# The Diffusion Road

**Live D3 instrument / frozen research edition of 18 September 2026.**

Website revision: r6. Source archive SHA-256: e5e67eb8a7bfa34620f6b6e0f608537a449a18a4acff5e3745a1d865acee674a.

One reference-date scrubber, two independent evidence lanes, and a clearly separated conditional future. The showroom can change quickly; the road carries the cars bought before it.

## Reader action

Scrub the historical record and compare new registrations with the latest preceding fleet snapshot. Then explicitly enter the conditional future; never glide unnoticed from observation into a scenario.

## The showroom lane

Choose one month or a trailing year. The D3 path and a normalized 100-unit rail show the BEV share of new registrations. The trailing year divides the sum of BEVs by the sum of all registrations; it does not average twelve percentages.

## The road lane

A separate 100-unit rail shows the latest preceding fleet observation on its actual reference date. Its publication date remains visible. Missing observations stay missing; units never migrate between these different populations.

## The conditional horizon

An explicit mode change reveals the supplied 2026–2040 scenarios. Choose a sales path and an assumed annual BEV exit rate. Dashed paths and hatched rails identify conditional arithmetic; the 1–5% exit range is not probability coverage.

## Published default

THE DIFFUSION ROAD / SELECTED READING
Mode: Recorded evidence.
Registration basis: Trailing twelve months / ratio of sums.
New registrations / Sep 2025–Aug 2026: 723,980 / 2,947,795 = 24.560053%.
Active fleet / 2026-07-01: 2,365,047 / 49,696,710 = 4.758961%.
Stock publication date: 2026-08-31.
Independent denominators. The reference-date join uses the frozen revised edition, not information known at the selected month’s end.
Each lane has its own 100 percentage-point units. Units are not vehicles and never move between populations.

## Technology

Locally bundled D3/SVG, shared URL state, exact fractional-unit fills, a linked temporal comparison, nearest-date inspection, keyboard stepping and an exact table. Hover previews a date; click, tap or keyboard selection commits it. No new chart library or WebGL.

## Rejected encoding

A literal vehicle-flow Sankey is not defensible: separate used imports, exports, scrappage and cohort survival are not observed. A racing speedometer would add a false speed metaphor to a share.

## Accessibility

Both rails use the same zero–100% scale, each with its own denominator. One unit means one percentage point, not one car. Native controls, a single keyboard timeline target, mobile stacking and an exact table preserve the reading. No quantitative marks travel or morph. The complete default reading survives without JavaScript.

## Sources and boundaries

German passenger cars / KBA. Monthly: January 2017–August 2026; six fleet reference snapshots: January 2022–July 2026. The reference-date join uses this frozen revised edition, not information historically available in the selected month. No fleet observation is inferred before 2022; no partial trailing year is substituted before December 2017.

The supplied conditional paths cover 2026–2040. Fleet and annual registrations are fixed at 49,486,487 and 2,857,591. Initial January 2026 BEVs: 2,034,260. Annual BEV exits of 1%, 3% and 5% are assumptions, not estimated lifetimes. Used imports and other net adjustments are zero. Model quantities can be fractional.

- https://www.schym.de/data/ev-diffusion/tables/germany_monthly.csv
- https://www.schym.de/data/ev-diffusion/tables/germany_stock.csv
- https://www.schym.de/data/ev-diffusion/tables/fleet_scenarios.csv

Live instrument: https://www.schym.de/blog/to-charge-or-not-to-charge#diffusion-road