A job post landed in my Upwork feed last year that stopped me for a second. It did not just ask for a portfolio link like most of them do. It said, in plain words, share 2 to 3 dashboards, ideally with visible DAX and the model behind them, not templates. Most clients tell you what they want. This one told me exactly what it did not want before I sent a single thing.

I had a pretty dashboard ready to go. Soft shadows, tidy cards, the kind that looks great as a thumbnail. I almost sent it. Then I read that last word again, not templates, and I understood that this client had hired someone before, been handed a downloaded theme with fake numbers inside, and paid real money for it.

A pretty dashboard proves you can pick a theme. A real project proves you can be trusted with the data.
01/WHAT THEY CHECK

What a client is really looking at when they open your link

When a client opens your first link they are not admiring your color choices. They are running a quiet test in their head. Can this person take a messy real file and make a decision fall out the other end. That one line, visible DAX and the model, not templates, is a client asking for the plumbing on purpose, because the front of a dashboard is the easiest part to fake and they already know it.

A reviewer spends maybe ninety seconds on your first link, and in that minute and a half they are checking three things. Whether the data looks like it came from real life. Whether the model behind it makes any sense. And whether the dashboard answers a question a manager would actually ask on a Monday morning, rather than showing a chart for every column just because the column was there.

WHAT A REVIEWER ACTUALLY DOES ON YOUR FIRST LINKOPENS LINK0sSCANS LAYOUT15sCLICKS FILTER30sOPENS MODEL50sREADS MEASURE70sDECIDES90sMESSAGE OR TAB CLOSEDninety seconds decides whether you exist to this client
fig 01 · what a reviewer checks in the first ninety seconds

If your project survives those ninety seconds you get a message. If it reads as a template they close the tab, and you never hear back, and you quietly decide the market is slow when really your link just answered the wrong question.

02/DEPTH OVER COUNT

One project you can defend beats five you downloaded

The instinct is to fill the portfolio. Five dashboards, ten, one per industry, so nobody thinks you are new. It backfires, because five shallow dashboards read as five templates, and the reviewer has scrolled past that exact gallery a hundred times this month. More of the same thing does not add up to proof.

One project that goes all the way through does something the gallery cannot. Start with a realistic messy dataset, not the clean sample file everyone downloads. Clean it with rules you can actually name out loud. Model it. Build one executive dashboard. Then land the whole thing on a single slide that says what a manager should do about the numbers. That full run is the thing nobody can copy off the internet, because the mess in the middle is yours.

ONE PROJECT, ALL THE WAY THROUGHRAW FILE3 DATE FORMATSCLEANNAMED RULESMODELFACTS + DIMSDASHBOARDONE SCREENSTORYTHE DECISIONTHE PARTS NOBODY CAN FAKE BY DOWNLOADING A THEMEclean rules, a real model, and a slide that says what to do
fig 02 · one project from a raw file to a decision on a slide

And when you present it, say the boring parts out loud. Here is where the source file had three date formats. Here is the row I dropped and why. That honesty is the exact signal the not templates client is hunting for, and most people hide it because they think it looks unfinished. It is the opposite. I wrote more about building the one screen a manager reads in the executive dashboard piece.

03/MODEL AND STORY

Show the model and the story, not just the charts

Most portfolios show the front of the dashboard and stop right there. But the client asked for the DAX and the model on purpose, so give them a second screen, the plain one. A small model diagram, one fact table, two dimension tables, and two or three measures written out with a single line each on why that measure exists at all.

Two measures beat twenty charts

You do not need a wall of DAX to look serious. Two measures a manager truly cares about, a rolling total and a same period last year comparison, shown with the reasoning behind them, tell a reviewer far more than a page of default visuals ever will. A measure with a reason attached is a person thinking. Twenty charts is a person filling space.

Then comes the story. A dashboard shows numbers. A story says what to do about them. So end your project on one slide that names the finding and the move, in the plainest words you have. I wrote a whole piece on turning data into a story, because this is the part that separates a report from advice, and advice is what people pay for.

TEMPLATE LOOKEVERY COLUMN CHARTED, NO QUESTION ASKEDJUDGMENT LOOKFACTSales YoY = DIVIDE(now, same period last year)ONE ANSWER, THE MODEL, AND A REASONvs
fig 03 · template look versus judgment look, side by side

What a reviewer feels in that contrast is the difference between a screen that dumps everything on them and one that points at a single answer and quietly says what it would do next.

04/WHERE IT LIVES

Where to put it so a busy client actually sees it

The project can be perfect and still lose if the client has to download a file or make an account before they can see anything. Put the dashboard where it opens in one click, a Power BI public link or a hosted page. Put the model screenshot and the story slide right beside it. And write three plain lines underneath, the question you asked, the data you used, and the decision it points to. That is the whole tour a reviewer needs.

If you do not have client work yet, you truly do not need it. A reviewer cannot tell whether the messy file came from a paying client or from an open dataset you found on a Tuesday night, they can only tell whether you handled it like someone who has done this before. I wrote about building a portfolio with no clients at all in this one, if that is where you are standing right now.

Here is where I stop teaching and give you the honest handoff. The framework above gets you one project you can defend. The version tuned to the exact role you are chasing, the dataset that mirrors their industry, the two measures their own manager already asks about, the slide written in their language, that shaping is the part I do with people one at a time. If you want that, start a project and tell me the job you are aiming at, and I will tell you straight whether your current link would survive those ninety seconds.

> Where this leaves you

So the short version. A client who writes not templates is telling you they have been burned, and the fix is one project you can defend from end to end. A real messy file cleaned by rules you can name, a model you are willing to show, and a slide that says what to do. Build that one thing, host it so it opens in a click, and describe the boring parts out loud, because the boring parts are the proof.

If you would rather have that project shaped around the exact role you want, the model that mirrors their data and the two measures their manager already watches, that tuned build is the done for you version I make with you. And every tool I ship along the way comes free with the membership, current and future, yours to keep. The pretty template was never going to get answered. A project you can stand behind will.