Dashboards that drive decisions

0:00 Dashboards that drive decisions

I've seen a dashboard with forty-three charts. Nobody could tell me what decision it supported. It was beautiful, it was expensive, and it was ignored. In this lecture you'll learn to build dashboards that drive decisions: how to design for a specific audience, the tools and connectors in twenty twenty-six, including Google's Data Studio, which you may know as Looker Studio, the design principles that make dashboards readable, how to keep them fast with BigQuery, and a one-page spec you can use for every build.

0:37 Why it matters

Why does this matter? Because a dashboard is a decision tool, not a data dump. The best ones answer a small set of recurring questions for a specific audience, at a glance, and show when something needs attention. If nobody would act differently after seeing a chart, it doesn't belong. Dashboards are also how your work is judged. A clear dashboard with honest commentary builds trust. A confusing one makes people doubt the data, even when the data is right.

1:12 Audiences

Start with the audience. Leadership asks: are we on track, and where should we invest? They need four to six K P Is versus target, trends and a channel summary, weekly or monthly. Channel managers ask what's working in their channel, daily or weekly, with spend, key events, cost per key event and creative breakdowns. Content and product teams need landing pages, engagement and funnels. And agency clients ask: is my money well spent? Outcomes, commentary and next steps, monthly. One dashboard can't serve all of them well.

1:50 Tools in 2026

Now tools. Google's free dashboarding tool is Data Studio. It was called Looker Studio from twenty twenty-two until Google renamed it back to Data Studio in April twenty twenty-six, and Looker Studio Pro became Data Studio Pro. Existing reports and links carried over automatically. It connects natively to G A four, Google Ads, Search Console, Sheets and BigQuery. Other tools, like Power B I, Tableau or Metabase, can use G A four through connectors or the BigQuery export. The concepts are the same everywhere: data sources, calculated fields, blending, and filters and controls.

2:30 Quotas and speed

One practical issue: quotas. The G A four Data A P I, which powers the live connector, has quotas. Dashboards with many charts and many viewers can hit them, especially on Monday mornings when everyone opens the report, producing errors instead of numbers. Mitigations include fewer charts per page, cached extracts, or routing through a small pre-aggregated BigQuery table. For any dashboard with more than a handful of regular viewers, the BigQuery route is usually the grown-up option.

3:04 Design principles

Here are the design principles. Top-left is prime real estate, so put the most important K P Is there. Every K P I needs context: versus target, previous period or last year. Show rates and totals together. Use consistent colors and definitions, with a glossary page. Minimize chart types: scorecards, line charts for trends, bar charts for comparisons, tables for detail. And annotate campaign launches, site releases and tracking changes. Think of it like a car dashboard again: few dials, clear warning lights, nothing decorative.

3:41 The commentary box

And the most important element on any dashboard is the commentary box. What happened, why, and what we're doing next. Numbers without narrative leave stakeholders to invent their own explanations, and they usually invent the wrong ones. Clients read the commentary first. The charts support it. If you only have ten minutes before a meeting, spend them on the commentary, not on another chart.

4:09 Smarter, not busier

Three techniques make dashboards smarter without making them busier. Calculated fields, like key events divided by sessions, or a CASE statement that groups messy campaign names into clean families. Blending, which joins sources, like ad spend from each platform with G A four key events, matched on date and campaign, so you can show cost per key event. And controls, like date, country and channel selectors, so one page serves several questions. Label every blended number with its sources, because a cost per acquisition built from two systems will never match either system exactly.

4:50 Example 1: an agency client dashboard

First example, a simple one. A small agency in Manchester manages paid social and search for an online florist. Its monthly client dashboard has four pages. Outcomes: orders, revenue and cost per order versus target, with spend blended from ad platforms. Channels: paid social, organic search and email, each with sessions, key event rate and revenue. Search: Search Console clicks and top landing pages with key events. And a glossary and data notes page with definitions, consent caveats and the attribution model used. Each month, three commentary bullets and one recommendation.

5:30 Example 2: Karachi agency

Second example, a business case. A Karachi agency's client dashboards were built on the live G A four connector. Every Monday morning, when fifteen account managers opened them at once, charts failed. The team moved each client to a nightly BigQuery summary table plus a Data Studio report per client, with a data freshness note saying updated daily at seven. Monday failures stopped, load times dropped, and they could finally blend ad spend and C R M outcomes on the same page.

6:06 Watch me do it, part 1

Watch me build the summary table. In BigQuery, I create or replace a table called daily channel K P Is. For each date and channel, where channel comes from the session traffic source last click record's default channel group, I count session starts as sessions, count purchases and generate lead events as key events, and sum purchase revenue. I cover the last four hundred days. Then I schedule it daily. And I note on the dashboard that these export-based counts will differ from the G A four interface, which uses different identity and modeling rules.

6:47 Watch me do it, part 2

Then I write the spec before building a single chart. Audience: leadership. Three questions: are we on track versus target, which channels moved, and what are we doing about it? K P Is with comparisons: revenue versus target, key events versus previous period, key event rate versus last year, cost per acquisition versus target. Filters: date, country and channel. Sources, with the BigQuery table, ad spend exports and C R M. Refresh time. A definitions page. And an owner with a monthly review.

7:23 Common mistakes

Common mistakes. Thirty charts on one page. No targets or comparisons, so nothing signals good or bad. Mixing attribution sources without labeling them. Dashboards that silently break when quotas are reached. Metrics named inconsistently, like users on one page and visitors on another. And no commentary, which leaves the most important part of the dashboard empty.

7:47 Recap

Recap. A dashboard is a decision tool for a specific audience. Every K P I needs context. Label every source and attribution model. Use Data Studio, formerly Looker Studio, or your BI tool of choice, and route busy dashboards through BigQuery summaries. Write a spec first, and never ship without a commentary box.

8:10 Try this now

Try this now. Sketch a one-page dashboard for a real audience using the wireframe and spec from the lesson. List each K P I, its comparison, and the question it answers. Write a sample commentary box for last month. If a chart doesn't connect to one of your three questions, remove it.