Contact Distances Online: corrections, missing cities and broken pages

What to include

The URL of the page and what looked wrong is enough to check a figure against the dataset. A screenshot is rarely needed; the address of the page almost always is.

• A distance that looks wrong — send the route and the figure you expected. Worth reading how the distances are estimated first: these are modelled from a straight line and a road factor rather than routed turn by turn, so a gap of a few per cent against a mapping service is the method rather than a mistake. A gap much larger than that is worth knowing about.
• A missing or misplaced city — the place name and state, and roughly where it should be. Cities come from a public gazetteer and small places do fall through it.
• A page that does not load, or a link that goes nowhere — the address you were on and the one you were trying to reach.

What happens to a reported distance

A reported figure gets checked against a real driving distance, and if the gap is real the route joins the ground-truth table the estimator is measured on. That table is not decoration: the published accuracy band is computed from it, and the build fails if the engine drifts past what these pages claim. So a single correction can move the number printed across the site.

That has already happened. A reader-supplied check on Aberdeen, South Dakota to Bismarck, North Dakota — 202 miles by road against the 167 quoted here — went into the table, moved the ninetieth-percentile error from 13.4% to 14.3%, and the published band moved with it. Routes chosen by someone else are worth more than routes chosen here, because a table built only from journeys we thought to test cannot surface the kind of error we did not expect.

Known limits, so they can be told apart from faults

Because the road figure is derived from a straight line rather than routed, it reads low wherever the road is far longer than the shape of the journey suggests. How the figure is put together is set out on the method page.

Short trips are where the percentage misleads most. Seven miles of road against a six-mile straight line is a large percentage and a small distance. A figure that looks far out is worth reporting either way.

A city pair with no page of its own is not a fault either. Around sixteen thousand places are in the dataset and any two of them can be answered from the calculator on the home page; the pages that exist are the routes people actually looked for. A pair that returns nothing at all, rather than an answer without a page, is worth an email.

What this site is

Distances Online is a road-distance reference for the United States, published as a free utility. It carries no advertising, runs no analytics and has no account system. Editorial questions and corrections both reach the same address.