Corrections

Every Zeltru Report edition stays as it was published. When a figure’s meaning, a label or a page’s arrangement is corrected afterwards, the correction is recorded here, dated, with a link to the part of the edition it concerns.

The Zeltru Report, Edition No. 1 (September 2026)

• 2026-09-16 · Concerns Section 3, Section 8

  In Section 3, the defining sentence was removed and the section now links to Section 8. This addresses arrangement only. It resolves no other outstanding item and does not close this edition’s revision window.

• 2026-09-17 · Concerns Section 3, the longest-listed postings

  Four rows in Table 2 and the 5,002-day figure quoted in Section 3 come from one platform, Lever, whose board publishes the date a posting record was created rather than the date the listing was published. No figure here has changed; what those particular numbers measure has. The full note is the 2026-09-17 entry on Lever, below.

• 2026-09-17 · Concerns the whole edition and its Brief

  The email sign-up box and its heading were removed from this edition and from its brief version. This addresses arrangement only. It resolves no other outstanding item and does not close this edition’s revision window.

• 2026-09-17 · Concerns Section 3 and Table 2

  Four rows in Table 2 and the oldest-posting figure in Section 3 come from one hiring platform, Lever, whose board supplies a single date — the date the posting record was created, not the date the listing was published. A record can predate its listing by years, and it survives the listing being taken down and put back up, so an age computed from it measures the record’s age rather than how long the posting has been open. The rows affected are Table 2 rows 1, 4, 5 and 10, and the 5,002-day posting quoted in Section 3. No figure in the edition has changed: each remains exactly what the board reported at press time, and the flag on row 1 already told readers its date predates the board’s own existence. What has changed is what we claim those particular numbers measure. The site no longer publishes an age of any kind for a Lever-sourced listing, and the longest-open ranking is now built only from boards whose dates are publication dates. The evidence is set out at what a posting date means. This corrects a label, not a figure. It resolves no other outstanding item and does not close this edition’s revision window.

• 2026-09-17 · Concerns the Brief

  The Brief’s oldest-posting line, and four rows of the full edition’s table of longest-listed postings, come from one hiring platform, Lever, whose board publishes the date a posting record was created rather than the date the listing was published. A record can predate its listing by years and survives the listing being taken down and put back up, so an age computed from it measures the record’s age rather than how long the posting has been open. No figure has changed; what those numbers measure has. The site no longer publishes an age of any kind for a Lever-sourced listing. The evidence is set out at what a posting date means. The full note is the 2026-09-17 entry on Lever, above.

• 2026-09-22 · Concerns Section 1, the instrument

  Figure 1 shows 477,481 postings carrying an employer date, labelled 95.6%, in a figure whose first bar is the 503,503 tracked total — so the percentage reads as a share of that total, and it is not. It is computed over 499,250 postings, the population of the coverage query at the moment it ran, which this edition stored but never displayed. Against the 503,503 tracked total the same numerator is 94.8%. Both published numbers stand and neither has been changed; what was missing is the denominator the percentage belongs to, and it is now stated. The two populations are read at different moments from the same collection, which is also why they differ.

• 2026-09-22 · Concerns Section 3, the longest-listed postings, Section 2, Section 5

  Every age in Table 2 is counted to August 31, 2026, one day later than the August 30, 2026 Table 2 and the rest of the edition are stamped with. All ten rows are exactly one day larger than the same arithmetic run to August 30, 2026: a posting dated 2017-06-01 is published here as 3,378 days, which is the count to August 31, 2026; to August 30, 2026 it is 3,377. The same is true of every other age figure in this edition: all were computed together on the morning of August 31, 2026, so the distribution, median and shares in Section 2 and the tables in Section 5 are counted to that day as well. Recomputed from our records, the Section 2 distribution and the platform table in Section 5 match their published values exactly when counted to August 31, 2026; counted to August 30, 2026, the median would be one day lower. An earlier version of this note said the distribution was not affected; it is. The figures are left as published.

• 2026-09-22 · Concerns Section 4, the removal lifecycle

  The 37 days stated in Section 4 is the elapsed count from July 24, 2026 to August 30, 2026; the window spans 38 calendar dates, both endpoints included. An internal note recorded 37 as the inclusive count, which it is not. Every removal figure in Section 4 is computed over the days themselves and is unchanged.

• 2026-09-22 · Concerns Section 5, industry, platform and geography, Section 8

  The Lever row of Table 4, and the reading of platform differences in Section 5 and in Section 8, rest on ages that are not posting ages. Lever’s board publishes the date a posting record was created rather than the date the listing was published, so its 2,429 postings — 0.51% of the 476,618 in the age statistics — carry record ages wherever a record predates its listing. That is what makes Lever’s 83.2% and 60.4% the highest in the table, and it means the gap between Lever and the other platforms is not evidence about how long those employers’ listings stay up. The same applies to any comparison drawn from it. This extends the note under Table 2, which covered four named rows, to the aggregate. Every figure in this edition stands as published; what the Lever figures measure has changed.

• 2026-09-22 · Concerns Section 8

  A question in Section 8 was reworded. It used a common term for long-listed postings that asserts why an employer keeps a listing up, which this report does not measure; it now asks whether these figures show why postings stay up. Its answer is unchanged, and so is every figure in this edition. This addresses wording only. It resolves no other outstanding item and does not close this edition’s revision window.

• 2026-09-24 · Concerns Section 1, the instrument

  This edition counted 27 postings — 0.006% of the 476,618 — from five boards since found not to belong to the employers they were counted under. Four were boards of other companies with similar names, counted under Strategy, Axon Enterprise, Quanta Services and Zebra Technologies; the fifth held a single test requisition on a board counted under Aon. What changes at the precision printed: counts only. The posting total falls by 27 to 476,591, and the counts listed 30, 60 and 90 days or longer fall by 20, 16 and 14. No percentage and no percentile changes: the median stays 28 days, the quartiles 13 and 67, and the shares stay 48.5%, 28.1% and 18.8%. Strategy, Axon Enterprise, Quanta Services and Aon are among this edition’s 1,037 employers only on the strength of those boards and hold no other posting here, so a corrected employer count is four lower; Zebra Technologies keeps 247 postings of its own and is unaffected. The figures stand as published. Edition No. 2 uses the corrected attribution on both sides of its comparison, so this fix cannot read as a change in the market.

Edition No. 1’s figures are as of August 30, 2026; the notes above say what changed and what did not.