Looker Studio Marketing Dashboard: Build One, Not Five

A marketing dashboard is worth building when it answers a fixed set of recurring questions with numbers whose definitions never move. Start from the three or four questions you are actually asked every month, put one number against each, and add nothing else. The work that makes it survive is not chart design, it is agreeing what each number means and where it comes from before anything is plugged in.

The reason the monthly report gets rebuilt by hand is rarely that nobody knows how to connect a data source. It is that the report is a pile of screenshots representing whatever each tool happened to show, with no shared definition holding them together. Automating that pile produces a faster pile.

This page is about the decision, not the button clicks. Looker Studio, like every product in this stack, moves its interface on its own schedule, so what follows describes what has to be true rather than which menu it currently lives under.

Start from the questions, not the connectors

Write down the questions you are asked, in the words the person asking uses. Most in-house marketing functions have four or fewer.

  • Are we spending what we planned to spend?
  • Is it producing more or less than last month, and than the same month last year?
  • Which channel or campaign is responsible for the change?
  • Is anything broken right now?

Everything on the dashboard has to serve one of those. A metric that serves none of them is decoration, however easy it was to add. This test does real work, because the temptation with a connector is to place every field it offers, and a page with forty numbers answers nothing.

Being specific about the question also settles the format argument. "Are we spending what we planned" is a single number against a target, not a time series. "Which campaign is responsible" is a table sorted by change, not by size, because the sorted-by-size table is the same every month and tells you nothing new.

The part templates skip: making two sources agree

Every gallery of dashboard examples shows finished charts. What they do not show is the joining problem, and it is the reason most first attempts get abandoned.

Two sources placed side by side on one page look like they belong to one system. They do not. Ad platform numbers and analytics numbers count different populations, file conversions under different dates and credit touchpoints under different rules, which is why key events and conversions are two objects with two rulebooks rather than two names for one thing. Put them in adjacent boxes with no labels and you have built a machine for generating arguments.

Three habits prevent that.

Label the source and the clock in the header, not the footnote. "Conversions, Google Ads, click date" and "Key events, GA4, event date" are the column names. Ugly, and it ends the recurring question.

Never place a total that sums across platforms. A blended conversion total invites double counting, because a single buyer can be claimed by more than one platform. If you need one headline figure, take it from the system that issues invoices.

Pick one channel grouping and write down its rules. Otherwise a campaign lands in three different buckets in three different boxes. Any channel breakdown you publish will also carry an unattributed bucket, and you should show it rather than hide it: what sits inside it is described in what direct traffic in GA4 actually is, and a dashboard that quietly drops it is overstating everything else.

What belongs on the page, and what does not

The useful division is not important versus unimportant. It is decision versus diagnosis.

Belongs on the shared dashboard Belongs in a tool, not on the dashboard
Spend against plan Search term detail
Result count and cost per result, defined once Auction and competitive metrics
Change against the previous period and the same period last year Anything requiring a filter to be meaningful
Channel or campaign contribution to that change Anything nobody has ever asked a question about
A visible flag when a data source has stopped updating Every field the connector offered

Diagnostic work does not belong on a shared page because it needs interaction, and a dashboard is read passively by people who will not filter it. The search terms report is the clearest case: it is one of the most useful screens in a Google Ads account and one of the worst dashboard tiles, because its value is in reading and acting on the long tail, not in a count. The same is true of impression share lost to budget, which answers a specific question about constraint at a specific moment and means very little as a trend line watched by people who cannot act on it.

The one exception worth building is a health tile. A small panel showing the most recent date each source returned data catches the failure mode that quietly destroys trust: a dashboard that keeps rendering a stale number long after a connection has broken.

Freshness, and why the dashboard is not showing today

A predictable complaint arrives about two weeks after launch: the numbers do not match what someone just saw inside the ad account.

Part of that is caching. Looker Studio's documentation states that "Each type of data source has its own default data freshness threshold, but you can adjust this threshold as needed", and the default interval documented for Google marketing product connectors is twelve hours (Looker Studio documentation, "Manage data freshness", checked 2026-09-07). So the page can legitimately be half a day behind the platform. Check the setting on your own data sources rather than assuming, because the adjustable range differs by source type and some are fixed.

Part of it is not caching at all. Conversion figures for recent days keep filling in after the fact, because ad platform conversion columns are filed against the click date. Yesterday is always provisional. Both effects together mean a dashboard is a poor instrument for looking at today and a good one for looking at the last complete week, and it is worth saying that on the page itself.

When one dashboard is the wrong answer

The honest limits, before the recommendation to build one.

Below a certain volume, a dashboard is theatre. If a channel produces a handful of results a month, a chart of it is mostly noise given a shape. A short written note beats a trend line at low volume, and the numbers you would have plotted still belong in a spreadsheet.

A dashboard cannot carry a caveat. Anything that needs "but" attached to it needs a sentence from a person. Attribution is the standing example: which model a report uses changes who gets credit, and the choice is a judgment rather than a setting, as attribution models in plain language sets out. A dashboard shows one model's answer with no indication that a different one was available.

Somebody has to own it. A dashboard has a maintenance cost: connections expire, campaign naming drifts, a platform renames a field and a tile goes blank. If nobody is responsible for it, it will be wrong within a quarter and people will go back to screenshots, only now with less trust.

It does not replace the commentary. The dashboard answers what happened. It cannot say why, and the meeting was always about why. The realistic goal is to stop rebuilding the numbers by hand so the time goes into the explanation instead.

What this page will not do

It will not tell you which tool to use, rank dashboard products, or quote what any of them cost. We have no basis for a ranking, prices date immediately, and a recommendation made without knowing your data sources would be guesswork dressed as advice.

Nor will it give you a template with fixed metrics. The set of numbers that belongs on your page depends on what your organization asks and what it can act on. A downloaded template is somebody else's answer to somebody else's questions, and its main effect is to fill a page with tiles nobody chose.

Everything product-specific above was checked against Google's own documentation on 2026-09-07. This area changes, so treat the freshness detail in particular as accurate on that date and verify it in your own report before relying on it.

FAQ

What should a marketing dashboard actually show? One number per recurring question, with its source and date basis labeled. For most in-house teams that means spend against plan, results and cost per result, the change against a comparable period, a channel or campaign breakdown of that change, and a health indicator showing when each source last updated.

Why does my Looker Studio dashboard not match my Google Ads account? Two ordinary reasons before anything is wrong. Cached data, since each data source type carries its own freshness threshold and the documented default for Google marketing connectors is twelve hours. And provisional recent days, because conversions are filed against the click date and keep being added after the fact.

Should I put GA4 and Google Ads conversions on the same page? Yes, but never in a combined total and never without labels. They count different populations on different clocks. Label each with its source and its date basis, and do not add them together.

How many metrics is too many? If a number on the page has never caused anyone to do anything differently, remove it. In practice a page that cannot be read in about a minute will not be read at all.

Is it worth automating if I only report monthly? Sometimes not. If the monthly rebuild takes an hour and the definitions are stable, automation saves an hour a month and costs maintenance. The stronger reason to build one is consistency, not speed: the same definitions every month, so month-to-month comparisons mean something.

Sources, read 2026-09-07

Looker Studio documentation: "Manage data freshness" (Google Cloud documentation for Looker Studio). Google Ads Help: "Understand your conversion tracking data" (support.google.com/google-ads/answer/6270625), for click-date reporting of conversions.

Interfaces and connector behavior change. This page describes what a dashboard has to do rather than where the buttons are, and carries a six month review date.

Similar Posts