recorded overnight first.

On a leaderboard the same comparison becomes a movement chip on each row: how many places that person, partner or queue has moved since the period before, with new marked as new rather than as a leap from a position nobody held. Click a name and the panel draws that subject's last twelve weeks, one point per week, worked out the same way.

On any tile that breaks a number down by person, partner or queue, Edit tile carries a section called Who's on this board. It offers three answers. Everyone is the default and is what every tile you already have is set to. Everyone except lets you pick the few to leave out. Only these lets you name the handful to show and leaves out the rest. The list you pick from is the board's own, so you are choosing between the people, partners or queues this tile actually returns rather than a directory.

The filter is applied where the number is worked out, not by hiding rows after the fact. That is what keeps the places reading 1, 2, 3 with no gaps, and it means a leaderboard that shows five people is genuinely a board of five, not a board of seven with two rows painted over.

Who's on this board decides which rows a leaderboard draws. Two filters beside it do something different and larger: they leave work out of the number itself. Partners on this tile and People on this tile each offer the same three answers, Everyone, Everyone except and Only these. If you are offboarding a partner and want tickets closed without their work in it, this is the control. Unlike the board filter it does not need a breakdown, so it works on a single figure, an aging histogram, a daily opened and closed line, or a board broken down some completely different way.

The two are separate on purpose, so you can leave out a partner, or a contractor, or both. The lists you pick from are the partners and people this tile's own number actually covers, so a tick always matches real work. Somebody you have already excluded stays in the list even in a week where they have nothing, so narrowing the timeframe can never quietly put their work back.

The filtering happens in the database, before anything is totalled. That matters more here than it does on a leaderboard: an excluded partner's tickets are not in the population at all, so they cannot reach a total, an average, a percentage, an aging bucket, a ranked row, or the Other bucket at the bottom of a long board.

Under Where it comes from there is an Advanced link. Open it and the tile tells you the real tables its number is read from and the read it goes through, named exactly as they are in the database. Nothing is paraphrased and nothing is hidden. This is for the moment somebody asks where a figure came from and a plain-English sentence is not enough.

Advanced tells you where a number comes from. It does not yet let you send a tile somewhere else. Choosing which of your connected systems answers a number is a change we want to get right, because two systems do not count the same events, and it will arrive as its own update.

Nothing else in the panel is changed quietly either. A tile where somebody has overridden any of Current's choices leaves the word adjusted on its face, and About this number lists each change in plain words. Asking Refine with AI to redraw a tile keeps the settings you chose by hand, and if any of them cannot survive the rebuild, Current tells you which ones before it does anything.

Two things the panel will not let you do. If you switch a leaderboard, donut, bar or table to one number with no breakdown, the shape it was using has nothing left to rank or slice, so the panel says which shape the tile becomes and saves that instead of leaving you a tile that reads empty forever. And while no queue is ticked, Save is switched off rather than accepting the click and answering with an error, because the panel already told you to tick one.

A dashboard is a twelve column grid, so a tile has a width as well as a height. Grab the grip in a tile's top left corner to move it. To resize, hover the tile and pull any of the eight handles: the four edges or the four corners. The tile keeps drawing its number while you size it, and a dashed outline shows you exactly where it will land and how big it will be before you let go. Widths and heights snap to the grid, and the board closes its own gaps: when one tile grows, its neighbour slides sideways into free space on the same row rather than dropping down and leaving a hole. A dashboard you arranged before the grid arrived keeps every tile's width and height, but its rows now pack upwards, so a short tile beside a tall one rises to fill the gap instead of leaving dead space. No number changes.

You can do all of it without a mouse. Tab to a tile's grip and press Enter. Arrow keys then move the tile, Shift with an arrow key resizes it, Escape puts it back where it was, and every step is read out. On a phone or a tablet the board becomes one column in the order you saved, and each tile carries a move up button, a move down button and a height bar under it, so you can arrange the board by touch. Moving a tile there rearranges the whole board without losing the widths you set on a laptop.

A tile sizes itself to the space it has, and so does everything drawn inside it. Make a tile bigger and the gauge dial, the donut, the chart, the leaderboard rows and the big number all grow with it; make it smaller and they shrink to fit rather than spilling over the edge. The big number keeps a readable floor on a narrow tile, so nothing is ever too small to read, and a large money figure on a very narrow tile shortens to something like 47.6M in your workspace currency rather than being cut off halfway through a digit, with the exact number on hover. Below a certain size a tile draws a simpler version of itself rather than a squashed one: a narrow gauge moves its verdict badge up beside the number, and a tile too short for a dial draws the same value against its target as a bar instead. Long names shorten to a single line and the whole name is on hover. A table or a leaderboard never draws wider than the tile it sits in. Leaderboard rows are one line each, with the person's initials, their place, and a bar behind the figure showing how they compare, and a tile shows as many rows as it has room for and says how many it is not showing. Leaderboards number their ranks with a proper tie rule, so three people on the same count all show first and the next person shows fourth.

