Case studies

Case study · AdventureWorks, Microsoft’s sample database of a bicycle maker

The apparent success

Store bike sales rose a fifth. Time to celebrate?

AdventureWorks’ bike sales to stores rose from the first half of to the second. Worth celebrating, on the face of it. So we split the sales by region, to see where the rise came from.

The contradiction

The biggest region barely moved.

Split by region, North America, of the sales, grew , while Europe grew and the Pacific more than quadrupled.

So the rise seems to come from the smaller regions.

The discovery

In North America, one bike replaced another.

Split again by bike type, North America’s Road bike sales fell a quarter, while Touring bikes, first sold on , added .

So the region barely moved because a fall and a rise cancelled out.

And the new bikes went to new stores.

Every Touring bike went to stores that first bought bikes in . The existing ones bought less.

So the rise came from new stores and a new line, while the business it already had shrank.

The check

The total might be zero. What moved it holds.

The rise’s 95% interval runs from to : it reaches below no change. But Touring’s rise and the existing stores’ fall each hold on their own.

  • Not the season. In the second half sold less than the first.
  • Fewer bikes, not cheaper ones. Existing stores took fewer Road bikes by the year’s end.
  • Not the discounts. Counted after them, the rise is .
  • But thin. Two quarters against two, and Touring is quarters old.

And no person has checked what it counts.

On AdventureWorks’ Living Map it is B2B Bikes Revenue, under B2B Channel Revenue. The machine matched what it counts, and no person has reviewed it.

What we’d decide

Proposed decision

Don’t plan an expansion on the .

In AdventureWorks’ sample data, which describes no real company, the rise is a new bike line sold to new stores, while the existing stores bought less.

Next: Find out why North America’s existing stores are buying less, before putting money into new ones.

Provisional · further evidence required

The strategy, from what we know and what we don’t

  1. Back the new line, and the stores that took it.

    Because: New stores brought a quarter, most of it Touring: more than the whole rise.

    Not known: Whether it lasts. Touring is quarters old.

  2. Find out why North America’s existing stores buy less, before spending on new ones.

    Because: They bought less, while Europe’s existing stores grew .

    Not known: Why. The records show what the stores bought, not why.

  3. Don’t copy a regional playbook yet.

    Because: The product’s AI proposed "Double down on Europe expansion playbook", but of Europe’s rise is new stores too.

    Not known: Whether new stores keep coming: .

What would change it: If North America’s existing stores return to their earlier orders while Touring holds for two more quarters, the rise is growth, and an expansion is worth planning.

The full analysis the combinations, the stores, the checks, what ran and what we worked out, the query

The definition

B2B Bikes Revenue: SUM(unitprice * orderqty) on sales.salesorderdetail, bikes sold to stores (category 'Bikes'). Matched by the machine; no person has reviewed it.

On AdventureWorks’ Living Map, B2B Bikes Revenue hangs under B2B Channel Revenue, beside B2B Components Revenue, B2B Clothing Revenue and B2B Accessories Revenue. Of the map’s metrics, a person has approved ; B2B Bikes Revenue isn’t one of them.

The rise, region by region

RegionChange a quarter95% intervalOn the first half
North America
Europe
Pacific

In the first half, North America made a quarter of the : . The three add up to the rise; North America’s interval crosses zero.

What ran

KPI Drill-Down ( s), then Deep Analysis in Trend mode ( s), on B2B Bikes Revenue by quarter, January 2012 to December 2013, the sample’s own dates: rows. The product rates its confidence : . How they work, on Product.

Region by bike type

The combination that moved North America.

RegionRoadMountainTouring
North America
Europe
Pacific

Each cell is the change in a quarter’s sales from the first half of to the second, with its interval beneath; no percentages inside a combination, as the product shows none. Touring was first sold to a store on .

Stores, by when they first bought bikes

Existing stores bought less; new stores brought the rise.

StoresFirst halfSecond halfChange a quarter
Existing ()
New in ()
North America, existing
Europe, existing

Existing stores bought their first bikes before ; they bought of Touring. Existing stores ordering each quarter: ; new stores: . Every store in the Pacific is new in .

The checks

What the rise is not, and where it is thin.

In , the second half sold on the first, so the rise isn’t a season. Existing stores took Road bikes a quarter; every store’s Road bikes went from to while each sold for more (): fewer bikes, not cheaper ones. Counted after discounts (each line’s total), the rise is , against before them.

How sure each part is

Two quarters against two.

Like the product, we compare the mean of the last two quarters with the two before, and give each part a percentile bootstrap interval on that change. With two quarters a side, that interval runs from the worst pairing of a second-half quarter with a first-half one to the best, so a part “holds” when every second-half quarter beat every first-half one, or fell short of it. That is a thin test, which is why the check names it; Touring’s rise doesn’t rest on it, since the line didn’t exist before .

What we worked out for this page

The saved run cut by region and country only.

The run crossed region with country, one inside the other, so its combinations added nothing, and it skipped the bike model, since one model sold in only two quarters and it needs three. So we rebuilt the run, to the cent, from Microsoft’s public copy of AdventureWorks (its 2025 release moves every date 4,017 days later; we moved them back), and cut it by bike type and by when each store first bought bikes, treating each part as Deep Analysis does. Deep Analysis crosses two dimensions at a time, never three, so region by bike type is the kind of cut it makes, though this run didn’t pick bike type. It can’t yet group stores by their first order, so that cut is ours.

By country

countries grew beyond doubt.

Only grew beyond doubt. have intervals that cross zero.

What doesn’t hold

Not yet, then.

The overall rise might be zero, and so might North America’s, the United States’, France’s and Canada’s. The product’s summary calls it "": its phrase for a rise it can’t confirm.

What the product’s AI wrote, and why it isn’t evidence

  1. ""
  2. ""
  3. ""

Written by the product’s AI after the numbers and quoted as it wrote them: none of it is evidence.

Who decided what

No person has reviewed the definition the run used. The decision above is proposed, not taken. Nobody has acted on any of it: AdventureWorks is Microsoft’s sample database of a bicycle maker, and none of this is a customer’s result.

About the data

AdventureWorks is Microsoft’s sample database: its figures, and its dates, describe no real company. Every figure in the regions and the check is the product’s own output, saved by the run on 27 September 2026, except each region’s share of the first half, which we worked out from its figures; the combinations, the stores and the checks we worked out from the same data, as “What we worked out for this page” says. How we check a case.

The query behind every number

The engine wrote this from the definition. No AI touches it; the same question gives the same query.

, …  the joins for each factor follow:  lines in all

Design partners

Bring your own question.

Design partners take the question their teams keep asking through their own map, with us.

Become a design partner