We have shipped enough dashboards to know that "we built a dashboard" and "the team uses the dashboard" are two completely different projects. The first one is two weeks of work. The second is a year of habits.
Most dashboards die in the same way. The owner gets excited, IT spends a month plumbing data, the agency hands over a beautiful Looker board, the team opens it twice that week, and by month two it is sitting unused while everyone goes back to the WhatsApp group and the gut feel.
Here is the pattern we use to make a dashboard part of the morning routine instead.
Rule 1: Start from the three numbers the owner already tracks in her head
Before we build anything, we ask the owner-operator what numbers she already tracks in her head. Not what numbers she thinks she should track. What numbers she already does.
For a kopitiam owner it might be: revenue today versus same day last week, the top-selling drink, and which outlet is short-staffed. For a distributor it might be: invoice value out, stock value in, and which sales rep has not closed anything this week.
These are the numbers the owner is already paying attention to. They are usually three to five. They are the spine of the dashboard. Everything else is supporting detail.
If you start from a template with 24 KPIs, you build a dashboard for nobody. If you start from the three numbers the owner already cares about, you build a dashboard that gets opened.
Rule 2: Mobile first, even if "real work" is on desktop
The owner-operator is on her phone. Always. Even if she has a laptop, she is checking the kopitiam from her phone in between everything else. The dashboard has to work on a phone before it works on a desktop.
This is not a "make it responsive" point. It is a "the first build target is mobile" point. Different decision. Mobile-first means the three numbers fit on one screen with no scrolling. It means the buttons are tap-sized. It means there is no hover state because there is no hover on mobile.
The analyst on the team gets the drill-down view on desktop. That is the second build. Not the first.
Rule 3: Ship in two weeks, then watch for a week before invoicing
This is the rule that most BI agencies will not adopt because it is unprofitable. We adopt it because it is the only way to know if the dashboard works.
Build the dashboard in two weeks. Hand it over. Spend the next week watching how the team actually opens it. Not asking. Watching. Open Vercel Analytics, check which views get hit, see what gets ignored. Sometimes go to the kedai and see if the manager has it pulled up on her phone during the morning shift.
What we find in week three is what makes the dashboard land. Usually it is one of these:
- The team is opening the dashboard but going straight to the drill-down view, never the summary. Move the drill-down to the top.
- The team is opening the dashboard once at the start of the shift and not again. Add the "current shift" view as the default, not "today."
- The team is not opening it at all. Find out why on WhatsApp. Usually they did not know about it, or they cannot remember the URL. Bookmark it on their phones during a 10-minute visit.
That third week is when we revise. Then we invoice.
The Kopitiam Saji example
Kopitiam Saji has two outlets. The owner used to make her decisions from a paper notebook and a printed POS report. She knew her numbers in her head. She did not know them across two outlets at the same time.
We built her a three-card dashboard for the phone:
- Today's revenue, both outlets, side by side.
- Top 3 items selling right now.
- Items at stock-out risk, sorted by how soon they run out.
That is it. No charts. No drill-downs. Three cards, three numbers each.
After the rewatch week, we moved the stock-out card to the top. She was opening the dashboard at 9am to figure out what to buy at the wholesaler on the way to the outlet. Revenue was secondary because she could look at that in the evening.
It has been six months. She opens it five times a day on average. The dashboard is part of how she runs the business now, not a project that got delivered and forgotten.
The one question that kills most dashboards before they ship
Before any line of code, we ask the owner: "If this dashboard did not exist, what would you do instead?"
If the answer is "I would just call the manager and ask," the dashboard might not survive. We have to make it faster than a phone call.
If the answer is "I would have to wait for the monthly report," the dashboard will work. Anything is faster than a monthly report.
If the answer is "I have no idea, this is all guesswork right now," we will probably ship something that gets opened daily because we are replacing nothing with something.
It is a question that filters out the dashboards that get built and forgotten. Ask it before you start. Save everyone the rewatch week.
If you want a dashboard that gets opened, send us a brief.