A tile's notes live behind About this number, at the bottom of the tile beside the dates it covers. Every caveat a number carries is still there, in the same words: which queues it counts, which clock it measured on, how the board is sorted, what was left out. They are one click away instead of stacked under the figure, so the tile reads as a number rather than a paragraph. Open it once and it stays open on every tile, and on every device you sign in from, so if you would rather always see them you only say so once.

Three things never hide behind that control, because they change what the number means: a warning that Current could not refresh this and is showing the last figure it read, a note that Current cannot yet confirm a security product has ever sent data, and the caveat on a tile Current composed itself. And where a hidden note would change how you read the figure, the tile leaves a short word on its face, such as includes alerts or adjusted. Click that word and the notes open. On a lobby TV nothing is hidden either, since there is nobody there to click: a note that applies to several tiles at once prints one time, in the strip along the bottom of the screen, and a note that belongs to one tile stays on that tile.

Colour is never the only signal on a dashboard. Where a tile carries a good, watch or at-risk verdict, it also shows a shape and the word, so the verdict survives colour blindness, a greyscale print and a washed out screen across the room. Hover the mark and it tells you the rule that produced it. On a bar or a trend chart the same rule applies: each threshold line prints its own rule in words, a toned bar carries a texture as well as a colour, and hovering a bar tells you the verdict behind it. Hover any chart and you get the date or the category, every series by name, and the value with its unit, plus its share of the total where that helps. A slice too small to round to a whole percent reads as under one percent, never as zero.

Motion means something here, and there is exactly one moment worth watching: the gauge arc sweeping up to its value while the number counts, with the verdict badge arriving a beat later. Everything else simply arrives. Tiles come in reading order, bars grow, and it is all over inside six tenths of a second. When a number changes, the tile eases to the new one so the change is visible. When a read fails, everything stops: the last good number stays on screen, the tile says so in words, and nothing re-animates, because a number that re-animates looks like a number that just arrived. If you have asked your computer to reduce motion, nothing moves and every number is correct on the very first frame.

The catalog covers the service desk: tickets opened, tickets closed, tickets closed by person, average first response, average resolution time, SLA compliance, the age of what is open right now, and opened versus closed day by day so you can see whether the desk is gaining or losing ground. Response and SLA gauges measure tickets that have closed; a ticket still open past its target shows up once it resolves. Assignment comes straight from your PSA, and a ticket nobody owns is counted as Unassigned rather than credited to anyone. Some desks never assign at all on their busiest boards, most often the one the RMM raises alerts on. When nearly every ticket in the window arrived with no person on it, a per-person breakdown would be a single Unassigned bar and nothing else, which reads either as a fault or as nobody having done any work. So the tile says that in words instead, on a laptop and on a wall screen, and points you at the whole-workspace figure. The same happens to a per-queue breakdown when the tickets carry no queue.

Which of these you are offered depends on the PSA you run, because a ticket number is only ever as good as what your PSA sends Current. Rather than draw a tile that would sit on zero forever, Current leaves that number out of the suggestions and says why if you ask for it.

Projects and sales are covered too: projects over budget, pipeline broken down by stage, deals expected to close this quarter using the same weighted rule as the Forecast report, and how long open deals have been sitting. Sales tiles follow the same access rules as the rest of the CRM.

Present mode fills the screen with your tiles, scaled for reading across the room, and rotates between dashboards when you pick more than one. For a screen with no keyboard, an admin can mint a TV link from the dashboard menu: a private URL any smart TV opens with no login, shown once at creation. Revoke it any time and the screen goes dark within a minute. What that link will show is decided tile by tile. It shows the service desk numbers everyone on your team can see, and it shows the recurring revenue tiles described further down this article, including the ones that name a rep or a partner beside a figure. Everything else stays off it: a cost rate, a per technician breakdown, an open deal and a margin never reach a TV link, whoever minted it. So a board with revenue tiles on it is a decision about where the screen hangs. The figures are the same ones anyone with CRM access already reads in the app, and a lobby is not the same room as a sales floor.

