Skip to content
Menu
The price study case

Competitor price study · system map

The price study, under the hood

Explore how public quotes became a workbook and a dashboard a European parcel operator’s pricing team could trust. Play the tour, pick one of the 6 journeys, or select any part to see what it does.

Figures from the project handover note

Journeys6 journeys through the system

PUBLIC SOURCESopen-source intelligenceCOLLECTa real browser sessionNORMALISEPython · one shapeDELIVER AND CHECKone pipeline, two outputsCLIENTEuropean parcel operatorWEEKLY MONITOR (ILLUSTRATIVE)Comparison sitesCompetitor pricingCollection agentAIBrowser sessionParserProvider listPrice tableWorkbookOne pipelineDashboardThree-way checkPricing teamCLIENTOpen sourcesWeekly readerAIChange logWeekly brief

Overview

Every figure starts as a public quote. Follow one through 16 parts and 22 connections, out to the workbook, the dashboard and the three-way check.

Drawn from the delivered project, with the client kept anonymous. The weekly monitor chapter is an illustration, not part of the client work.

By the numbers

competitor retail prices collected
5,093
of 237 destinations compared, gaps left blank
226
headline figures checked three ways, all agreed
37
workbook formulas, recalculating with zero errors
4,808

Figures from the project handover note, 2026. The client stays anonymous.

How each number was counted
5,093 competitor retail prices collected
Project handover note, 2026.
226 of 237 destinations compared, gaps left blank
Project handover note, 2026.
37 headline figures checked three ways, all agreed
The check before delivery: from the raw data, the workbook formulas and the dashboard pipeline, 2026.
4,808 workbook formulas, recalculating with zero errors
Project handover note, 2026.

The journeys, step by step

The same 6 journeys the map plays, written out. Each step names the parts it connects.

01A price is collectedCollect · 5 steps
  1. 1The pricing team needs one answer: where do its export prices sit against every competitor, destination by destination?Pricing team → Collection agent
  2. 2An AI agent runs the collection through a real browser session: the part that would have taken weeks by hand.Collection agent → Browser session
  3. 3It gathers quotes from public price-comparison sites.Comparison sites → Browser session
  4. 4And from a competitor’s public pricing service, in the same session.Competitor pricing → Browser session
  5. 5Every quote is handed to Python, in whatever format its source used.Browser session → Parser
02Cleaning into one tableNormalise · 4 steps
  1. 1Quotes arrive in different formats, with carriers named a dozen ways.Browser session → Parser
  2. 2Python parses each quote into one shape and maps dozens of carrier spellings onto a single list of 15 providers.Parser → Provider list
  3. 3The result is one consistent table of 5,093 competitor retail prices.Parser → Price table; Provider list → Price table
  4. 4Where a destination has no quote, it stays blank. 226 of 237 destinations are compared, and the gaps are shown rather than estimated.Price table → One pipeline
03One pipeline, two deliverablesDeliver and check · 6 steps
  1. 1A single pipeline takes the clean table.Price table → One pipeline
  2. 2It builds the workbook with live formulas: 4,808 of them.One pipeline → Workbook
  3. 3And the dashboard: a standalone report with charts, in the client’s own language and branding.One pipeline → Dashboard
  4. 4Both come from the same table through the same pipeline, so they cannot disagree.One pipeline → Workbook; One pipeline → Dashboard
  5. 5The pricing team reads the market at a glance, on any device.Dashboard → Pricing team
  6. 6Its analysts change an assumption in the workbook and watch it flow through.Workbook → Pricing team
04Checked three waysDeliver and check · 4 steps

All 37 figures agreed

  1. 1Every headline figure is calculated from the raw data.Price table → Three-way check
  2. 2Then again through the workbook formulas.Workbook → Three-way check
  3. 3And a third time through the dashboard pipeline.Dashboard → Three-way check
  4. 4All 37 headline figures matched, and the 4,808 formulas recalculate with zero errors. Only then did the client see them.Three-way check → Pricing team
05Where AI is usedAI · 4 steps
  1. 1An AI agent does the collection and cleaning: the slow, repetitive bulk of a market study.Collection agent → Browser session
  2. 2Every quote passes through the same Python parsing and the same list of providers.Parser → Provider list
  3. 3The agent’s work is not trusted on its own. Every headline figure is recomputed by three independent routes.Price table → Three-way check; Workbook → Three-way check; Dashboard → Three-way check
  4. 4People see the figures only after the check has passed.Three-way check → Pricing team
06Weekly monitor (illustrative)Illustration · 5 steps
  1. 1An illustration, not the client project: the same approach, run on a weekly schedule, becomes a standing monitor.Collection agent → Weekly reader
  2. 2It reads more open sources than prices: product pages, job ads, ad libraries, reviews, public registers and tariff PDFs.Open sources → Weekly reader
  3. 3Screenshots, PDFs, images and text are each read in their own form, and every change is logged with its evidence.Weekly reader → Change log
  4. 4The week’s changes are condensed into a short brief.Change log → Weekly brief
  5. 5It ends in one decision, for example whether to answer a competitor’s price cut, and on which routes.Weekly brief → Pricing team

Want a system like this, documented like this?

Tell us about the process or the product. We'll reply within two business days with a first view of what we'd build.

Start a project