Your dashboards are beautiful and useless

I have a confession. Over the past eight years, I’ve built somewhere around 200 custom dashboards. Looker Studio, Tableau, Power BI, custom React apps with chart libraries I can’t even remember the names of. Beautiful work. Color-coded. Responsive. Interactive filters. Drill-down capabilities.

Most of them are dead.

Not broken. Just abandoned. Sitting in shared drives and bookmarked folders, pulling data from APIs nobody checks on anymore. I know this because I’ve gone back and looked at the access logs. Out of the last 40 dashboards I built for clients in 2024 and 2025, exactly 7 were still being opened weekly after six months. Seven.

That’s a 17.5% survival rate. Which means I spent roughly 82% of my dashboard-building time creating expensive digital wallpaper.

So I started asking myself: what’s different about the ones that survived?

The dashboard graveyard

Here’s the pattern. A client comes to me and says “we need a dashboard.” I ask what they want to see. They say everything. Revenue, sessions, conversion rate, bounce rate, average order value, pages per session, new vs returning users, top landing pages, exit pages, device breakdown, geographic distribution, campaign performance by channel and sub-channel, product category performance, search term analysis…

You get it. They want the entire analytics platform rebuilt in a prettier skin.

So we build it. Fourteen tabs. Fifty-something charts. Every metric anyone could possibly want. The client sees the first draft and gets excited. “Can we also add…” Yes, we can add that. We always can.

Launch day arrives. Stakeholders gather around a conference room screen. Everyone nods approvingly. The CMO says “this is exactly what we needed.”

Two weeks later, three people are using it. By month two, it’s just the analyst who helped build it. By month four, nobody.

What went wrong? The same three things that always go wrong.

Problem 1: No question, just metrics

A good dashboard answers a specific question. “Are we on track for Q2 revenue targets?” is a question. “Show me all the data” is not a question. It’s a request for a database browser with better fonts.

I worked with an e-commerce brand last year that had a 22-tab Looker Studio report. Twenty-two tabs. I asked the head of marketing what question she used it to answer. She paused for about ten seconds and said “I don’t really use it, I just go straight to GA4.”

She went straight to GA4 because at least there she could type a question into the search bar. The dashboard gave her everything except what she actually needed at any given moment.

Problem 2: No context, just numbers

A number without context is meaningless. Revenue is $340,000 this month. Is that good? Bad? Expected? I have no idea unless you show me what it was last month, what the target is, and what the same month looked like last year.

I see this constantly. Someone builds a dashboard full of big number cards. Revenue: $340K. Sessions: 890K. Conversion rate: 2.3%. These are facts, not insights. They tell you what happened without any framework for whether you should care.

The dashboards that survive always include comparison. Always. Period vs period, actual vs target, this year vs last year. Preferably two of those three.

Problem 3: Nobody owns it

Dashboards rot faster than you’d expect. A UTM convention changes, a new product category gets added, someone renames a campaign medium from “email” to “Email” and suddenly half the filters break. Without an owner who notices these things — someone running a regular analytics audit — a dashboard goes from “source of truth” to “that thing with the wrong numbers” in about six weeks.

The seven dashboards that survived from my batch? Every single one had a named person responsible for maintaining it. Not a team. A person.

Three dashboards that actually work

After enough failures, I started noticing what the survivors had in common. They fell into three categories. Every organization needs exactly these three, and probably not more.

1. The Monday Morning dashboard

This is the one the marketing director opens at 9:15 AM on Monday with a coffee. It answers one question: “How did last week go, and are we on track?”

Here’s what goes on it. Five KPIs, max. For most businesses that’s revenue (or leads), sessions, conversion rate, cost per acquisition, and one business-specific metric like average order value or trial-to-paid rate. That’s it. Five.

Each KPI shows the current value, the previous period value, and the percentage change. Color-coded: green if it improved, red if it dropped, gray if it’s within normal variance (I use plus or minus 5% as the threshold, but you should calibrate this to your business).

Below the KPIs, one chart: a weekly trend line for the past 12 weeks. Not daily, weekly. Daily is noise for this audience. They want to see if the trendline is going up, down, or flat.

One filter: date range. That’s the only interactivity. No channel breakdown, no device split, no geographic drill-down. Those belong somewhere else.

The entire dashboard fits on one screen without scrolling. This matters more than you think. The moment someone has to scroll, engagement drops. I’ve watched people in user testing sessions stop at the fold like it’s a wall.

