Sponsored formats — spec
Table of Contents
1. Purpose
The wal.sh adtech surface models seven distinct sponsored formats. Each has its own catalog shape, disclosure convention, and industry antecedent. This spec is the format taxonomy: what each format is, historically, and where it lives in the tree.
The point of naming them is that most public conversation about "native advertising" flattens them all to one category. They are not one category. Each format carries a different disclosure convention, a different physical dimension set, a different pricing model, and a different measurement failure mode. Reading them as one class hides the differences.
2. Format catalogue
2.1. Native (in-list product-shape)
Antecedent: Amazon Sponsored Products [1997- ] and Google Shopping
Ads [2002- ]. Disclosure: a Sponsored badge in the card corner
plus rel"sponsored"= on the outbound link
(WHATWG, n.d.). Failure mode: badge visibility
degrades on mobile; the "in-grid ad indistinguishable from organic
product" pattern is the class of complaint documented by ANA
(Association of National Advertisers 2023). wal.sh unit: product archetype in
sponsored-research (5 SKUs, Amazon-shape).
2.2. Product (SKU card)
The generalisation of native for e-commerce. Distinct from native
in that the card carries product schema (price, reviews, prime-
eligibility). ACOS-priced: advertiser bids per click as a fraction
of the SKU price. wal.sh unit: product archetype; the
:acos ##Inf field is the running joke — every SKU is $0, so ACOS
is undefined, so the "ad" cannot cost anything to serve.
2.3. Listing (real-estate / classified card)
Antecedent: Redfin [2004- ], Zillow, Craigslist. Disclosure: a
"Sponsored" ribbon on the card, distinct visual treatment. Failure
mode: the difference between "sponsored listing" and "featured
listing" and "organic listing" is invisible to most readers; the
ribbon convention was standardised precisely because it wasn't
(Interactive Advertising Bureau 2017). wal.sh unit: listing archetype (4
entries, Redfin-shape).
2.4. Local (slot-0 reserved)
Antecedent: Yelp [2004- ], Google Local Pack. Disclosure: "Ad"
label above the pinned unit; the unit occupies slot 0 above the
organic list. Slot-0 reservation is the industry pattern for local
inventory. Failure mode: slot-0 reservation is non-displacing
by convention (the organic list starts at slot 1), but the
above-fold click share is dominated by slot 0. wal.sh unit:
local archetype (4 entries, Yelp-shape).
2.5. Arbitrage (query card, not a result)
Antecedent: Yahoo "Related Searches" [c. 2007- ], Bing
"related searches", Taboola/Outbrain widgets
(Taboola, n.d.), (Outbrain, n.d.). Disclosure: a
container-level "Sponsored" label; the individual card carries no
disclosure. Failure mode: the click resolves to another ad
surface, not content. Documented in ANA 2023 MFA
(Association of National Advertisers 2023) and Vekaria et al. (Vekaria et al. 2022)
as the primary MFA-inventory pattern. wal.sh unit: arbitrage
archetype; templates × idf-tail hapax terms; deterministic per
(git-sha, query).
2.6. Display (banner)
Antecedent: HotWired 1994 (Zeff and Aronson 1997), DoubleClick 1995 (DoubleClick, Inc. 1998), the entire IAB Standard Ad Unit Portfolio (Interactive Advertising Bureau 2017). Disclosure: "Ad" or "Sponsored" text on or near the banner. Failure mode: banner blindness — 30+ years of measurement literature. wal.sh unit: sponsored-display format with per-banner provenance affordance.
2.7. Exit-intent (overlay)
Antecedent: Bounce Exchange (Wunderkind) [c. 2012- ]
(Bounce Exchange (Wunderkind), n.d.). Fortune's "Before you leave" unit is
the widely-cited implementation. Disclosure: container-level in
the source pattern; per-card in our version (R32). Failure mode:
back-button hijacking, universally blocked by browser policy — the
class of exploit our R28 refuses to implement. wal.sh unit:
exit-intent blended-grid overlay with 6 cards, 50/50 organic/
sponsored.
