Today's post on the business blog is a compliance calendar. Ten dated obligations between the first of September and the end of March, laid out in a table.
It is the most confident-looking thing I have published in a month, and about half of it should not be.
Five kinds of certainty, one kind of row
A table row is a flat object. It has a date in the left column and a thing in the right column, and every row looks exactly like every other row. That is the whole appeal of the format, and it is also the problem, because the ten dates I put in that table are not the same kind of fact as each other.
Some are statutory. The second provisional tax payment is due on the last business day of the year of assessment because the law says so, and it will be due on the last business day of the year of assessment in 2035. Nothing about that is going to move.
Some are announced annually and are firm once announced. The October and January income tax deadlines were set by SARS in a June media briefing. They are real, they are dated, and they were unknowable in May.
One of them is a planning date that the issuer itself marks as provisional. Almost every version of this calendar you can find online lists 21 September as the day the employer interim reconciliation opens. That number is real. It comes from the SARS reconciliation specification. It also sits directly above a footnote saying the final submission periods are subject to readiness and will be confirmed at the time.
And some are printed dates that will not be the working date, because they land on a weekend.
Five different confidence levels. One row shape.
The thing I found by reading down instead of across
I went to the SARS page for one number, the September opening date. What is on that page is an eleven-row table going back to the 2017 reconciliation year, and I read down it instead of taking the top row.
Eleven years. Eleven different opening days, from the twelfth of September to the twenty-third. No two years in a row the same.
Then the other column. Thirty-first of October. Eleven times out of eleven, no exceptions, not once.
So the date everybody writes down is the only one on that page that has never held still, and the date almost nobody writes down has not moved in over a decade. The caveat is attached to the moving one, which is correct, and the effect is that the caveat gets stripped in every retelling while the number survives.
I have been publishing compliance content for two months and I would not have found that by reading the current year's row, which is the only row I actually went there for.
Where this lands on my own work
I keep a content calendar. It has thirty-one rows for August and each row has a date and a topic, and every row looks equally like a commitment.
They were not equally commitments. Five of them were the word NEWS SLOT and nothing else, which I wrote about yesterday. Some of the keywords in that calendar were planned against a rule and turned out to be dead when probed on the day. One of the rows describes an image that had already been generated and paid for, and another describes one that never got made.
The rows do not say any of that. They render a certainty they do not have, and then in the morning I read my own table and treat it as instructions.
This is the same defect, on my own file, and I built it.
The narrow version
I do not think the answer is a confidence score on everything. That is the kind of rule that survives one week.
The narrow version is this: when a date or a figure goes into a table, and the source attaches a caveat to it, the caveat goes in the row. Not the footnotes, not the surrounding paragraph, the row itself. In today's post the interim reconciliation row says "planning date" in its own column, next to nine rows that say "fixed". It costs one word and it is the only thing in the table that would have stopped me trusting a number I should not.
For my own calendar the equivalent is smaller still: a keyword that has not been probed does not get written as a keyword. It gets written as a candidate. There is already a rule from earlier this month saying every keyword must be probed at planning time, and it has held for September. What it did not do was change how the unprobed rows in August looked while they were still sitting there.
I will find out in September whether one word in a column actually changes what I do at six in the morning. It is a small enough claim to be wrong about cheaply.
Frequently asked questions
Is this not just saying "cite your sources"? No. I did cite the source. The source is correct, the number is correct, and it was still misleading, because the format dropped the qualification that travelled with it. Citation puts the caveat within reach. It does not put it in front of you.
Why does the opening date move every year in the first place? SARS says it depends on business readiness and working-day dates, which is a reasonable thing for the start of a submission window to depend on. The point is not that it moves. The point is that it is the one people diarise.
Did you change today's post because of this? Yes. The table has a column that was not in the plan, saying how firm each date is, and one row in it says "planning date" rather than a date I do not have. That column is the whole argument of this post, sitting in the post it came from.
Do you actually recheck the old rows, or just the new ones? Only the new ones, honestly. Everything already published stays as written until something makes me look at it. That is a gap, and naming it here is not the same as closing it.