Build time: about 4 hours in Looker Studio, including data source setup. That’s it. Four hours for a dashboard people actually use every week for years. Compare that to the 40-hour monster that dies in a month.

2. The Something Broke dashboard

This one exists for a different audience and a different moment. It’s for the analyst or the ops person who gets a Slack message saying “our ads aren’t tracking” or “why did traffic drop 40% overnight.”

The Monday Morning dashboard won’t help here because it updates weekly and shows trends. This one needs to be near-real-time and show anomalies.

What goes on it: hourly session count for the past 48 hours, with a reference band showing the expected range based on the same day/hour from the previous four weeks. If the current line drops below the band, something’s probably broken. If it spikes above, something interesting happened (or your bot filter is down).

I add event firing rates for key conversion events. If your purchase event was firing 50 times per hour yesterday and now it’s at 3, your tracking broke. This catches tag manager deployments gone wrong, consent tool updates that blocked scripts, and site changes that broke data layer pushes.

A simple server response code breakdown is useful too. A spike in 500 errors explains a lot of traffic drops.

I set up alerts on this one. Email or Slack notification when traffic drops below 60% of the expected range for two consecutive hours. Two hours, not one, because single-hour dips happen all the time and alert fatigue is real. I learned this the hard way after setting up instant alerts that pinged the team at 3 AM because of a CDN cache refresh.

Build time: about 8 hours, more if you’re setting up BigQuery streaming exports for real-time GA4 data. Worth it, though. This dashboard has literally saved clients from running blind for days.

3. The Board Meeting dashboard

This is the most misunderstood one. Executives don’t want dashboards. They want stories. They want to know: did the thing we invested in work, and what should we do next?

Stop putting interactive Looker Studio reports in front of your C-suite. I’ve watched a CEO try to use a date filter during a board presentation and accidentally set the range to a single day, making all the numbers look catastrophically low. The room panicked for thirty seconds until someone fixed it.

What works instead: a static, curated report. Yes, static. Export it to PDF or build it in Google Slides. Pick the three things that matter most this quarter. Show each one with a chart that has a clear annotation explaining what happened.

For example: “Revenue grew 12% QoQ. This was primarily driven by a 23% increase in paid search conversions after we restructured campaigns in February. Organic growth was flat.” One chart. One annotation. One insight.

The board doesn’t care about your bounce rate. They don’t care about pages per session. They care about revenue, margin, customer acquisition cost, and growth trajectory. Give them four slides with those four things, each with a single chart and two sentences of context.

Build time: about 2 hours per quarter. But those two hours require an analyst who understands the business, not just the tools. The hard part isn’t building the chart, it’s knowing what story to tell.

Design principles I’ve learned the hard way

One question per view. Every screen on your dashboard should answer exactly one question. If you can’t articulate the question, delete the screen. I literally write the question as the tab title now. “Are we on track this week?” is a better tab name than “Overview.”

Comparison is mandatory. Never show a metric without something to compare it against. A number alone is trivia. A number with a trend is information. A number with a trend and a target is actionable.

Annotations save dashboards. Put text on your charts explaining what happened. “Launched new landing page Feb 12” next to the conversion rate spike. “iOS 18.3 update” next to the traffic dip. Without annotations, every anomaly becomes a mystery that someone has to investigate. With annotations, people trust the dashboard because it acknowledges reality.

Build for the viewer’s worst day. Your dashboard will be used by someone in a rush, probably on a laptop with a small screen, probably distracted by twelve other things. They will not hover over tooltips. They will not click filters. They will not read your clever footnotes. Make the main point obvious at a glance.

Kill the dashboard before it dies naturally. Set a review date. Every quarter, check the access logs. If nobody opened it in the past 30 days, archive it. Don’t let zombie dashboards accumulate. They erode trust in your whole reporting infrastructure because someone eventually stumbles across one with outdated numbers and assumes all your data is wrong.

The uncomfortable truth

Here’s what I’ve come to accept after years of this: most dashboards fail because they’re built for the wrong reason. They’re built because someone asked for a dashboard, not because someone had a question they needed answered repeatedly.

The best reporting setup I’ve ever built for a client was three dashboards and a monthly narrative email. Total build time was about 20 hours. It replaced a system of eleven dashboards, four automated reports, and two weekly manual exports that took an analyst half a day to compile. Nobody missed the old system.

Fewer dashboards, used more. That’s the whole strategy. It’s not exciting and it won’t win any design awards, but it works. And in analytics, working is the only metric that matters.