A tile the link will not show does not simply go missing. It leaves a marked placeholder in its slot, so the board keeps its shape and anyone looking at the screen can tell the difference between a tile Current is holding back on purpose and a link that has stopped working. The placeholder says nothing about the tile it stands in for, because the whole reason those tiles stay off a wall screen is that their headings can name a person, a partner or a figure. Open the dashboard in Current to read the number.

Nobody stands at a lobby TV, so the TV links panel tells you how that screen is doing. Beside when the link was last opened it prints how many of the tiles it can show reported a number on that refresh, and names separately how many are held back on purpose, so a full board of gated tiles never reads as a screen in trouble. If the screen keeps asking but has not finished a board in a while, the panel says when it last drew one instead of quoting numbers from that older refresh. A tile that could not be read this minute keeps the last number Current had for it, says so on the tile, and tries again on the next refresh about a minute later.

The screen itself says so too, which matters more, because the panel is only read by whoever thinks to open it. After a few minutes with no successful refresh the wall puts a wide banner across the top: the time the numbers on screen are from, in your workspace's own time zone, and a line saying Current has not been able to refresh the screen. The numbers stay where they are and dim, rather than the board going blank, because old numbers labelled as old are more use to a room than an empty wall. When a refresh gets through the banner disappears on its own with no announcement.

Two things read differently on a wall than on a laptop, and both are deliberate. A TV shows no comparisons against the previous period and no movement arrows on leaderboard rows, so the same board is slightly plainer there. And a workspace running several screens at once shares a small queue: if three of your TVs are refreshing at the same moment, a fourth waits about half a minute and takes its turn inside the same minute. Nothing on that screen changes while it waits.

The screen's orientation matters too. A TV mounted sideways, portrait style, gets its own layout: fewer, wider tiles stacked taller, with full-size gauges instead of squeezed ones. Rotate the screen and the board reflows on its own, without a reload, in Present mode and on the TV link alike. Text on a wall screen never drops below a size you can read from across the room, whichever way the screen is turned.

Wall screens open in dark by default, both in Present mode and on a TV link, because a bright white board in a dim office is the thing people ask us to turn down first. In Present mode there is a sun and moon control in the bar at the bottom of the screen, so whoever is driving can switch that session to light without changing anything on their own desk. A TV link has nobody standing at it, so it takes its instruction from the address instead: add ?theme=light to the end of the link and that screen stays light. Anything else on the end of the link is ignored and the screen stays dark, so a typo can never leave a lobby TV blank.

Dashboards can be light or dark and you pick which. The sun and moon control sits in the dashboards toolbar, on the right, next to the dashboard name. It changes the dashboards page only: your sidebar, your projects, the CRM and every other screen stay exactly as they were. The choice follows you rather than the machine, so switching to dark on your laptop means the board is dark when you next open it on another computer.

Nothing about the numbers changes with the theme. The same tiles, the same figures, the same explanations under each one. Colors were rechecked for readability in both, so the arcs, bars and chart labels stay legible in dark rather than fading into the background, and the light board looks exactly as it did before.

A hero tile is a deep ink card with the figure in big type, its change and its trend line sitting underneath. It is what turns a board into a cockpit: two or three numbers that are plainly the headline, with the supporting tiles reading as ordinary cards below them. Hero tiles look the same in light and dark, so a board reads the same way whichever theme you are in.

By default Current decides by size. Make a number tile large enough and it becomes a hero tile; make it smaller again and it goes back to an ordinary card. That works well until you resize a tile to fit a row and its look changes as a side effect, which is why you can also just say what you want.

Open the tile menu, choose Edit tile, and look for Tile style. It offers three answers. Automatic is the default and behaves exactly as described above: big tiles get the dark hero look. Always hero keeps the dark look whatever size the tile is. Standard card keeps it an ordinary card however big you make it. Once you pick Always hero or Standard card, resizing the tile never changes its look again.

The ready-made dashboards under Start from a set already name their own headline tiles this way, so a starter board opens as a couple of big numbers over dense supporting rows rather than a screen of identical dark tiles. Your choice also survives a refine: ask Current to change a tile and the style you picked stays, the same way your queue and timeframe choices do.

The explain caret is unchanged. About this number still sits under every tile, in both themes and at every tile size, and still opens How it's calculated and How to read it.

Current builds a tile one of two ways, and the difference is worth knowing before you put one on a screen in the office.

First it looks in a vetted catalog: numbers an engineer wrote, named, and tested, covering four families. Service desk (tickets opened and closed, satisfaction when CrewHu is connected). Sales and CRM (open pipeline, won deals, won value, activity, by rep). Projects and time (hours logged, tasks completed, overdue work, budget used, your own hours). Security and devices (phishing caught, open alerts, backups failing, device compliance, warranties running out). A catalog tile is plain: a person wrote that query and it is covered by tests.

