DeckPlanner
Home/Corrections

Corrections log

This page logs substantive corrections: a wrong number, a wrong formula, a citation that pointed at the wrong clause. Formatting and typo fixes are not logged. Nothing is removed.

Why this page exists

A calculator site's only real asset is that its numbers can be trusted. A site that quietly edits its errors is asking to be trusted without evidence. This log is the evidence.

It also serves a practical purpose: if you found a result here that later changed, this is where you find out why.

Log

2026-10 — Share image was an SVG

What was wrong: every page pointed og:image at og.svg. Facebook, X, LinkedIn, Slack and WeChat all refuse SVG for link previews, so every shared link would have rendered as bare text with no thumbnail.

Changed to: the card is now rendered to og.png at 1200×630 by _og.js, with the dimensions and alt text declared as og:image:width, og:image:height and og:image:alt, and a matching twitter:image.

Effect: none on the pages themselves. The failure was invisible in a browser and would only have appeared when someone shared a link.

2026-10 — No favicon.ico

What was wrong: the pages declared only favicon.svg. Browsers and crawlers still request /favicon.ico by default, so that path returned a 404 on every visit.

Changed to: a 32×32 favicon.ico now ships alongside the SVG, plus an 180×180 apple-touch-icon.png for iOS home screens.

Effect: none on ranking. It removes a recurring 404 from the server log and gives iOS a proper home-screen icon instead of a screenshot of the page.

2026-10 — Stair stringer remaining depth

What was wrong: the stringer page originally described remaining depth as board depth minus a projection term involving the riser and tread. The calculator computes it as board depth minus riser height, measured vertically, and the two did not agree.

Changed to: the formula text now matches the calculation, and the worked example states the vertical measure explicitly.

Effect: no numerical change to any result. The explanatory text was wrong, not the arithmetic.

2026-10 — Retaining wall base gravel

What was wrong: the worked example stated 0.11 cubic yards of base gravel per foot of wall, which implies a trench three feet deep. The actual figure for a 12-inch-wide, 6-inch-deep base is 0.019 cubic yards per foot.

Changed to: the example now reads 0.56 cubic yards for a 30 ft wall, matching the calculator.

Effect: the calculator was correct throughout; only the example text was wrong. A reader following the example would have ordered roughly six times too much stone.

2026-10 — Retaining wall block length read as 4 inches

What was wrong: the block-size selector encodes two numbers, height and length, as a single option value. The standard 8x12 block was written value="8x4", so the script read the length as 4 inches instead of 12. A 30 ft wall reported 90 blocks per course instead of 30, and 378 blocks total instead of 126.

Changed to: the option value is 8x12 and the script default matches. Verified in a real browser at 30 ft: 30 blocks per course, 126 total.

Effect: a three-fold overcount on every use of that block size, from the day the page went up until this test. This is the second dimension-encoding error of the same shape found on a site built this month, and the reason every calculator now has a hand-computed baseline committed to the test file: reading the formula would not have caught it, because the formula was right and the input was wrong.

2026-10 — Code tables unreadable on a phone

What was wrong: the three-column code tables were allowed to compress. At 375px the value column collapsed to one word per line, so a requirement read "Deck / surface / more than / 30″ above / grade". The numbers were right and the page did not overflow, which is why the first pass of testing passed it.

Changed to: every table sits in a horizontally scrollable container with a 470px minimum width, so a narrow screen scrolls one table instead of shredding it.

Effect: no change to content. All 48 tables across 44 pages were affected.

2026-10 — Four pages shipped without a title tag

What was wrong: the contact, privacy, editorial standards and corrections pages had no <title> element at all, and no meta description fallback. Every page in the build reads its title from one field, and those four pages never set it. Nothing failed: the template rendered, the page looked correct, and the tag was simply absent.

Changed to: each page has an explicit title, and the build now refuses to write any page missing a title, description, h1 or body, or with a duplicate slug or Chinese characters in the body. It exits non-zero and names the page.

Effect: search results for those four URLs would have shown the bare domain instead of a title. The gate caught a fifth problem the same day it was added.

2026-10 — Two calculators missing from the tool index

What was wrong: the pergola shade and sod calculators were written and published, but never flagged as tools. The flag drives the tool index, the WebApplication schema and the sitemap priority, so both pages worked, were linked from their section, and were absent from /calculators with no structured data.

Changed to: the build now derives the flag from the page markup — if a page renders a calculator, it is a tool page — so the flag cannot be forgotten again.

Effect: the tool index went from 31 entries to 33, which is the real count.

2026-10 — Pergola shade at an angle

What was wrong: the first draft described shade coverage as a single percentage, which is only true when the sun is directly overhead.

Changed to: the tool now reports overhead coverage and coverage at a stated sun angle separately, with the projection term explained.

Effect: a 2x6 rafter at 24 inches covers 22.9% overhead and about 32% at a 45-degree sun. The single figure understated what people actually experience.

How to report something

The contact page explains what to include. A report with the inputs, the result, and your own calculation is usually enough to confirm or refute an error within a day.

What is not logged here

Questions people actually ask

Have any calculators given wrong numbers?

Two text errors were found and corrected in October 2026 — one in a worked example, one in a formula description — but neither affected a calculator's actual output. Both are logged above in full.

Why publish corrections at all?

Because a site that quietly fixes its errors gives readers no way to judge how careful it is. The log is evidence rather than a claim.

Are old corrections removed?

No. Entries stay, dated, even when the error was embarrassing.

Related