Descriptive → Prescriptive

The ladder

Analytics is often split into four types that form a natural ladder — each rung more valuable, and harder, than the one below it.

Each rung depends on the ones below it. You cannot diagnose a problem you have not described. You cannot prescribe a remedy for a cause you have not identified. The ladder is not decorative — it is the logical order of how evidence accumulates.

One running example: an online store’s sales

To make this concrete, follow a single real situation across all four rungs.

The setup. Sparq Store sells electronics online across three regions: US, EU, and Asia. Last quarter’s revenue was $1.2 M — down 8 % versus the previous quarter (which came in at $1.3 M). The data team is called in.

Rung 1 — Descriptive: “What happened?”

Descriptive analytics summarises and reports on data that already exists. The output is a KPI — a Key Performance Indicator, meaning a single number tied to a specific business goal — or a dashboard full of them.

The goal here is not to explain or predict; it is to see clearly. Most companies live on this rung: weekly revenue reports, daily active user counts, monthly churn rates.

At Sparq Store:

• Total revenue: $1.2 M (down $96 k vs last quarter)
• Orders: 14,200 (down 6 %)
• Average order value (AOV — total revenue divided by number of orders): $84.51 (down 2 %)
• Revenue by region: US $680 k (+2 %), EU $280 k (−28 %), Asia $240 k (+4 %)

The EU number stands out immediately. That is all descriptive analytics does — it surfaces the signal. It does not yet tell you why.

Rung 2 — Diagnostic: “Why did it happen?”

Diagnostic analytics drills into the data to find the root cause — the underlying reason an outcome occurred, as opposed to a symptom (a visible side effect). Techniques include segmentation (splitting data by category), drill-down (zooming into a subset), and correlation analysis (checking whether two things move together — say, a price rise paired with falling orders — to point at a likely cause).

At Sparq Store: The team slices EU revenue by product category, customer segment, and time. They find:

• The EU drop was entirely concentrated in the first six weeks of the quarter.
• A price increase was rolled out in the EU market on the quarter’s first day (laptops and tablets went up an average of 12 %).
• EU conversion rate (the percentage of site visitors who complete a purchase) fell from 3.1 % to 2.2 % immediately after the price change.
• The US and Asia markets saw no price change — and no conversion drop.

Root cause: the EU price increase caused a conversion drop large enough to cost the company approximately $96 k in revenue.

Rung 3 — Predictive: “What will happen?”

Predictive analytics uses historical patterns — often through statistical models or machine learning — to forecast future outcomes. It introduces uncertainty: a prediction is never a certainty, only a probability-weighted estimate.

At Sparq Store: The team builds a simple linear trend model on the past eight quarters of EU revenue. They also factor in the current conversion rate. The model’s output:

• Base case (prices unchanged): next quarter EU revenue ≈ $261 k — a further decline of 7 % as some customers do not return.
• Total company next quarter: ~$1.11 M (US and Asia stay flat; EU deteriorates further).
• Confidence interval: $1.03 M to $1.19 M at 90 % confidence (meaning the forecast could easily miss by up to $80 k).

Predictive analytics adds a time arrow — it projects current reality forward. But it only tells you what will happen if nothing changes. Deciding what to change is the next rung.

Rung 4 — Prescriptive: “What should we do?”

Prescriptive analytics recommends a decision or action by combining the predictions above with constraints, objectives, and trade-offs. Techniques include optimisation (finding the mathematically best option given constraints), simulation (running many “what if” scenarios), and decision analysis.

This is the rung where data analysis directly drives business strategy.

At Sparq Store: The team models three options:

Recommendation: revert EU prices. Net expected recovery of +$104 k exceeds the margin sacrifice of $18 k by a factor of nearly six. If the team wants to protect margin, the targeted-discount option is second best.

Why the ladder metaphor holds

Three properties make the ladder more than just a taxonomy:

1. Dependency. Each rung genuinely requires the one below it. If your descriptive data is wrong (the revenue figure itself is inaccurate), your diagnosis will chase the wrong signal. If your diagnosis is wrong (you blamed EU conversion when the real culprit was a shipping delay), your forecast will miss the true driver. Bad foundations collapse the whole structure.

2. Increasing difficulty. Descriptive analytics requires clean data and a good reporting tool. Diagnostic analytics requires analytical skill, domain knowledge, and the ability to rule out alternative explanations. Predictive analytics requires statistical or ML expertise plus enough historical data to train a reliable model. Prescriptive analytics requires all of the above, plus explicit modelling of decisions, constraints, and trade-offs. Each rung demands more.

3. Increasing value (when done right). A dashboard tells you what happened. A prescription tells you what to do. The prescriptive recommendation directly changes the business decision, which directly changes outcomes. That is as close as analytics gets to being the business itself.

Putting it together

The CEO’s four questions from the opening are now answered — and each maps cleanly to a rung:

1. “What exactly happened?” — Descriptive. Revenue was $1.2 M, down 8 %, driven almost entirely by a 28 % drop in EU.
2. “Why did it drop?” — Diagnostic. A 12 % EU price increase cut conversion from 3.1 % to 2.2 %, costing ~$96 k.
3. “Will it drop again next quarter?” — Predictive. Yes, likely to ~$1.11 M unless prices change.
4. “What should we do?” — Prescriptive. Revert EU prices; expected net recovery of +$104 k.

None of those answers would have been possible without the one before it. The ladder is the logic of evidence-based decision-making made explicit.

Next

Metrics vs. KPIs — not every number worth measuring is a KPI (a key metric tied to a strategic goal). Choosing what to measure, and what not to, is where descriptive analytics either pays off or wastes everyone’s time.