If nothing in the catalog fits what you asked for, Current neither refuses nor fakes it. It works the number out from your own data, and it tells you that is what it did.

One bar across the top drives every tile below it: Today, Week, Month, Quarter, and Year, plus trailing 7, 30, and 90 days and any custom range you pick on the calendar. Change it and the whole dashboard recomputes together, so two tiles can never quote different windows at each other. Your choice is remembered for next time, per person.

A tile can also keep a window of its own. Open Edit tile and pick a timeframe there, say Last 30 days, and that tile stays on it while the rest of the board follows the bar at the top. The tile's own scope line prints the window it actually used, so a pinned tile can never be mistaken for one following the bar, and the pin survives a refresh, Present mode and the TV link, where the tile's stamp names its window too. Tiles that only exist as of this moment offer no timeframe control at all.

Every window on a dashboard is cut in your workspace's time zone: today, this week, this month, this quarter. That is set once, in Settings, in the Workspace basics section, and so is the currency every money tile reads in. The zone starts on the one Current already used to time notifications and the lead response clock, so you do not have to touch it. Some numbers did move the day it shipped: the forecast tile and the sales reports used to cut their months and quarters at UTC midnight, so on any workspace outside UTC a deal that closed late on the last evening of a month could sit in the wrong month. It now sits in yours. Only a tenant admin can change the zone, the card shows what the clock reads there before they save, and every change is written to the audit log.

It matters because a day has to mean the same thing to everyone reading the same tile. Before this, a dashboard built its window at the clock of whoever was looking, the forecast tile counted in UTC, and a TV board used the clock of the screen it played on. A project manager in Grand Rapids and an engineer overseas could open one saved tile and read two different numbers, with nothing on the tile to explain why.

So every tile prints its calendar under the number, next to the window it covers, and so does the Reports header and every report you print. A TV board prints the window it is showing and whose calendar it belongs to, along the bottom.

Average first response and average resolution time can be measured two ways, and it is your choice. On the working hours setting, only the hours your desk is open count: nights, weekends and your holidays are left out. On the round the clock setting, it is plain elapsed time. Every workspace starts on elapsed time, so nothing changed for you on the day this shipped.

It matters because every published response time figure in this industry is measured on open hours, including the one Current itself shows on the first response tile, which is 60 minutes. Measure the wall clock instead and a desk that works business hours can never reach it: a ticket that arrives at 5pm on Friday and is answered at 9am on Monday scores 3,840 minutes, when on an eight to five desk it took one working hour. On Autotask it was worse, because Autotask works its own SLA due times out against your SLA calendar, so a green SLA compliance tile could sit right beside a red first response tile off the same tickets.

You set it once, in Settings, in the Workspace basics section: when the desk opens, when it closes, which days it works, or that you work around the clock. Your holidays come from the list you already keep under Holidays, the same one the project scheduler skips, so the two can never disagree about a day off. Each of those two tiles says under its number which of the two clocks it counted on, in the app, in Present mode and on a TV board, so a figure nobody can place is never on screen. Only a tenant admin can change it, every change is written to the audit log, and a schedule that cannot be true, such as a closing time earlier than the opening time or no working days at all, is refused before it can be saved. One schedule covers the whole workspace rather than one per partner.

Break average first response or average resolution time down by person, queue or partner and the board ranks fastest first. Rank 1 is your quickest responder. That is the opposite of every other leaderboard in Current, where the biggest number wins, so the tile says under its number that it is ranked fastest first. It is not a setting: on a number where lower is better there is only one correct order, and asking to reverse it gets you a plain answer saying so rather than a tile that does nothing.

Every row says how many tickets it is an average of, because that is usually the difference between a slow week and a slow technician. A row built on fewer than five tickets is shown below the ranked rows without a rank and says so, so a 0.0 off two tickets can never sit at the top of the board. Nobody is dropped: a name that vanishes off a board is worse than a name with a small number beside it. When the board is showing eight of twenty five, it now tells you how many it is not showing.

Four numbers most managed service providers run their week on now have proper tiles, each built on one server side read and each one naming what it could not measure.

• ＋Reactive ticket hours per endpoint, the number usually called RHEM. The hours your team logged on service tickets, divided by the workstations and servers your RMM manages, reported as a monthly rate. Under about half an hour per endpoint per month is the industry bar. Ask for it by name, or pick the starter suggestion.
• ＋Endpoints per partner. Your managed fleet, per partner, counted from every RMM this workspace has connected.
• ＋Managed services margin. A month of recurring revenue, less the cost of the reactive service desk work that served it. Tenant admins only, because it is built on cost rates. Margin for a single account, rather than the whole desk, is also readable by sales leadership and account executives in that account's renewal dossier.
• ＋Effective hourly rate. The same month of recurring revenue divided by the reactive hours worked in it, so you can see what a flat fee agreement really pays per hour. Tenant admins only.

