SGX Financial Report & Disclosure Document Watch

Singapore Exchange financial-report and disclosure documents as they are lodged: annual, interim, quarterly, business-review, sustainability and summary reports plus information and disclosure statements, with the issuer, the document date and the SGX document link. SCOPE: this is the open financial-reports document feed. It is NOT the full SGX company-announcements stream, which SGX keeps behind an authorised API.

Source

SGX public financial-reports document feed (public, logged-out)

• https://api.sgx.com/financialreports/v1.0

Public, logged-out only — no API key, login, cookie or CAPTCHA.

Two dates, and which one means "new"

Every record carries both, and they are not the same thing:

field	what it is
document_date	the reporting period end the document covers — and the feed's sort key
broadcast_date	when SGX actually published it — the arrival clock

The gap between them is large and normal: measured across 6,000 records on 2026-07-27, the publication lag runs to a median of 105 days and a p90 of 114. SoftBank Group's annual report, for instance, carries document_date 2026-03-30 (its 31 March financial year end) and was broadcast on 2026-07-27 — SGX's own announcement page for it says Period Ended 31/03/2026 and Date & Time of Broadcast 27-Jul-2026 16:30:17.

Both dates are in the change signature. A watch that only signed document_date would call a four-month-old period date "stale" on the day the document was published.

How deep this Actor reads, and why it isn't a page count

SGX sorts this feed by document_date. Because publication lags the period date, a document broadcast today is not near the top of the feed — measured, a trailing-30-day arrival sits at median rank 103 and was found as deep as rank 2,362.

So depth is decided by broadcast coverage, not by a page number. This Actor keeps reading until 10 consecutive pages carry no document broadcast inside broadcast_window_days (default 30). Ten is measured, not guessed: the largest observed gap between two pages carrying a trailing-30-day arrival was 6 pages.

pages is a hard cap (default 40 = 4,000 records) that bounds the read if coverage is never reached. Hitting that cap is reported as a named PARTIAL — "DEPTH CAP" — never passed off as a complete read. Raise pages or narrow broadcast_window_days if you see it.

For reference, an earlier build of this Actor read a fixed 2 pages of 100. On the day it was measured that window saw 89 of the 94 documents broadcast in the trailing 30 days; the other five were not late, they were permanently invisible.

Pricing

Event	Price (FREE tier)
Actor start (apify-actor-start)	$0.02 one-time
Source check (source-check)	$0.10
Per delta (report-document-delta)	$0.05

source-check ladder, GOLD-floored: $0.10 / $0.09 / $0.08 / $0.067. Charged once per run after the source enumeration terminates — seed runs and genuine-zero runs included. It is not charged when the source refused the read: a run that never read the source has not performed a source check.

The per-delta value event (report-document-delta) is the primary event and is priced at $0.05 per delta at the FREE tier. Ladder, GOLD-floored: $0.05 / $0.045 / $0.04 / $0.0335 (FREE / BRONZE / SILVER / GOLD-PLATINUM-DIAMOND). source-check still charges; it is no longer the headline event. Billed deltas are capped at 500 per run and any overflow is reported as a withholding rather than charged.

Snapshot mode is a PREVIEW

watch_mode: false returns at most 25 records, hard-capped in code regardless of any input, labeled output_mode: "preview". This actor sells change detection, not bulk record export.

Watch semantics

The first watch run stores a private baseline and emits zero deltas. Later runs emit NEW / CHANGED / REMOVED keyed on the source-native id with before / after and changed_fields. Billed deltas are capped at 500 per run; the overflow is reported as a named PARTIAL withholding, is never billed, and is re-detected on the next run.

Terminal states

NORMAL · SUCCEEDED-0: GENUINE_EMPTY · SUCCEEDED-0 (no change) · PARTIAL · BLOCKED. A zero with no evidence FAILS loud rather than exiting 0 quietly.

This feed is seasonal. Budget for the season, not the average.

Arrivals are measured on the broadcast clock, and they are lumpy by construction: SGX issuers file annual reports in a cluster a few months after their financial year ends, and 31 March is the common one.

Measured 2026-07-27 across 6,000 records:

window	documents broadcast	rate
trailing 7 days	4	4.0 / week
trailing 30 days	94	21.9 / week
trailing 90 days	249	19.4 / week

The daily shape inside that 30-day window shows the season plainly — 9, 12 and 13 documents on 13, 14 and 15 July, then 4, then 1, then roughly one a day through to the end of the month.

So the 90-day rate is the one to budget against, not the 30-day peak and not a quiet week. At $0.05 a delta the value event carries roughly 89% of a weekly-run bill in season and about 62% out of it; on a daily schedule those become 54% and 19%. Either cadence is defensible here — unlike the low-frequency watches in this fleet, this feed genuinely moves most days in season, so a daily run buys real timeliness rather than paying for checks.