There's a lot to be said about how terrible looking at charts is. They're too often not used the right way. When I look at a chart, I want it to be a map - something that tells me where I am so I can determine where to go next. When I want to know if things are okay, I want a value.
For example: If I want to know if my house price is good, I don't want a chart showing how much it has gone up or down. I want a number that tells me what it's worth versus what I paid for it. Simple. Direct. Actionable.
The Dashboard Problem
As a data engineer, I see this constantly: dashboards full of charts, scrolling chart after chart. And they have serious problems:
What Charts Cost You:
- Time waste: Every chart needs interpretation. You have to look at the axes, understand the scale, decode the trend.
- Screen real estate: Charts take up enormous space that could show more actual information.
- Compute expense: Rendering charts is expensive. Even with rollups and pre-aggregation, charts consume far more resources than values.
- Cognitive load: Each chart is another thing to process, another decision tree in your head.
What Values Give You:
- Instant comprehension: "Status: GOOD" or "ROI: +23%" requires zero interpretation.
- Minimal space: A single value fits anywhere. You can pack dozens where one chart would go.
- Cheap computation: One calculation, one number, done. Orders of magnitude less expensive.
- Actionable: You know immediately whether you need to do something.
Data has become the new IT - an expanding, often unnecessary expense. Everyone still touts "data is the new oil," but most organizations are drowning in dashboards that nobody uses effectively.
When Charts Are Actually Useful
I'm not saying charts are always bad. Sometimes they're exactly what you need:
- When you're looking at the whole population: CIE colour gamut charts show you the entire colour space. That's appropriate - you need to see the distribution.
- When you don't know where you are: Maps are charts. If you're lost and need to figure out where to go next, a map is essential.
- When you're exploring: Initial data exploration often benefits from visualization. But once you understand the pattern, codify it into a value.
As a general rule: If you're looking at the whole population or don't know your current state, charts are fine. Otherwise, you want a value.
The Bitcoin Example: SIGAF
I built a project called SIGAF (Should I Give A F***) to solve exactly this problem for Bitcoin price tracking. The question was simple: Should I care about what Bitcoin's price is doing right now?
I could have built a dashboard with:
- Price charts over various timeframes
- Moving averages
- Volume indicators
- RSI, MACD, Bollinger bands
All of which would require me to:
- Look at multiple charts
- Interpret the indicators
- Synthesize a conclusion
- Decide whether to care
Instead, SIGAF gives you this:
One value. One glance. You know immediately whether Bitcoin's momentum is strong, weak, or somewhere in between.
How It Works: Geometry Over Time Series
SIGAF doesn't use traditional technical analysis. It uses pure geometry.
The system:
- Normalises the entire price history against the all-time high (ATH) block and price. Everything becomes coordinates in a unit box where ATH = (1, 1).
-
Calculates geometric areas
representing different aspects of price
behaviour:
- A_past: Area before ATH block
- A_since: Area after ATH block
- A_vol: Volatility area (distance from current trend line)
-
Computes ratios:
- R = A_since / A_past: How much "weight" has accumulated since ATH
- H = A_vol / A_macro: How much volatility relative to macro trend
- Performance score P: Euclidean distance from current position to (1,1), normalised to 0-1 range
The formula for P:
$$x_c = \min(1, x_n)$$ $$\text{dist} = \sqrt{(1 - x_c)^2 + (1 - y_n)^2}$$ $$P = 1 - \frac{\text{dist}}{\sqrt{2}}$$
This uses ALL of the available data - every single price point in history - but does so in an extremely cheap and efficient way. No fancy indicators. No machine learning. Just geometry.
Why Geometry Works
Traditional technical indicators try to extract signals from time series data. They're computing things like "is this going up?" or "is this overbought?" But they do it in time series space, which is expensive and often noisy.
Geometric analysis in normalised space is fundamentally different:
- It's coordinate-invariant: Doesn't matter if Bitcoin is at $100 or $100,000 - the geometry is what matters
- It's computationally trivial: Calculate areas under trapezoids. That's it.
- It captures macro structure: The shape of the price history relative to ATH tells you the momentum story
- It's a single value: All that complexity reduces to P, R, H, and a simple status
The Broader Point
If I were in charge of a data organization, I would put guardrails around dashboards:
- You can generate all the values you want
- But not charts - unless you can justify why a value won't work
I would be very interested to see how that improves effective KPI tracking.
The hypothesis: Teams would be forced to think clearly about what actually matters. They couldn't hide behind pretty visualizations. They'd have to define success as a number, then track that number.
Think Twice Before You Chart
Before you make your next dashboard chart, ask yourself:
- Do I know where I am and have a destination? If yes → you want a value, not a chart.
- Am I looking at the whole population to understand its shape? If yes → chart might be appropriate.
- Could this be reduced to a single status, score, or metric? If yes → make it a value.
Values over charts. It's faster, cheaper, clearer, and more actionable.
SIGAF proves it. No charts. Just geometry. One value that tells you what you need to know.
Should you care about Bitcoin right now? The dashboard tells you in one glance. No interpretation required.
The SIGAF project is open source. Find it at: github.com/datatypesio/sigaf