Reactive means service tickets people asked for. Tickets your monitoring raised automatically are left out, and so are the queues an admin marked as project work under Integrations, so project time never lands in a service number.

A sales gauge counts your sales pipelines: the boards whose type is set to new business, expansion or renewal in the pipeline manager. It does not count every pipeline you own. A projects board, an intake board or an acquisitions board is real work, but folding it into an open-pipeline figure answers a different question than the one the gauge is asking, and it can make the number several times larger than the pipeline your team is actually selling.

Every sales gauge says on its face which pipelines it counted, so you never have to guess at what is in the figure. To add another board, open the gauge, go to the Filters section and tick it in the pipeline list. To go back, click Use my sales pipelines. Leaving it alone is the better default whenever it fits: a gauge that names its own list stops following your workspace, so a pipeline you create next month will not join it until somebody ticks it.

Every sales number in Current values a deal the same way, and it is the same rule the Pipeline and Forecast reports use, so two screens can never disagree about the same deal. Monthly recurring times twelve, plus any one-time amount. Three deal types carry an exception worth knowing:

• ＋A renewal reads its renewed monthly amount, not the new-revenue field, because that is where a renewal's money is recorded.
• ＋An expansion counts only the increase it brought. A partner going from a hundred a month to a hundred and fifty is fifty a month of new money, not a hundred and fifty, so counting the whole figure would credit you with revenue you already had.
• ＋A project deal counts once. Its money is a quote price rather than a monthly bill, so it is never multiplied by twelve.

Ask for monthly recurring revenue and Current draws it from the contracts your PSA already holds. It adds up every recurring contract line that is active today and states each one as a monthly figure, so an annual line is divided by twelve and a quarterly line by three before anything is totalled. That is the same number the company record and the renewal dossier print, which is why they agree. Autotask contracts, ConnectWise agreements and HaloPSA contracts all feed it, so the tile works the same way whichever PSA you run.

Each line is normalised on its own period, not the contract's. A service quoted per year inside a contract billed monthly is still a yearly line, and treating it as monthly would understate that account by a factor of twelve. This is the single most common reason a revenue figure looks wrong somewhere else, and it is the reason Current reads the line rather than the contract header.

One-time work, project work and time and materials are not in it. Recurring means recurring. If you want what you invoiced, that is a different question and a different number, and mixing the two is how a monthly figure ends up being read as a good month rather than as a bigger book.

There is a ready-made board for all of this. Open Add a tile, choose Start from a set, and pick Recurring revenue: it lands with the total, the bridge, retention, churn, and leaderboards by rep and by partner, trimmed to what your workspace can actually draw. The total also sits at the top of the money group on the MSP Scorecard.

A total that moved two hundred dollars is usually two five figure numbers cancelling out. The bridge tile is the answer to that: it starts at what you billed at the beginning of the window, draws each thing that moved, and ends at what you bill now.

• ＋New business. An account that has never had a recurring contract with you before, so a renewal that replaces one agreement with another is never mistaken for a first sale. An account you have only ever billed for one-off or time-and-materials work still counts here the first time it signs an agreement.
• ＋Reactivation. An account that was on your book before, was paying nothing at the start of the window, and is paying again by the end of it. The bar appears only when at least one account came back.
• ＋Expansion. An account that already had recurring revenue and now has more.
• ＋Contraction. An account that already had recurring revenue and now has less.
• ＋Churn. An account whose last active recurring contract ended inside the window.
• ＋Acquired. Accounts that arrived with a business you bought, kept separate so a purchase never reads as organic growth. The bar appears only once at least one account is marked, and the tile says so when none are.

Current works this out by comparing each account's recurring revenue at the two edges of the window rather than by classifying events, and that is what makes a renewal behave. A contract that ends while its replacement starts inside the same window is one account with two contracts, so it lands as whatever it actually changed by, or as nothing at all if it renewed flat. It never appears as a churn and a new sale at once, which is the double count that makes deal based revenue reports hard to trust.

A leaderboard of recurring revenue by rep has to answer a question your PSA does not: which person a contracted dollar belongs to. Current asks you once, when the tile is built, and writes your answer onto the tile where you can see it and change it.

