A stable multi-year seasonal ratio is a strong structural-decline detector — when the ratio breaks, the cause is not the market

Claim

Growth rates are contaminated by everything at once — price, volume, mix, macro, share, one-offs. A seasonal ratio divides most of that out, because the numerator and denominator come from the same year, the same customers, the same pricing and the same competitive position. What remains is the calendar.

Budget calendars are sticky. Advertisers spend more in Q4 because of holiday retail and brand-budget flush; enterprises spend more in Q4 because of fiscal-year-end. Those behaviours survive recessions. So when the seasonal ratio breaks, the explanation cannot be "the market is soft" — a soft market lowers both halves.

Evidence — TTD, guided H2 2026

Rebuilt from XBRL back to the IPO era, not just the three years the vendor feed carries:

Year	H1	H2	H2/H1
2018	$198.0M	$279.3M	1.411
2019	$280.9M	$380.2M	1.354
2020	$300.1M	$535.9M	1.786
2021	$499.8M	$696.7M	1.394
2022	$692.3M	$885.7M	1.279
2023	$847.1M	$1,099.0M	1.297
2024	$1,075.8M	$1,369.0M	1.272
2025	$1,310.0M	$1,586.2M	1.211
2026E	$1,403.9M	~$1,394M	~0.993

Eight years, minimum 1.211 — spanning a pandemic, a 2022 ad recession, and a secular deceleration from +43% to +18% growth. Even extrapolating the visible flattening trend (−0.03/yr) predicts ~1.17 for 2026, implying H2 of $1,643M against a guided-implied $1,394M.

~$250M of H2 revenue missing versus trend ≈ 9% of full-year revenue. Management attributed the weakness to macro (CPG and auto softness). The ratio says the calendar broke, and 2022's ad recession did not break it.

The check that made it stronger: 2026 is a US midterm year. Political advertising peaks in Q3–Q4 and is a tailwind to the very half that collapsed. Correcting for it made the core decline worse, not better — Q3 ex-political ≈ −19% to −21% YoY versus the −12.1% guided.

How to apply

1. Build the ratio from XBRL, across every year available, not from the 3–5 years a vendor feed carries. https://data.sec.gov/api/xbrl/companyconcept/CIK{cik}/us-gaap/RevenueFromContractWithCustomerExcludingAssessedTax.json
2. Check the ratio's own trend before calling a break. A slowly flattening ratio is normal as a business matures and diversifies; extrapolate that trend and compare against it, not against the historical mean.
3. Look for a seasonal tailwind in the broken half. If the collapsing period should have been helped (election year, a product cycle, an extra week), the underlying decline is worse than the headline.
4. Distinguish guided from disclosed. Companies usually guide one quarter. The far quarter of the ratio is often your estimate, not theirs — state that, and give the sensitivity. On TTD, a Q4 of $850–900M instead of the assumed $744M moves H2/H1 to 1.10–1.14, changing "unprecedented in eight years" to "worst in eight years." That difference was worth ~12pp of scenario weight.
5. The inverse is also useful. A company whose seasonal ratio holds through a bad guide is more likely genuinely cyclical, and more likely to mean-revert.

What would falsify this

A genuine change in the customer mix or business model that legitimately alters the calendar — e.g. a shift from advertising to subscription revenue, or an acquisition with an opposite seasonal profile. Check for those before treating the break as diagnostic. Absent such a change, the calendar is not something a soft market moves.

Related

• [[pitfall-contra-revenue-reclass-inflates-revenue-growth]] — correct the series for presentation changes before computing the ratio, or the break may be an artifact.
• [[principle-primary-source-beats-vendor]] — the XBRL API is the source that makes the long series possible.

History

• 2026-08-07 — the Fundamentals agent found the erasure over three years; the Moat (quantitative) agent extended it to eight via XBRL and added the political-tailwind check, which is what converted it from a suggestive observation into the decisive one.