Dashboards Are for Looking Up, Data Stories for Understanding

Hand someone a dashboard and you have technically shown them the data. But unless they already know what question to ask, you may have shown them very little.

A dashboard is a room full of switches with no sign saying which one matters. Twelve charts, forty filters, a date range, a dropdown - all sitting there, equally weighted, waiting for a visitor who already knows exactly what they came to find. Most people don't. They glance, feel vaguely uninformed, and leave. The data was all there. The point was nowhere.

I build the other thing. Simpler Story turns a dataset into a guided, scrolling narrative - the kind where the chart redraws itself as you read, one claim at a time, until you reach the end understanding something you didn't when you started. The stories built on it have crossed 3 million views in the past year - one on how Israelis rank their cities' reputations against what the numbers actually say got picked up by outlets like Ynet and N12. Rather than explain the platform in the abstract, let me just build one in front of you, and you'll see where the interesting decisions - and the dangerous ones - actually live.

The raw material is always a mess

Here's the file I'll work from: Israeli salary data - specifically the National Insurance Institute's wage report for the first half of 2025, the kind of thing that comes out of a government portal as an export nobody expected a machine to read. Merged header cells. Units hiding inside the values - is that column shekels, thousands of shekels, gross, net? A footnote three tabs over that redefines a category halfway through the years. Hebrew text running right-to-left through columns laid out left-to-right.

So the first day isn't design at all. It's reading. Not the schema - the schema lies, or at least flatters. The header says "average wage" and means six different things across the sheet. I go through the rows until I know what each column actually holds, then normalize it into something a chart can be trusted to draw: consistent types, units pinned down once and explicitly, one row per observation. This is where "monthly" and "annual," "gross" and "net," "median" and "mean" get nailed down so that nothing downstream can quietly conflate them. Almost all of the honesty in the final piece is won right here, before a single pixel exists. Get it wrong now and every beautiful chart that follows is a beautiful lie.

Finding the one thing worth saying

Now the editorial act, the one I'm actually accountable for. The clean data can support a hundred charts. A story gets one thesis, or it's just a slideshow with transitions.

I could publish the average salary and call it a day. But the average salary is a trap: not a false number, a misleading one when used as shorthand for ordinary experience. A handful of very high earners drag the mean up above what almost anyone actually makes. In this data the average wage comes out at ₪15,098 a month, while the median - the person standing exactly in the middle - earns ₪10,586. Both figures are exact; both are true. But quoting only the ₪15,098 describes a comfortable earner who mostly doesn't exist, and hides the ~₪4,500 gap that is the actual story: the distance between the number everyone repeats and the money most people take home. The spine wasn't something I decided in advance and went looking for - it fell out of the cleaned rows the moment the distribution was legible. That ordering matters, and I'll come back to why.

Notice what just happened, though. The instant I chose that gap as the thesis, I took a side. This is the thing dashboards get to pretend they never do. Forty charts shown at equal weight feels neutral, but choosing to present everything at once is itself a choice, and an unaccountable one, because it offloads the hard part onto a reader who doesn't have the context to do it. Showing everything isn't more neutral than choosing; it just hides who did the choosing. So I'd rather make the choice out loud and own it, which means staying alert to how close a point of view sits to propaganda. That's what the rest of the build is really about avoiding.

Building the scroll, one claim per step

A story on Simpler Story is a sequence of steps, and the discipline is brutally simple: each scroll step is exactly one claim paired with exactly one state of the chart. The step says what to notice; the chart shows it; the transition between steps is the argument.

So the salary story opens on a single big number - the average - because that's the number the reader arrived believing. Scroll, and a marker slides in showing where that average actually falls in the distribution: up in the thin right tail, far from the crowd. Scroll again, and the median lands where most people actually are, well below. Each step is allowed to assert only what its bound chart genuinely shows. No step gets to sneak in a claim the picture doesn't back up.

This is where legibility and honesty start pulling against each other. To make the story readable, I have to simplify: aggregate, round, drop categories, choose one framing instead of five. Simplification isn't the sin; it's the job. Showing the distribution instead of the average is a simplification that serves the truth. Truncating the y-axis so a modest rise looks like a cliff is a simplification that devours it. My test is simple: does this help the reader see what's actually there, or does it manufacture a reaction the full data wouldn't support?

Where a point of view turns into propaganda

Back to why the ordering matters - why the thesis has to be discovered in the data rather than imposed on it.

The failure mode of narrative data is seductive and common: start from the conclusion you wish were true, then reverse-engineer the charts to sell it. The mechanics look identical to honest storytelling from the outside - same animations, same one-claim-per-step discipline, same confident voice. The only difference is causal direction. Did the claim come out of the rows, or did the rows get bent to fit the claim?

I hold that line by treating the data as allowed to win. If, halfway through building the salary story, the distribution had turned out roughly symmetric - if the average had actually described most people fine - the thesis would have had to die, not the data. The reader gets guided through the evidence, never around it. And there's a structural backstop: every finished story also leaves the dataset reachable. Filters, interactions, and a dataset-grounded chat let skeptical readers ask their own questions instead of accepting my framing on faith. The guided scroll is the default, not a cage - and a point of view you can immediately fact-check reads as an argument you'll defend, not a wall built to stop people looking.

The part that makes people actually finish

None of this matters if nobody scrolls to the end, and this is where I'll admit the uncomfortable thing: a story that doesn't make you feel something doesn't get finished, and a story nobody finishes informs precisely nobody. So the emotional pull isn't decoration I could take or leave. The salary piece has a "here's where your salary falls on this curve" moment, and that moment is the whole reason the median lands with real force instead of as a footnote.

But feeling is exactly where overstatement creeps in - the breathless superlative, the "surged" that was a nudge, the color that codes a neutral number as alarming. So the rule I actually hold isn't "stay dry." It's that the affect can be as vivid as I can make it, but it has to attach to a number that's true at full precision. The reveal can be dramatic; the value it reveals is the real one. The "you are here" marker hits because it's your actual position, not a rounded-up version chosen for impact. Feeling is how the truth becomes hard to ignore - it was never permission to bend it.

What's left when the scroll ends

The finished salary story takes a couple of minutes to read and leaves you with one true thing you didn't have before: the number everyone quotes isn't the money most people see, and now you know roughly where you actually stand. A dashboard could have held every one of those same figures and left you with nothing, because it would have made you do the one job I'm here to do - decide what matters - at the exact moment you had the least context to do it.

That's the trade the whole platform is built on, across nearly twenty public stories and counting. A dashboard gives you everything and calls it transparency, then leaves you to find the one thing that matters on your own. A good story points you straight at that thing, shows you where it sits in the rest of the data, and still lets you go back and check the rest yourself.

Guiding someone's attention, done honestly, isn't the opposite of informing them. For most readers it's the only version of the data they'll actually finish, which makes it the only version that informs them at all.