Upbuild
Written by Uptimal, the team that builds and operates Upbuild and runs marketing analytics for multinational clients across the globe. Everything below reflects how the product actually works, including the parts that are deliberately boring.
Tuesday morning, and the analytics lead of a multi-market group has just got the answer she needed: spend and orders by market for the month, with GA4 orders next to platform conversions, in one table. It is right. She checked. And now comes the moment every analyst knows: the answer is about to be pasted into a slide, where it will be wrong by Thursday and impossible to question for ever after.
In Upbuild the path from that answer to a standing report is one sentence long. Here is what happens along it, and when it is worth taking.
Ask, then check the numbers
It starts in AI Insights, the shared chat on every account the organisation has shared with her.
"Spend and orders by market across EMEA, this month against last, with GA4 orders next to platform conversions."
The answer arrives in seconds as a table and a chart. Before anything goes near a dashboard she does what she would do with an analyst's work: she checks it. Every figure comes straight from the platform, and account names and numbers are copied from the data, so she can open Google Ads and see the same number. "Show me the campaigns behind Germany" gives her the rows under the total. "Is last month's UK figure complete?" gets a plain answer: yes, or here is what is missing and where.
This is the step decks skip. A chart pasted into a slide cannot be asked anything. It can only be believed.
"Add this to the EMEA weekly dashboard"
Satisfied, she types one more line:
"Add this to the EMEA weekly dashboard."
The table becomes its own tile on the dashboard, next to the charts the region already reads every Monday. Upbuild AI adds it as the next numbered version of the dashboard, and she opens that version first: the tile sits where she wants it, the title reads the way the region will understand it, the comparison windows are labelled. Then she publishes. People with the link only ever see what is published, so nobody saw a half-finished page.
Nothing was rebuilt. The query that answered the question is the query that now feeds the tile, so the number on the dashboard and the number in the chat cannot drift apart.
Dashboards, exports and alerts, built by asking.
Upbuild is an AI platform for marketing data: AI Dashboards, AI Insights, AI Data Exports and AI Alerts, with a read-only MCP server as part of its toolset that brings your accounts into Claude, ChatGPT, Copilot Studio and Gemini Enterprise. Every tool, live data. Free for 14 days on the connector and 7 days on the platform.
Get a demo Start free →The tile refreshes with the dashboard
From here the tile belongs to a dashboard that looks after itself. With AI Dashboards every dashboard refreshes on the cadence the team set, every 30 minutes, every hour or once a day, in the team's timezone, whether anyone opens it or not. The date range in the question was stored as a rolling window ending yesterday, so "this month against last" in September still means this month against last in November.
Every page shows when its numbers were last updated. If a platform is unreachable during a refresh, the tile keeps its previous rows and the dashboard shows a banner instead of a blank chart, and the admins get an email, once a day, saying what happened and when the last good data was. Reconnecting takes seconds and changes nothing she built.
Version saved, link shared
Adding the tile created a numbered version, with who made the change and why. Any version can be restored with one click, so if the region decides in October that the old layout was better, October gets it back without anyone rebuilding anything. The page and its data queries version separately, which means restoring a layout never quietly undoes a data fix somebody shipped in between.
Sharing is a link. The regional leads open the EMEA weekly dashboard with no Upbuild account, no seat and no login to Google Ads, Meta or GA4. The link can carry a password, shows only the published version and can be switched off at any time, and each dashboard lives on its own secure address. For a client rather than a colleague, the same link is the whole hand-over: see sharing a dashboard with a client.
And because dashboards belong to the organisation and not to a login, the EMEA weekly keeps refreshing after the analytics lead changes jobs. An admin removes her, her access ends, and the dashboard carries on for whoever takes over.
Why this beats copying a chart into a deck
A chart in a deck is a photograph of a number. It was true at the moment it was taken, it cannot be asked a follow-up, and it will be presented as current for as long as the deck circulates. Someone always has to rebuild it next month, usually by hand, usually slightly differently.
A tile is the number. It refreshes on a schedule, it is versioned, it can be opened by anyone with the link, and the chat can be asked about it next week: "Why did the UK tile move?" One source feeds the chat and the dashboard, so the Monday meeting stops arguing about which slide is right and starts arguing about what to do.
There is still a place for a deck, and for a BI tool such as Looker Studio when many stakeholders need governed self-serve exploration over a modelled warehouse. But most of what marketing teams paste into slides every week is a tile nobody had time to build. Now building it is one sentence.
When to keep it as a question, and when to promote it
Not every good answer belongs on a dashboard. A dashboard is a page people read every week without asking. The chat is where people find things out. The test is simple.
Keep it as a question when it is an investigation: why Spain moved, whether a launch is working, what happened on the day the tag broke. Those answers matter once, and the working stays in the chat for anyone who needs to see how the conclusion was reached.
Promote it when it has been asked twice, when someone else needs it without you in the room, when a client or a region will read it every week, or when a decision hangs on that number on a fixed day. "Which markets are over budget?" is a question in March and a tile by May, and by June it is also a rule in AI Alerts that sends the answer before anyone asks.
A useful habit is a short review at the end of each month: which tiles were opened, which questions were asked twice, and which answers from the chat are now being pasted into slides by hand. The pasted ones are the next tiles.
And retire tiles. A dashboard where every answer anyone ever liked was promoted is a deck with a refresh button. The versions make this safe: remove a tile, and if the region misses it, the version with it in is one click away.
Frequently asked questions
Do I have to rebuild the chat answer as a chart before adding it to a dashboard?
No. The table or chart in the answer becomes its own tile, fed by the same query that answered the question, with one prompt such as 'add this to the EMEA weekly dashboard'.
What happens to the tile when the data changes?
It refreshes with the dashboard, every 30 minutes, every hour or once a day. Date ranges are stored as rolling windows ending yesterday, and if a platform is unreachable the tile keeps its previous rows and the dashboard shows a banner.
Who sees the dashboard after I add a tile?
Anyone the dashboard is shared with, through a view-only link that needs no Upbuild account and no platform login, with a password if you want one. They only ever see the published version, never a draft.
Can I undo an added tile?
Yes. Every change is a numbered version with who made it and why, and any earlier version can be restored with one click. The page and its queries version separately, so restoring a layout never undoes a data fix.