Open in app ↗

Learn · The LUCID Framework

Five stages between raw data and a decision someone actually makes.

LUCID is not a charting style and not a tool. It's a thinking sequence: the order in which a data storyteller earns their audience's trust and moves them to act. If you've ever built a dashboard nobody used, LUCID explains why, and fixes it.

L
Leakage
Find where value is escaping
U
Unification
Bring the evidence together
C
Causality
Prove why it's happening
I
Illumination
Make the insight visible
D
Decision
Hand over a clear action

The five stages, in plain language

One example, carried through every stage

The fastest way to understand LUCID is to watch one story travel through all five stages. Our running example: you analyse sales data for a small coffee chain and notice revenue is slipping. Here's what each stage asks of you.

L
Leakage
Where is value escaping?

Every data story starts with a leak: money, time, customers, or trust quietly draining away somewhere. Beginners often start with "here's my dataset." LUCID starts with "here's the problem." If you can't name what's leaking and roughly how much, you don't have a story yet; you have a spreadsheet.

The question you answerWhat is being lost, where, and what is it costing?

Coffee chain exampleWeekday afternoon sales have dropped 18% over three months at 4 of 12 stores. That's roughly ₹2.4L a month leaking, enough to matter and specific enough to investigate.
U
Unification
Bring the evidence together

A leak rarely explains itself from one table. Unification means pulling the relevant sources (sales, foot traffic, staffing, weather, competitors) into one coherent picture, and being honest about what each source can and can't tell you. This is where you establish that your evidence is trustworthy and complete enough to reason from.

The question you answerWhat evidence do I need, and does it all describe the same reality?

Coffee chain exampleYou join POS sales with store footfall counters, shift rosters, and a list of nearby café openings. Now you can see that footfall fell only at the 4 stores. Sales per visitor is unchanged.
C
Causality
Prove why, don't just correlate

This is the stage beginners skip, and the one that separates a data story from a chart dump. Two things moving together is not an explanation. Causality means testing the candidate explanations and ruling out the ones the data doesn't support, and saying so honestly when you can only show a strong association, not proof.

The question you answerOf all the possible reasons, which one does the evidence actually support?

Coffee chain exampleThree hypotheses: new competitors, staffing cuts, menu price rise. The price rise hit all 12 stores, ruled out. Staffing is unchanged, ruled out. All 4 declining stores gained a competitor within 500 m in the same quarter. That's your driver.
I
Illumination
Make the insight impossible to miss

Only now do you design visuals, and each chart exists to make the causal finding obvious to someone who wasn't in the analysis. Annotation does the storytelling: mark the moment the competitor opened, label the gap, remove everything that doesn't serve the point. One illuminating chart beats ten decorative ones.

The question you answerCan a busy non-technical reader see the finding in ten seconds?

Coffee chain exampleA single line chart per affected store, with a vertical marker labelled "Competitor opens ↓". The sales line visibly bends at each marker. No dashboard needed.
D
Decision
Hand over a clear action

A story that ends with "interesting, right?" has failed. The Decision stage converts the finding into options a specific person can act on, with the trade-offs priced. Who decides, what are their choices, and what happens under each? If your story doesn't change what someone does next week, return to stage L.

The question you answerWhat should who do, by when, and what does each option cost?

Coffee chain exampleFor the ops head: (a) launch a 3–5 pm loyalty offer at the 4 stores, est. cost ₹40K/mo vs ₹2.4L leak; (b) match competitor pricing on the top 3 items; (c) accept the loss and monitor. Recommendation: (a), review in 8 weeks.

Common confusions

What beginners get wrong, cleared up

"LUCID is a tool or a chart template."

It's a thinking sequence. You can execute it in Anthrena Studio, Excel, Python, or on a whiteboard. Tools change; the five questions don't.

"Leakage means data leakage, like in machine learning."

Different concept entirely. In LUCID, Leakage is the business leak: the value quietly escaping (revenue, customers, hours) that gives your story a reason to exist.

"I should start by making charts, then find a story in them."

Charts come fourth, at Illumination. If you chart first, you decorate whatever pattern you happen to notice. LUCID makes you earn the chart: problem → evidence → cause → then visual.

"Causality means I must run a formal causal inference model."

It means you must reason causally: list rival explanations, test them against the data, and be honest about confidence. Sometimes that's a model; often it's disciplined elimination, like the coffee example.

"The stages are a strict waterfall. You never go back."

You'll loop. Unification often reveals the leak was mis-stated; Causality often demands new evidence. The sequence is the spine of the final story, not a cage for the analysis.

"A dashboard with filters counts as the Decision stage."

A dashboard hands the thinking back to the reader. Decision means you did the thinking: named options, priced trade-offs, made a recommendation someone can accept or reject.

Keep this handy

The one-glance cheat sheet

LUCID in one table

StagePlain meaningYou're done when…Beginner trap
L — LeakageName the problem and its costYou can state what's leaking, where, and how muchStarting from the dataset instead of the problem
U — UnificationAssemble trustworthy evidenceAll relevant sources describe one coherent pictureUsing whatever single table was easiest to get
C — CausalityTest why it's happeningRival explanations are ruled out or ruled in, honestlyPresenting a correlation as the cause
I — IlluminationDesign visuals that carry the findingA non-expert sees the point in ten secondsTen pretty charts, zero annotations
D — DecisionDeliver options and a recommendationA named person can act on it next weekEnding with "interesting insights" and no action

Your first LUCID story: pick a dataset you already know, write one sentence per stage before touching any tool, and only then build. If the five sentences hold together, the story will too.