I used to think an “insight” was just pointing out the highs and lows. The biggest variances. The odd outliers.
Sorry – that was lazy. Those things aren’t insights. They’re data points with an attitude.
An insight is not “Sales are down 12%.”
That’s a statement.
It’s only an insight when you can explain: why it happened and what to do next. Until then, you’re not offering insights. You’re just serving up noise in a nice chart.
The Definition That Actually Matters
Here’s mine:
Insights are the combination of cause and consequence – they show why something has happened and what you can do about it.
That means:
- Spotting an unusual value isn’t enough.
- Surfacing a trend line isn’t enough.
- Even having a dashboard that “highlights what’s important” isn’t enough.
If your dashboard can’t explain impact and action, it’s not generating insights. It’s generating trivia.
The Analyst’s Craft: A Systematic Method
So how do you get there? It’s not magic. It’s not even “genius.” It’s a craft – and it’s systematic:
- Start with the anomaly. Something shifted. Something’s off.
- Form hypotheses. What are the possible causes?
- Test them. Bring in other datasets, segment the numbers, or compare time periods.
- Translate into business meaning. Does this matter to revenue, cost, risk, reputation?
- Frame the options. Do we fix, prevent, or exploit?
Why AI Isn’t There (Yet)
This is where people get it wrong. They assume:
“AI can crunch more data than me, so surely it’ll find better insights than me.”
No.
AI is brilliant at surfacing anomalies and patterns. But it has two fatal blind spots:
- It doesn’t know what it doesn’t know.
- If your datasets are incomplete, AI won’t think to fill the gaps by getting more data.
- If a critical external factor isn’t in the numbers, it won’t look for it.
- Example: AI spots that sales fell 10%. It analyses seasonality, geography, channel mix. But it misses the fact that a competitor launched a discount campaign in the same week – because that data isn’t available to it.
- An analyst can ask: “What’s missing here? What other data should I bring in?” and go hunt it down.
- It can’t judge real-world significance.
- AI will happily flag a thousand correlations. But it won’t tell you which one actually matters to your business model.
- Example: HR turnover spikes. AI notices more sick days and a shift in gender ratio. It doesn’t know that a new manager started six months ago – and that morale collapsed. An analyst can connect dots across datasets and lived experience.
The Human Advantage: Curiosity + Context
Where AI stops at “the pattern,” analysts push further with questions AI doesn’t even know to ask:
- “What data points are missing here?”
- “What else could be influencing this?”
- “Who can I speak to in the business to validate this?”
That’s the difference. Analysts don’t just crunch numbers – they chase blind spots.
Back to our HR example:
- AI flags the spike.
- Human digs deeper: exit interview themes, Glassdoor reviews, salary benchmarks, competitor hiring ads.
- The insight: a leadership problem compounded by external poaching.
- The action: coach the manager, fix comp packages, and protect your best people.
That’s the gulf. Until AI can self-diagnose gaps in its own field of vision, it can’t replace analysts.
Data Storytelling is Dead
You’ve probably heard the mantra: “Tell stories with data.”
Let’s be blunt – that’s not enough. Nobody cares about your pretty scatterplot if it doesn’t explain what to do differently tomorrow.
You don’t tell stories with data.
You tell stories with insights.
The data is just the backdrop. The insight is the plot twist. The action is the ending.
Example:
Data story: “Channel A outperformed Channel B.”
Insight story: “Channel B flopped because its campaign overlapped with a public holiday – next time, reallocate the budget and expect a 15% lift.”
See the difference? One’s bedtime reading. The other’s business-critical.
So What? (And Why You Should Care)
Because calling numbers “insights” when they aren’t is professional malpractice.
Your execs don’t need another dashboard that just points out what they already know. They need context, causation, and action.
That’s what earns trust as an analyst. That’s what drives real decisions. And that’s why – until further notice – insights remain a human superpower.
Takeaway for Your Next Dashboard
Before you hit publish, ask yourself:
- Am I showing just what happened?
- Or have I explained why it happened and what to do next?
- And most importantly: have I hunted for blind spots the data alone can’t reveal?
If you can’t confidently answer “yes” to that last one, you don’t have insights. You have noise.
And no business needs another noisy dashboard.
Where to Practice This (Without the Pressure)
The truth is, asking these kinds of questions – “What’s missing here?”, “What else could be influencing this?”, “Am I falling for the obvious story?” – is tough when you’re the only analyst in the room.
That’s why I created the Tableau Insights Community (TIC).
It’s a friendly, supportive space where you can:
- Test your thinking and challenge your own assumptions.
- Learn how other analysts spot blind spots in their data.
- Practice turning patterns into true insights before you take them back to stakeholders.
If you want to build dashboards that deliver more than surface-level reporting – dashboards that actually drive decisions – TIC is the best place to sharpen that skillset.
👉 You can read the full offer here: TIC Coaching and Community Overview
