Skip to content
woodfrog
Analytics · Deep dives

Questions a dashboard can't answer.

Bring us the hard question about your business. We go deep into your data, your processes and the people who run them, and answer it with evidence in 3 to 6 weeks: what drives what, by how much, what works best, and what happens if.

Build your brief ↓Try the what-if ↓
Five kinds of question

The questions that decide things.

01What drives what, and by how muchWhich factors move the number, in which direction, and at what rate?What moves churn? What explains yield differences between shifts?
02Why did it happenWhat changed, and which part of the change was price, volume, mix or cost?Why did margin fall when volume rose?
03What is the most effective wayOf the ways we run this process, which gives the best result, and why?Which dispatch, staffing or maintenance routine works best?
04What ifIf we change this, what happens to the things we care about?What happens to profit if we raise price? If a line is down for a day?
05Where is the moneyWhich customers, products or routes carry the profit, and which lose it?Which 10% of customers carry the margin, and which cost us money?
The question canvas

Turn your question into a brief.

Five choices, and you get a structured investigation brief: the question, a hypothesis tree, the methods and data we would use, and how big it is. Nothing is sent unless you choose to send it.

01What is the area?
02Which number do you want to move?
03What can you change?

Pick any. If you pick none we start with the two most common.

04Where does the data live?
05What would make you act?
Your investigation brief

What is driving gross margin, by how much, and what would happen to it if we changed list price and promotions?

Hypothesis tree
Gross margin
We can change
List pricePromotions
We cannot change, but must account for
Competitor pricesDemand seasonalityInput cost swings
Methods we would use
  • Price elasticity model
  • Price, volume and mix bridge
  • Scenario model
Data we would need
  • Sales by product and month
  • Price and discount history
  • Promotion calendar
  • Unit costs
Suggested size: Sprint, 2 to 3 weeksYour data looks ready to start.
  1. W1.1Agree the decision and the threshold.
  2. W1.2Profile the data: sales by product and month and price and discount history.
  3. W1.3Interview the people who run the process, and walk it end to end.
Email it to us
See how we would approach this →
How an investigation runs

Slow on purpose, and rigorous.

  1. Step 1ScopeOne decision and one threshold. "We will act if the answer is above X." We agree what a good answer looks like before any analysis.
  2. Step 2ImmerseInterviews, a process walk-through, data profiling, and the real rules behind the numbers. The people who do the work know things no table records.
  3. Step 3AnalyseDriver models, elasticities, root-cause tests, process mining or simulation, chosen for the question. The simplest method that answers it.
  4. Step 4Test the evidenceEvery finding gets a confidence range, its assumptions and a sensitivity check, and is tested on data the model did not see.
  5. Step 5Read out and hand overA findings report, the driver model, a scenario workbook you keep, a ranked action list, and the code so another analyst could reproduce it.
The what-if playground

Move a lever. See what it does.

Change the price and watch profit, volume and margin move. Then change how sensitive volume is to price, and notice that the best price jumps. That number is not in any dashboard. It has to be measured from your own history, and measuring it is what a deep dive does.

Move the levers

A made-up baseline. Type your own numbers if you like.

Profit a month₹1,80,045+₹20,045 vs today
Units a month18,141-9.3% vs today
Margin41.0%+3.0 pts
Break-even units13,953to cover fixed cost
Profit against price changeBest in range: +24.0% (₹2,06,452)
-15.0%0.0%+15.0%+30.0%
Your priceBest price in range
What matters mostChange in monthly profit when each input moves, from largest to smallest. This is the question a deep dive answers with your real numbers.
Unit cost ±10%
Baseline volume ±10%
Fixed cost ±10%
Price ±10%
Price sensitivity ±25%

An illustration on a made-up baseline, using a constant-sensitivity model. Not a forecast. Real answers come from your own price and volume history.

What you get

An answer you can act on, and the model to keep.

A written answerA findings report a board can read: the question, the answer, the evidence and the limits.
The driver modelThe effect of each factor, with a range, so you know what matters and by how much.
A what-if workbookInteractive scenarios you keep and run yourself, with every assumption in plain view.
A ranked action listWhat to do first, each with its sized value and an owner.
Evidence, not opinionA confidence range, the assumptions and a test on data the model did not see, for every finding.
Reproducible workThe code and data steps, so another analyst can check it or extend it.
Three sizes

Fixed scope, fixed fee.

Sprint2 to 3 weeksOne question and one decision. A written answer with the evidence, and a short readout.Best for: A sharp question you already have.
Deep diveAbout 6 weeksA few linked questions, the driver model and a scenario workbook you keep, plus a leadership readout.Best for: A question that touches several parts of the business.
Standing decision modelQuarterlyThe model kept current, refreshed each quarter, with an analyst on call for the next question.Best for: A decision you make again and again.
Six example investigations

See how we would approach it.

These are use cases, what we would do, not client results.

Questions

Deep dives, answered.

What is a deep dive?

An analyst-led investigation of one business question. We go deep into your data, your processes and the people who run them, and answer with evidence: what drives what, by how much, what works best and what happens if you change something.

How is it different from a dashboard?

A dashboard describes what happened. A deep dive explains why, measures how much each factor matters, and lets you test changes before you make them. Most deep dives end in a better dashboard, but the dashboard cannot give you the answer alone.

Is it a model that runs by itself?

No. It is a piece of focused expert work with a written answer. You also keep the model and a scenario workbook, so you can run your own what-ifs afterwards. If the answer needs to run every day, we build that next.

How long does it take and what does it cost?

A sprint takes 2 to 3 weeks and a deep dive about 6. Scope and fee are fixed before we start, and never billed by the hour.

What do you need from us?

One decision to inform, a named person who knows the data, access to the data under an NDA, and an hour with the people who run the process in the first week.

What is the question you cannot answer today?

Tell us. We will say whether a deep dive is the right way to answer it, and what it would take.

See our dashboards