3. Comparison table
| Format | Disclosure | Displaces organic? | Antecedent | ORACLE |
|---|---|---|---|---|
| Native (product) | badge + rel=sponsored |
yes (in-grid) | Amazon SP | sponsored_research product |
| Product (SKU) | schema + label | yes | Google Shopping | sponsored_research product |
| Listing (card) | ribbon | yes but labeled | Redfin | sponsored_research listing |
| Local (slot-0) | label | no (above result 1) | Yelp / Google Local | sponsored_research local |
| Arbitrage (query) | container-level | n/a (outbound click) | Yahoo Related | sponsored_research arbitrage |
| Display (banner) | on/near banner | no (side rail) | HotWired / DoubleClick | sponsored_display |
| Exit-intent (overlay) | per-card (ours) | no (overlay) | Bounce Exchange | exit_intent |
4. Disclosure conventions (summary)
Each format has a canonical disclosure pattern. The formats disagree on where the label goes, how prominent it is, and whether it applies per-item or per-container. This is the industry state, not a design choice.
- Per-item badge (native, product, listing)
- Per-item ribbon or label position (listing "Sponsored" ribbon)
- Slot-level annotation (local slot-0 "Ad" label)
- Container-level (arbitrage, exit-intent in source pattern)
- On-creative (display banner "Ad" text)
WHATWG's rel"sponsored"= (WHATWG, n.d.) is the
outbound-link standard that cuts across all formats; every wal.sh
sponsored surface uses it.
5. Related literature
- IAB Standard Ad Unit Portfolio (Interactive Advertising Bureau 2017) — the reference for banner dimensions across display.
- Zeff & Aronson, Advertising on the Internet (Zeff and Aronson 1997) — the earliest published survey of what would become the format taxonomy.
- Taboola / Outbrain publisher documentation (Taboola, n.d.), (Outbrain, n.d.) on the arbitrage / recommendation-widget mechanics.
- ANA 2023/2024 MFA studies (Association of National Advertisers 2023), (Association of National Advertisers 2024) on the industry-scale failure modes across formats.
- Zeng, Kohno & Roesner on bad ads (Zeng, Kohno, and Roesner 2020) — the measurement-scale catalogue of format-level abuse.
- Ali et al. on delivery discrimination (Ali et al. 2019) — bias is format-agnostic; each format carries the same auditability failure.
- Bounce Exchange (Wunderkind) (Bounce Exchange (Wunderkind), n.d.) on exit-intent as a distinct format.
rel"sponsored"= HTML Living Standard (WHATWG, n.d.).- Vekaria et al. on ad-inventory pooling (Vekaria et al. 2022) on the arbitrage-format economics at industry scale.
6. Cross-references
- sponsored-research/spec.org — native, product, listing, local, arbitrage archetypes.
- sponsored-display/spec.org — display banner format.
- exit-intent/spec.org — exit-intent overlay format.
- attribution-audit/spec.org — the audit surface every format appends touchpoints into.
- web-history/spec.org — the 30-year survey each format's antecedent references.
- browser-adtech-infra/index.org — the top-level ownership map; the chumbox / recommendation-widget taxonomy this note formalises.
7. Open questions
- Video / VAST format. IAB VAST video ads (Interactive Advertising Bureau 2019) are the eighth format in the industry taxonomy. wal.sh has no video content and does not model VAST; the closest analogue would be an autoplay hover-preview which is deliberately not implemented.
- AMP-only formats. Google AMP (AMP Project 2015) introduced its
own ad-format subset (
amp-ad,amp-ad-exit). wal.sh does not support AMP; the format taxonomy is HTML-only. - Sponsored-audio / podcast dynamic-insertion. Off-topic for a text-first research site; noted for taxonomy completeness only.