There is a low tolerance for dashboards that look like a spaceship and explain nothing about money.
Sessions up. Bounce rate down. Engagement looking spicy. Meanwhile nobody can answer the only question the owner asked: “Are we making more money than we spend to get the phone to ring and the appointment to stick?”
Vanity charts are comfort food. Dollar dashboards are vegetables. you try to eat vegetables.
Start from the outcome, walk backward
Pick the outcome that pays you. Kept appointment. Admission. Closed deal. Qualified booked call. Whatever actually moves the P&L.
Then walk backward only as far as you can defend:
- outcome
- sales or admissions step that predicts it
- contact (call, form, chat)
- click or impression
- spend
If a tile on the dashboard is not on that chain or a known leading indicator for that chain, it is optional. Optional means it does not get to sit in the middle of the screen like a celebrity.
Attribution will be messy. Show ranges, not fake precision
Multi-touch is real. People see an ad, search later, call from a missed-call agent, show up because their mom nagged them. Your pixel did not witness the mom.
So I show:
- spend by channel
- inbound contacts by channel where the tracking is honest
- opportunities and revenue with clear definitions
- a note on what we cannot see
It is better to show “callable leads from paid search this week” and “admissions with first-touch paid within 14 days” as separate truths than invent a single magical ROAS number that assumes omniscience.
Speed metrics belong next to money metrics
In a lot of service businesses, speed-to-first-contact predicts revenue better than clever copy.
Put time-to-call, after-hours capture rate, and missed-call recovery next to CPL. Otherwise marketing gets blamed for ops leaks, or ops gets blamed for weak demand, and everyone argues with different charts.
One screen. Shared definitions. Fewer religious wars.
Definitions on the dashboard or they do not exist
What is a lead? What is qualified? What is a kept appointment? What timezone? What happens when a duplicate call shows up twice?
Write it under the chart in human language. If two teams disagree, fix the definition before you build another visualization.
AI can help summarize anomalies. It cannot save you from two departments using the word “lead” to mean different animals.
Design for Monday decisions
A good weekly dashboard answers:
- What did we spend?
- What did we get?
- What broke in the pipe?
- What will we change this week?
Red/yellow/green is fine if the thresholds are real. Sparklines are fine if someone will look. Thirty filters nobody uses are not fine.
Also like a small “data health” row: pixel firing, call tracking gaps, CRM fields missing. Bad data deserves a louder alarm than a soft engagement dip.
Kill charts that never change a decision
If a metric has not driven an action in a month, archive it. Dashboards accrete like junk drawers.
The goal is not a museum of analytics. The goal is a cockpit.
Where AI helps without becoming the point
- draft anomaly explanations for humans to verify
- fill missing categorization when rules fail (with review)
- narrate a weekly digest that links back to the same source charts
Do not replace the dollar chain with a chatbot that “insights” at you. Make the chain visible. Then maybe add a narrator.
A Monday ritual that sticks
Open the dollar dashboard before the vanity one. Ask what changed in spend, contact, and kept outcomes. Assign one owner for the biggest leak. Close the laptop.
If that ritual takes more than fifteen minutes, the dashboard is too clever. Simplify until a busy operator can run it without a data translator in the room.
Bottom line
Show dollars. Show the pipe that produces dollars. Admit what you cannot attribute. Put speed and ops leaks on the same screen as spend.
If your dashboard cannot survive the question “so what do we do differently on Monday,” it is art. Hang it somewhere else.
