RedHub Weekly: The Week Blank Cells Stopped Scoring

RedHub AI Editorial5 min read

A woman at a workshop bench holds a pencil over the one empty box on a ticked paper grid, lit in red
Jump to a section12

RedHub Weekly: The Week Blank Cells Stopped Scoring

TL;DR

  • What shipped: four rounds of fixes to how our workbooks and engines read bad input, plus a site audit pass.
  • What broke: a blank cell could score as if you had typed 0, and our fix for that briefly made an empty table read clean.
  • The numbers: 121 of 161 workbooks had at least one affected column. We rebuilt 136 buyer zips on October 27 and 162 on October 28, and read every one back from storage.
  • What to do: if you bought on or before October 28, re-download from your RedRig page.

What changed at RedHub this week?

Between October 24 and 28, 2026, we changed how RedHub workbooks and engines handle bad input. A blank cell could compute a result as if 0 had been typed, and 121 of 161 workbooks had at least one column where that happened. Those cells now show CHECK INPUT, and engines refuse empty files in one line. We rebuilt 136 buyer zips on October 27 and 162 on October 28. Anyone who bought on or before October 28 should re-download.


This is the first RedHub Weekly. The format is simple: what shipped, what broke, and what the numbers said. This week the second list is longer than the first, and that is the honest shape of it.

A band with a hole in it

The smallest fix first. Several products scored to one decimal place but described their bands in whole numbers: "55 to 74," then "75 and up." A score of 74.5 belonged to neither. The band text now reads "55 to under 75," on the product page, the FAQ, the workbook and the zip. Nobody would have lost money on this. It was still a sentence that did not describe what the product did.

A mark of 0.4 became 0

The next one was worse. We ran every engine against inputs nobody is supposed to type: half marks, fractions, blanks, text, numbers out of range. In 27 products the engine and its own workbook disagreed on at least one of those inputs. The common case was a mark below 1. The engine rounded it down to 0, and a 0 tripped a gate that the workbook only fires on an exact 0. Same input, two different answers, from the same purchase.

Now the engine refuses that mark and names the row and column. We chose a refusal over a guess because a guess is the defect.

A blank cell scored as if you had typed zero

Then we cleared every input cell in every shipped workbook, one at a time, and recalculated. Of 161 workbooks, 121 had at least one column where an empty cell still produced a result: a verdict, a band, a flag, or a gate, computed as if you had entered 0. The engine for the same product refused the blank. The spreadsheet quietly answered.

Those rows now show CHECK INPUT, and so does every summary that counts or sums them. A fully empty row still reads as empty. Every filled row recalculates exactly as before, which we checked per product on the shipped sample and three filled variants.

Then the fix made an empty table look clean

This is the one we did not see coming. Teaching an empty row to drop out of the count meant that a table with no rows at all had nothing left to count. In about 32 products, an empty table read as a clean result. A buyer who opened the workbook and looked at the dashboard before entering anything would have seen a pass.

With no rows in use, every verdict cell now reads CHECK INPUT. Engines given a file with no data (header only, blank lines, all commas, an empty folder) refuse it in one line that names the file.

What broke on the site

Three things, in rising order of embarrassment.

The missed-call calculator had no H1 in its served HTML. The heading rendered on screen, but only after the page loaded in the browser, so a crawler reading the raw page found none. An Ahrefs audit caught it.

One product page said its engine and workbook "return the same verdict on every input." That stopped being true when we added the CHECK INPUT guards, because the engine now refuses inputs the workbook still scores. We re-measured: the two agree exactly on 7,645 cases, and the engine refuses all 4,001 blank or out-of-range marks it was built to refuse. The page now states that scope instead of "every input." The underlying gap is still open and goes in next week's list.

And one of our own calibration checks, the kind that confirms a gate can still fail, had been failing quietly since October 25. The guard it looked for had moved, and the check kept looking in the old place. It reads the new place now.

The numbers

We rebuilt 136 buyer zips on October 27 and 162 on October 28. Each one was uploaded, then pulled back down from storage and compared, before we called it shipped. On the site side, the diagram images behind our tool pages went from 705 MB to 172 MB across 872 files, and six explainer pages now carry Article schema.

What we can't tell you

We can't tell you how many buyers saw a wrong answer because of any of this. Your workbook runs on your machine, and your results never reach us. Any number we gave you there would be invented.

What we can tell you is where the fixed files are. If you bought on or before October 28, your downloads are on your RedRig page, and each click signs a fresh link to the current version.

Decision Guide

  • You bought a workbook or engine on or before October 28, 2026: re-download it. The file you have may score a blank cell.
  • You bought after October 28: you already have the fixed version.
  • You are mid-way through a filled workbook: every filled row recalculates the same, so you can copy your inputs into the new file.

FAQ

What is RedHub Weekly?

A short weekly note on what shipped at RedHub, what broke, and what the numbers said. It covers fixes to our own products as well as new work.

What does CHECK INPUT mean in a RedHub workbook?

It means a row in use is missing a required value, or the table has no rows yet. The workbook shows it instead of computing a result from a blank cell.

Will my filled-in workbook give different results after the update?

No. Every filled row recalculates exactly as before. Only rows with a required blank, and tables with no rows at all, read differently.

How do I get the updated files?

Go to your RedRig page at redhub.ai/rig. Each download link is signed fresh when you click it, so it always points at the current version.