# The core datasets: sales, registrations, and the parc
A carmaker announces "record sales" of 200,000 units in a quarter. A rival analyst says the same brand sold 170,000. Neither is lying. One is counting vehicles shipped to dealers. The other is counting vehicles actually registered to customers. That 30,000-unit gap is the difference between a factory metric and a market metric, and mistaking one for the other has wrecked more forecasts than any recession.
This lesson separates the three datasets that anchor almost every automotive analysis: wholesale shipments, retail registrations, and the parc (the installed base of vehicles on the road). Get these straight and everything downstream, from market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.Voir la définition complète → to aftersales revenue, gets cleaner.
Wholesale = units the manufacturer ships to its dealer network or importers. Also called "factory sales" or "sell-in."
This is what a carmaker books when a truck of new cars leaves for the dealer lot. It is a production and supply-chain signal, not a demand signal. Automakers like it because they control it and it hits their revenue.
The trap: dealers can be stuffed with inventory. If sell-in runs ahead of actual customer purchases, cars pile up on lots. High wholesale numbers can mask weak real demand.
Registrations = units actually titled and registered with a government authority when a customer takes delivery. Also called "sell-out."
This is the closest proxy for true consumer demand. A car is registered once, to one owner, in one jurisdiction. That makes registrations the cleaner basis for market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.Voir la définition complète →.
In the US, registration data is aggregated commercially. The classic source is R.L. Polk, now part of S&P Global Mobility (Polk was absorbed via IHS, then IHS merged with S&P Global in 2022). In Europe, national registration figures are collected by bodies like ACEA (the European Automobile Manufacturers' Association), which publishes monthly registration data for free.
Key distinction: wholesale is a company-reported figure; registrations are a government-sourced figure aggregated by third parties. They rarely match in any given month.
The parc = the total stock of vehicles currently in operation (VIO) in a market. Think of it as everything registrations have added, minus everything that has been scrapped or exported.
Simple identity:
Parc(this year) = Parc(last year)
+ new registrations
- scrappage
- net exportsThe parc is the foundation of the aftersales business: parts, service, tires, insurance, financing on used cars. A new-car sale happens once. A vehicle in the parc generates revenue for 15+ years.
Rough scale, as of the mid-2020s (estimates, verify against current S&P Global Mobility or national data):
Compare that to annual new sales (US new light-vehicle sales run around 15 to 16 million a year as of 2024 to 2025 estimates). The parc is roughly 18 times larger than one year of sales. That ratio alone explains why aftersales is where durable margin lives.
Market shareMarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.Voir la définition complète → should be computed from registrations, not wholesale. Here is why it matters.
Suppose in a given month:
Wrong calculation (using wholesale):
50,000 / 400,000 = 12.5% shareRight calculation (using registrations):
40,000 / 400,000 = 10.0% shareA 2.5 percentage-point error, purely from mixing datasets. Repeat that across a year of planning and you misallocate marketing spend, production, and incentives.
Aftersales demand scales with the parc, not with new sales. If you size a parts inventory or a service-network plan off new-registration volume, you will systematically under-forecast.
Worked example. Say you want to estimate annual brake-pad replacements for a brand:
Brand parc in market: 4,000,000 vehicles
Avg brake-pad replacements
per vehicle per year: 0.30 (industry rule-of-thumb estimate)
Annual demand = 4,000,000 x 0.30 = 1,200,000 setsNew registrations that year might be only 250,000. If you had sized your parts business off new sales, you would have planned for a fraction of true demand. The parc drives the number.
Wholesale spikes at quarter-end (automakers pushing units to dealers to hit targets) look like demand surges but are inventory shifts. Registrations reveal what customers actually did. Watching the gap between the two is itself a signal: a widening gap means inventory is building, a classic precursor to discounting.
🎬 [VIDEO: "How Car Sales Are Actually Counted" - youtube.com - a clear primer on the difference between shipments, registrations, and vehicles in operation]
These datasets are only as good as their governance. Concrete checks:
Unit of counting. Is one row one VIN (Vehicle Identification Number, the unique 17-character code stamped on every vehicle)? VIN-level registration data is the gold standard because it lets you dedupe and track a vehicle across its life. Aggregate counts cannot.
Vehicle scope. Passenger cars only, or light commercial vehicles too? Does it include heavy trucks? A "total market" figure is meaningless until you know the vehicle-type definition. ACEA and S&P Global use specific segment definitions; do not blend across sources without mapping them.
Fuel-type classification. In 2026 this matters enormously. Is a plug-in hybrid (PHEV) counted as "electrified," "BEV" (battery electric vehicle), or its own category? EV market-share debates are frequently just definitional fights. Pin down the taxonomy first.
Geography and registration lag. A vehicle registered in one region but operated in another distorts local parc estimates (fleet and leasing companies register centrally). Registration data also lags: month-end figures get revised.
Scrappage assumptions. Parc figures depend on how you model vehicles leaving the fleet. Scrappage is estimated, not directly reported in most markets, so two providers can show different parc totals for the same country. Always ask which scrappage model was used.
Vérification des acquis
1. A carmaker reports high wholesale shipment numbers while customer purchases at dealerships are actually declining. What does this scenario most likely indicate?
2. Why are retail registrations considered a cleaner basis for calculating market share than wholesale shipments?
3. An analyst wants to gauge current supply-chain and production activity at a manufacturer. Which dataset is the most directly relevant, and why?
4. Select ALL correct answers about the distinction between wholesale shipments and retail registrations.
Sélectionnez toutes les réponses correctes.
5. Select ALL correct answers about why confusing the core automotive datasets damages analysis.
Sélectionnez toutes les réponses correctes.
A mature automotive analysis uses all three in sequence:
1. Wholesale tells you what the factory pushed out and what revenue the OEM (Original Equipment Manufacturer, the vehicle maker itself) booked.
2. Registrations tell you what customers truly bought, and therefore real market sharemarket shareThe percentage of total industry sales your company captures in a given period. It measures competitive position relative to rivals in a defined market.Voir la définition complète → and demand trends.
3. Parc tells you the size of the future aftersales, insurance, and used-car opportunity.
Reconcile them. Wholesale minus registrations, tracked over time, is your dealer inventory proxy. Cumulative registrations minus scrappage builds the parc. When someone quotes a single "sales" number with no source, your first question is always: sell-in or sell-out?