Coding and AI classes in Gütersloh

import csv

points = {}
with open("guetersloh_charging.csv", encoding="utf-8") as fh:
    for row in csv.DictReader(fh):
        points[row["operator"]] = points.get(row["operator"], 0) + int(row["points"])

total = sum(points.values())
print("Operators:", len(points), " charging points:", total)

running = 0
for rank, (name, n) in enumerate(sorted(points.items(), key=lambda p: p[1], reverse=True), start=1):
    running += n
    if rank <= 3 or running / total >= 0.8:
        print(rank, n, f"{100 * running / total:.1f}%")
    if running / total >= 0.8:
        break

The largest operator, the town's own utility company, Stadtwerke Gütersloh, runs 94 of the 262 points, 35.9 percent on its own. The next two add only 16 and 14. By the sixteenth operator the running share reaches 80.2 percent, so 16 of 39 operators, about two in five, hold four fifths of all public charging points in the register. The other 23 share the remaining fifth. Analysts call this kind of check a Pareto analysis, after the observation that a few often account for most.

singles = sum(1 for n in points.values() if n == 1)
print("Operators with a single charging point:", singles)
half = 0
for rank, n in enumerate(sorted(points.values(), reverse=True), start=1):
    half += n
    if half * 2 >= total:
        print("Operators needed for half of all points:", rank)
        break

Rows taken from the Bundesnetzagentur charging register (version dated 1 September 2026, fetched 3 November 2026) where the place is Gütersloh; reused under CC BY 4.0 with credit to Bundesnetzagentur.de. Ranking, totals and percentages are worked out by us.

Why does this belong on a page about AI? Because data for AI is usually lopsided in exactly this way. In a pile of text, a handful of words make up a large part of everything written; in a pile of photos, a few kinds of picture dominate. A model trained on such data sees the head of the list thousands of times and the tail hardly at all, so it tends to do well on common cases and poorly on rare ones. A learner who has sorted and added up shares by hand knows to ask, before training anything, how the examples are spread.

1. Children, 8 to 12. Count the colours of 30 sweets, sort the colours from most to fewest, and add them up in order. After how many colours have you reached 24 sweets?
2. Teenagers. Run both versions, then change the stopping point to 50 percent and check that you get 4.
3. Adults. Do the same for any list at work: customers, products or support tickets. How many items make up four fifths?
4. Going further. Repeat the analysis counting stations instead of charging points. Does the largest operator's share go up or down, and why?