• ＋The account manager on the company. The person your PSA already has against that account. This is the default and it is the best populated, because it is synced rather than typed.
• ＋Whoever won the business. The owner of the won deal behind the account, falling back to the account manager where there is no won deal.
• ＋Split evenly across the account's co-owners. For books that are genuinely shared, again falling back to the account manager.

Accounts that no rule can answer collect in an Unassigned row rather than being dropped, so the board still adds up to the workspace total. A board that quietly loses a quarter of your revenue to a missing field is worse than one that shows you where the field is missing.

Changing the rule changes the leaderboard and nothing else. The total, the bridge and retention are the same numbers whichever rule you pick, because they are about the book rather than about people.

Anyone with CRM access can open these tiles, the same wall that already governs won revenue by rep and the recurring revenue on a company record. Cost rates and per technician cost stay tenant admin only, as they are everywhere else.

Recurring revenue tiles may go on a TV link, and that includes the ones that name a rep or a partner beside a figure. This is a change from how TV links used to work, and it is deliberate: a revenue wallboard with the names taken out is not a revenue wallboard. The rest of the ceiling is unchanged. A tile behind the manager or admin wall, a cost rate, a margin and an open deal never reach a TV link, and a link never gains a permission the person who minted it did not have.

So the decision is about the room, not the software. A screen showing what each partner pays you belongs where your team works, not in a lobby where those partners sit down. Mint the link for the board you mean, and put a board of service desk numbers on the screens that visitors see.

Accounts that came with a business you bought are counted in their own bucket on the bridge, so buying a book never shows up as a record month of new business. Current reads that from your PSA rather than asking you to tag anything by hand: a tenant admin or sales lead points Current at one company field in your PSA under Settings, Integrations, on your PSA's card, and any account with a value in that field is treated as acquired. The value itself is the label, so an account whose field reads Northbridge IT shows as Came with Northbridge IT on its company record.

Nothing is assumed when no field is chosen. The bridge says no acquisitions are marked rather than showing an empty Acquired bar, because no acquisitions and no field configured are different facts and only one of them is about your business. Company fields come across from all three supported PSAs, so the setting works whichever one you run. They arrive on the first sync after you connect, and a field your team typed in Current works just as well. One difference is worth knowing on ConnectWise and HaloPSA: neither tells Current which company fields still exist, only which ones are on the companies it has read. So if you delete a company field in one of those two, Current keeps the field and the values it last saw rather than guessing it is gone — which means companies can go on reading as acquired from it. Pick a different field in this setting, or clear it, when you retire the one you were using.

Current has two views of recurring revenue movement and they answer different questions, so seeing different numbers is expected rather than a fault.

The MRR movement report, in the Sales category of Reports, reads your deals. It is about selling: what your team closed in the period, split into new business, expansion, downgrades and churned renewals, with retention worked out on those deals. It moves when a deal is marked won.

The MRR bridge on a dashboard reads your contracts. It is about the book: what is actually being billed at the two edges of the window. It moves when a contract starts, changes or ends in your PSA.

They separate whenever selling and billing happen at different times, which is most of the time. A deal won in March whose contract starts in April lands in March on the sales waterfall and in April on the bridge. A price rise applied to a contract with no deal behind it shows on the bridge and nowhere on the sales waterfall. An account that quietly stopped renewing shows as churn on the bridge whether or not anybody recorded a lost deal. Each tile names which of the two it is reading, so the number on the screen always says where it came from.

The SLA tile prints its own denominator, in the form "based on 14 of 3,902 resolved tickets": the first number is how many of your resolved tickets carried an SLA verdict from your PSA at all, and the second is how many were resolved in the window. When fewer than twenty carried a verdict, the percentage is hidden and a sentence takes its place, because a green 100 percent off three tickets is not a fact about your desk. Break the tile down by partner, queue or person and the same floor applies to each row, with anything left out counted on the tile face. A TV board applies the same floor too, so a lobby screen can never show a percentage your laptop is hiding.

Every dashboard starts personal. It is yours, nobody else sees it, and you can keep as many as you like: one for your morning check, another for the week you build business reviews.

Project managers and tenant admins can also publish a shared dashboard the whole team sees. A shared dashboard has one layout, so its tiles stay where its editors put them and everyone else reads it as published. Creating, renaming, sharing, unsharing, and deleting all live in the dashboard switcher at the top of the lens. Sharing asks you to confirm first, because it is the one action that widens who can read your numbers.

The dashboard opens no side doors. It mirrors the walls Reports already has, and the database enforces them underneath, not the interface.

