N2S, D2S and S2S: the arbitrage models

Arbitrage models are named by where the visitor is bought and where the money is earned. The second half is always "search", meaning a search feed. The first half is the traffic source. The abbreviations look technical, but each simply answers the question: what was this person doing a moment before they arrived?

A café can find customers by handing out flyers on the street, putting up a poster at the station, or paying to be listed first on a map app. The coffee is the same. The cost of each customer, and how thirsty they are when they walk in, is completely different.

The four models

Native-to-search

Native advertising places sponsored headlines in a content recommendation widget, the "you may also like" grid under an article. N2S was the workhorse of the 2010s. Its strengths are scale and price. Its weakness is intent: a person who clicks a curious headline is rarely about to buy, so the article and the search terms must do the work of finding the minority who are. Placement quality varies enormously between host sites, so operators lean heavily on a blocklist of sites that send poor visitors.

Social-to-search

Paid social works the same way with a different engine behind it. The platform's algorithm chooses who sees the ad, guided by the revenue signals the advertiser feeds back. That feedback loop is powerful: tell Meta which clicks earned money and it finds more people like them. It is also the main source of account risk, because social platforms police ad content closely and an account ban stops everything.

Display-to-search

D2S covers banner, video and discovery inventory. In practice the term is now used most often for campaigns bought through Google's own ad products, where the same company sells the traffic and supplies the feed. That does not make policy lighter. The ad still passes Google Ads review, and the landing page still has to satisfy the AFS and RSOC rules.

1. An ad in the feed: Asha is scrolling a social app and sees an ad: “SUV lease deals compared for 2026”. The arbitrageur paid the app to show it. This is the paid traffic half of Social-to-search.
2. She taps it: The tap costs the arbitrageur about $0.20 (illustrative). Asha lands on a content page: a real, readable article about leasing an SUV, on a site the arbitrageur runs.
3. The article offers related searches: Inside the article sits a labelled block of related search terms such as “SUV lease deals near me”. This is Google’s RSOC unit. The page also tells Google the wording of the ad she came from (referrerAdCreative) so the terms can be checked against it.
4. First click: choosing a topic: Asha taps one term. Nobody is paid for this click; it simply tells the system what she wants to search for. The share of visitors who do this is the lander CTR.
5. A results page with ads: A search results page opens on the same site. Its sponsored listings are supplied by the search engine and marked as ads. Advertisers who bid on that phrase compete to appear here.
6. Second click: the one that pays: Asha clicks a leasing company’s ad. The advertiser is charged, say, $1.00, and a share flows back to the arbitrageur. Two clicks, one paid: that is the two-click flow.
7. The advertiser judges the result: If Asha asks for a quote, the advertiser got what it paid for: an advertiser conversion. If visitors like her rarely do, the engine learns that this traffic is worth less, and prices it down or cuts it off.

Search-to-search

S2S is the oldest model and the one closest to pure arbitrage. The operator bids on a keyword where clicks are cheap, and sends the visitor to a results page whose ads pay more. The visitor already has intent, so conversion for the end advertiser can be good. Margins are thin because both prices are set by the same kind of auction, sometimes by the same search engine.

It is also the most constrained. The buying side has rules on destination quality: search engines do not want their ad clicks to land on pages that are mostly more ads. The selling side has rules on how a query may reach the results page. Feeds usually require explicit approval before any S2S traffic is sent.

One click or two

Some operators add a Pre-lander, an extra page before the article, to warm the visitor up or to filter out the uninterested. Each added step loses people, so it only pays when those who remain are worth more.

The economics side by side

An illustrative comparison, with invented numbers. An N2S campaign buys visitors at 8 cents; 12% reach an ad click at an RPC of 90 cents. RPV is 0.12 × 0.90 = 10.8 cents, a spread of 2.8 cents. An S2S campaign buys visitors at 45 cents; 55% click an ad at 95 cents. RPV is 0.55 × 0.95 = 52.25 cents, a spread of 7.25 cents. S2S earns more per visitor but there are far fewer cheap keywords to buy, so N2S can still produce more total profit. Neither figure is a benchmark.

Risks by model

• N2S: uneven placement quality, accidental clicks on mobile widgets, and headline styles that drift into clickbait.
• Social: strict ad review, sudden account loss, and creatives that tire quickly (ad fatigue).
• D2S: broad reach with weak intent, and placements on apps where clicks are often unintended.
• S2S: thin margins, strict rules on both sides, and exposure to a single engine if it both sells and buys.