tolga boroglu.

The “Numbers Don’t Lie” Lie: Making the Wrong Decision with the Right Data

•4 min read•English•Tolga Boroğlu
#data-analyst#data-visualization#data#data-analysis#data-science

Numbers don't lie

As data analysts, we spend most of our day writing complex SQL, processing large data with PySpark and designing dashboards in tools like Metabase and Power BI. But that is only the tip of the iceberg. Underneath, our real job is to understand business needs, make sense of the data and help decision-makers set the right course. We don't just pull data; we spot anomalies and gaps in it and turn them into insight the company can act on strategically.

Now that AI tools (LLMs) are part of our daily work, writing code takes seconds. That speed brings a sneaky risk with it: silent errors.

AI and "Supposedly" Flawless Code

Writing a prompt and testing the SQL or Python it produces is no longer enough. Code that runs without errors does not mean the result is "correct". Invisible logical errors are much harder to find.

Every piece of code written without planning, without understanding the spirit of the data architecture, carries a hidden bill: time. Say you want to add a new business rule to code that an AI wrote in one flat, non-modular block full of nested subqueries. The structure collapses and you have to rewrite it from scratch. Every query written without CTEs (Common Table Expressions) or a clear logical structure is bound to mislead decision-makers sooner or later.

The Butterfly Effect of a Technical Mistake in the Boardroom

Let's look at a technical example of how a small mistake by an analyst can lead management into a wrong decision worth millions:

Suppose you are analysing customer retention and total revenue for the company's growth strategy. You need to join the main customer table with the transactions (or claims/costs) table. In the query you wrote, or had an AI write, you used a LEFT JOIN and overlooked a one-to-many relationship.

  • What happened? The query ran without a single error.
  • What happened behind the scenes? Because of multiple records in the transactions table, the revenue in the customer table was duplicated (rows multiplied).
  • What showed up on the dashboard? Total revenue and active customer value looked 30% higher than they really were.
  • The result: Looking at these inflated metrics, management decided, "Things are going great, we can put a huge budget into new customer acquisition." In reality existing customers were being lost, and the budget should have gone into retention.

This is exactly where "numbers don't lie" falls apart. Numbers don't lie, but a badly designed data model or a carelessly written query tells the biggest lie of all.

The Dashboard Illusion: Lots of Data, Zero Focus

Even with a flawless technical foundation, mistakes in presentation (visualization) can mislead managers or push them into panic.

  1. Visual clutter: Putting 25 KPIs, colorful pie charts and huge tables on one dashboard is not success. It pulls the decision-maker away from what they actually need and costs them focus. "A dashboard that shows everything says nothing."
  2. The wrong chart: Showing a trend over time with a bar chart, or putting two unrelated metrics on a dual axis so they appear correlated, can cause management to panic and make wrong operational decisions.

In Conclusion

An analyst's real value is not in writing complex algorithms but in safely building that dangerous bridge between data and the business. We are not operators who carry out requests ("pull these metrics") as given. We are strategic partners who ask "why?", build technical foundations that can withstand future change, and tell the story the data needs to tell in its simplest, most honest form.

AI gave us speed, but getting to the right decision still depends on our analytical thinking and our healthy scepticism of data.