Each tile stamps how long ago it last read your data and refreshes itself about every 90 seconds while the page is open. That is how fresh the tile is. How fresh the data underneath is depends on the system it came from, and Current never claims to be fresher than its source. Hover the stamp and it names the cadence behind that tile.

Ask for a number Current cannot reach and it will not guess and will not draw an empty tile. It names the integration that holds the data, says why that data is needed, and points you at Settings → Integrations. Here is the path from there:

1. 1
   Read which integration it named

   A satisfaction tile needs CrewHu. A phishing tile needs Inky. A warranty tile needs ScalePad Lifecycle Manager. Current names the one it needs rather than saying no data.
2. 2
   Connect it

   Settings → Integrations, find that card, and follow it through. Connecting an integration is a tenant admin action, so if that is not you, send them the name of the card.
3. 3
   Give it one sync cycle

   A freshly connected integration has nothing in Current until its first sync lands. Use the table above for how long that is, or press Sync now on the card where it offers one.
4. 4
   Ask again

   The sample prompts and everything Current will build widen on their own the moment a connection is live. Nothing to turn on for the dashboard itself.

A safety tile will never tell you everything is fine when Current does not know. Backups failing, EDR alerts open, SOC incidents and phishing caught are the four that matter most here, and each one now says which of these three it is hitting:

1. 1
   Waiting on the first data

   The product is connected but has never sent Current a record yet, so the tile says it is waiting on the first data from that product and names it. It fills in on its own with the first sync. This is the difference between a real zero and an empty feed, and Current keeps them apart by remembering the first record each product ever delivered.
2. 2
   The read failed

   A tile that already had a number keeps showing it, prints "Couldn't refresh this. Showing the last number Current read." underneath, and lets the Updated stamp in its header keep aging so you can see how long it has been. A tile that never got a number says it could not read this one. Either way it retries by itself about every 90 seconds.
3. 3
   Only part of it could be read

   On a very large fleet a tile can hit its reading limit. It then says the number would be incomplete rather than showing a smaller one, and points you at Settings, under Integrations, to check that the product is syncing.
4. 4
   The window was never measured

   SOC incidents over the last 90 days goes further, because a zero there reads as an all clear. Some RocketCyber tokens refuse the 90-day history request outright while the open-incident counts keep arriving. Until Current has completed one 90-day read for your workspace, a zero across every account with incidents open right now is a question nobody has answered: the tile says there is no figure it can stand behind rather than printing a zero, and names the refusal when RocketCyber has told Current about it. Once that read has run, a quiet 90 days shows a real zero, with the number of older incidents still open printed beside it so nobody takes it for an all clear.

Devices compliant follows the same rule and goes one step further. If any RMM you have connected cannot be read, it refuses the whole percentage instead of publishing the RMMs that did answer, because a Mac-only 100 percent over thousands of unmeasured Windows machines reads as good news. It also prints how many devices actually reported an answer, and says on its face when it is mixing Windows patch compliance from Datto RMM or ConnectWise RMM with Mac policy compliance from Addigy, since those are two different checks.

A gauge can be arithmetically right and still mislead, in three ways that used to be invisible. Each one now prints a line under the number rather than leaving you to guess. None of them stops the tile drawing, and none of them can be collapsed out of the way, because a wall screen has nobody to click the caret.

1. 1
   The feed says connected and has gone quiet

   If your PSA has not delivered anything for six hours or more, every tile fed by it says how long ago Current last heard from it, in hours up to two days and in days after that. The number is real; it just counts what had arrived by then. A connector that looks green in Settings can still be hours behind, and this is the only place that shows.
2. 2
   The timeframe starts before your history does

   Current can only count what it holds. A ticket tile set to a period that ends before the oldest ticket in the mirror refuses instead of printing a confident zero, and gives you the date to move the timeframe to. A period that starts earlier and ends inside your history still draws, with a line saying the earlier stretch is empty rather than quiet, so a year long chart with four flat months at the start is not read as a terrible spring.
3. 3
   Logged hours have their own start date

   Reactive hours per endpoint, effective hourly rate and managed services margin read your PSA’s ticket time, which usually begins the day the connection was switched on, not the day your tickets do. Those tiles are measured against the oldest logged hour Current holds and say so by name, so a period before it is refused rather than answered with zero hours, a hundred percent margin and an infinite hourly rate.
4. 4
   The check itself has not landed

   Before a board draws, Current works out what this workspace can report on. If that check is slow or fails, the tiles still draw and each one says the number was drawn without it, while the check retries behind you. The line clears itself the moment the check lands. It never appears over a check that has already answered.

