The Daily Business Dashboard You Check on Your Phone
Most owners do not need another report. They need three numbers on their phone before the first meeting, and they need to trust them enough to act without opening anything else.
That was the premise of the working session I co-presented with Tim Holum, co-founder of Coulee Tech, at the AI Business Retreat 2026 in Spooner, Wisconsin. We took a fictional company’s financial data and built a working dashboard live on the projector, in front of the room, in under an hour. No mockups. Real data in, real dashboard out.
Why do most dashboard projects die?
Not for technical reasons. They die because nobody decided what the dashboard was for.
The usual sequence is that someone asks for visibility, a project gets scoped, and what comes back is a page with fourteen charts on it. It gets opened enthusiastically for two weeks and then never again, because nothing on it changes what anyone does that day. Fourteen charts is not visibility. It is a filing cabinet with a login.
The discipline that makes this work is the same one from hiring your first AI employee: get specific enough that somebody can check it in under a minute. A dashboard nobody can act on before their coffee is cold is a report, and you already have too many of those.
What number do you actually need every morning?
Ask it this way instead: what is the number that, if it moved the wrong way overnight, would change what you do today?
For most small and mid-sized businesses it is a short list. Cash in the bank against what is committed. Receivables past due, and by how long. Work sold but not yet delivered. Sometimes utilization, sometimes pipeline, rarely revenue on its own. Revenue is the number people ask for first and act on least, because by the time it moves, the decision that moved it was made weeks ago.
Three numbers is usually right. Five is the ceiling. If you cannot say out loud what you would do differently when a number goes red, it does not belong on the morning view. Put it on a monthly review where it belongs.
What does it take to build one now?
Less than it did, and the reason is worth understanding rather than just believing.
The pieces have not changed: a connection to the system your numbers already live in, a query that pulls the right rows, a calculation that turns rows into the number you actually mean, and a page that renders it on a phone. What has changed is that describing each of those in plain English now produces working code, which collapses the build from a project into an afternoon.
That does not eliminate the hard parts. It relocates them. The hard parts are now definitional. Does “revenue” mean invoiced or collected? Does a job count when it is sold, scheduled, or finished? Is a customer who pays on day sixty past due at forty-five? Every business answers these differently, and if you do not answer them deliberately, the tool will answer them for you and be confidently wrong on your phone every morning.
We walk through this build in more depth in turning a database into a dashboard, and the specific case of pulling profitability numbers out of accounting data in calculating EBITDA from QuickBooks.
Why did we use fake company data on stage?
Deliberately, and it is worth copying.
Building the first version against invented data lets you argue about definitions without anyone defending their department. When the receivables number looks wrong on a fictional company, the conversation is about the rule. When it looks wrong on your real company, the conversation becomes about whose numbers those are, and the design work stops.
Build the shape first on data nobody owns. Connect it to the real source once the definitions have survived an argument.
What does it need to be trusted?
A dashboard is only useful if you stop double-checking it. Four things get you there.
A timestamp. When did this last update? A number with no time on it gets verified every single time, which defeats the purpose.
A visible definition. One line on the page saying what the number counts. Not in documentation. On the page.
One source per number. If cash can come from two systems, it will eventually disagree with itself, and the first disagreement kills the whole dashboard’s credibility permanently.
A name attached. Somebody owns each number and gets told when the feed breaks. Silent failure is worse than no dashboard, because a stale number that looks fresh will get acted on.
That last one is where most of these break down in practice. The data pipeline fails on a Tuesday, nothing errors visibly, and the dashboard keeps showing last week’s cash position with today’s date on it.
Start with one number this week
Pick the single number you would want on your phone tomorrow morning. Write down exactly what it counts, in one sentence, including the edge case you are least sure about. Find out which system holds the underlying data and whether anyone can currently query it.
That is genuinely the hard part. The rendering is an afternoon.
If the answer to “which system holds it” turned out to be “four systems and a spreadsheet,” that is the common case, and it is the problem worth solving before anyone builds a chart. We do this work with businesses across Houston and Dallas Fort Worth, usually starting by getting the data into one place that can actually be queried.