A failed read and a genuinely quiet week used to look identical on a money tile as well. Open pipeline value, Deals won, Won revenue, Won revenue by rep, Quotes won and CRM activities logged all went quiet when their read timed out or was refused, so a tile could read as though nothing was won while the read behind it was failing. They now report the failure instead, and the whole tile library follows one rule: an empty tile means the query ran and matched nothing. Anything else says what went wrong.

• ＋The could-not-refresh line described above is not only for safety tiles. Any tile that keeps a number through a failed read prints it, in amber, and it now shows in the app, in Present mode and on a TV screen, because a wall screen is where a frozen number does the most damage.
• ＋A tile with no earlier number to keep says it could not read this one, instead of showing a zero.
• ＋No tile puts database wording on its face. A read that takes too long says it took too long and that Current will try again in a moment; the technical detail goes to the browser console, where an engineer can find it.
• ＋An empty tile is worded for the kind of tile it is. A right-now tile no longer says there was no data in this window, because it has no window: it says there is nothing to show right now and points you at refining what it counts. A tile with a timeframe says nothing matched that timeframe.

Two things about a ConnectWise service ticket were read wrongly, and correcting them moves ticket numbers in both directions at once. Autotask and HaloPSA workspaces are not affected.

1. 1
   A closed ticket now counts as closed

   Current used to treat a ticket as closed only when ConnectWise had stamped a close time. Any ticket ConnectWise considered finished without one stayed open forever, so Tickets open now and all four aging buckets read too high, Tickets closed read too low, and the opened versus closed chart showed a backlog that only ever grew. Current now trusts ConnectWise's own closed flag. Where there is no close time it uses the ticket's last update as the closing moment, and the tile says so under the number, because a resolution time built that way is an upper limit rather than a measurement.
2. 2
   Tickets closed by person now measures the person it was dispatched to

   It used to measure the ticket owner. ConnectWise desks dispatch work by assigned resource and often never set an owner, which is why that leaderboard was mostly Unassigned. It now reads the dispatched resource and takes the name from your own resource list, so bars move, some names appear for the first time, and the owner is used only on a ticket with nobody dispatched. A ticket dispatched to somebody Current has not synced yet stays under Unassigned rather than being credited to the wrong person.
3. 3
   Current re-reads the last 24 months once

   Tickets already stored have to be re-read for the corrections to reach them, so Current walks the last 24 months of ConnectWise tickets one time, a chunk at a time across your normal syncs. Numbers settle over a few hours to a few days depending on how many tickets you have. Nothing is deleted and nothing needs switching on.
4. 4
   The dispatched person now actually arrives

   ConnectWise's ticket list does not carry who a ticket was dispatched to, so for a while Current asked for it and got nothing back, and every ConnectWise ticket was stored as unassigned. Current now asks for that one field separately and matches it against your own resource list, so a per-person board on a ConnectWise desk shows real people instead of a single Unassigned bar. Tickets already stored are corrected by the same 24-month re-read described above. If ConnectWise will not answer the extra request, Current leaves your assignees exactly as they are rather than reporting that nobody was on them.
5. 5
   A board says how much of it is unassigned

   Plenty of desks assign some work and not the rest, and a board where half the tickets carry nobody still draws a tall Unassigned bar at the top that reads like a person's score. When a quarter or more of the tickets behind a board arrived with no assigned person, or with no queue, the board still draws and now says the share on its face, on a laptop and on a wall screen alike. Above about nineteen in twenty it stops drawing altogether and says why, as it did before.
6. 6
   Two records, one name

   A PSA sometimes carries the same company as two records, and Current mirrors each one, so a tile narrowed to that partner counts the tickets on one of them. The tile now says when another record carries the same name and its own tickets, and it says it on a wall screen too. It is matched on the name alone, so it can also be two genuinely different businesses that share one - you know which. Merging the two records in your PSA is the permanent fix for a real duplicate.

Connecting something is not always the answer, and Current will not pretend it is. Some numbers are not in Current at any price yet, because the data behind them never arrives from anywhere: technician utilization, first contact resolution, escalation rate, quote win rate, blended gross margin, revenue per technician. Ask for one of those and Current says in one plain sentence what it does not hold and why, rather than sending you to Settings to connect an integration that would not have helped. Reactive hours per endpoint used to be on this list and is not any more: Current now mirrors your PSA's service ticket time, so it draws that number for real. See The four MSP gauges above.

Then it offers the nearest numbers it can genuinely draw, as buttons you can click straight into the box. Every suggestion is checked against your own permissions and what your workspace has connected before it is offered, so what you are shown is a tile Current will really build. If none of them clears those walls for you, Current says so instead of showing an empty row.