Traffic sources · read at the source
Where you buy traffic, and whether the math works.
The offer index covers where a publisher earns. This covers where they spend: what a channel costs, the smallest useful budget, and whether the commission on the other end can cover it.
Same rule as the offer index: every figure shows the date it was checked, or it is not shown.
The question this section answers
Whether a paid channel pays for itself is decided by the commission, not the channel.
The math is the same for every channel and every program: it works when the clicks you pay for to get one sale cost comfortably less than that sale pays you. An expensive product with a high commission can cover an expensive click. A flat bounty almost never can.
DigitalOcean · Droplets pays $5.00 on a first-year customer. At one sale in a hundred clicks, that is the most a click can cost.
Teachable · Business Suite pays $1,076. The same arithmetic, 215 times the budget.
Reddit Ads, at the bottom of its range, read 2026-09-08. Everything below that line is unbuyable at any commission.
The whole argument in one chart
The most you can pay for a click, against what a click costs
Each bar is the break-even click price for one real commission in the index, if one click in a hundred becomes a sale. The tick marks are what a channel charges per click. A channel clears when its tick sits left of the bar's end. Double the sale rate and every bar doubles.
Commission figures are computed from the index: the rate, the plan price and the window each advertiser publishes. Click prices are only drawn where this site has read one at the source — 2 of 7 sources carry one. The sale rate is yours to set; 1% is a common starting assumption for software, and nobody has measured it for creator traffic.
Before you buy a single click
Check the program's rules first. Some ban paid traffic outright.
A program that forbids paid placement does not make an exception for a page in between: an ad pointed at a page carrying that affiliate link breaks the terms and can cost you the program. Every researched offer page now shows what its program allows, and the index says so on the row. Our own experiments follow the same rule.
What the measured evidence says
Most sales from a recommendation never touch a link.
Podscribe's Q1 2026 benchmark, across 97,000+ podcast campaigns, finds pixel-based matching captures about 4.7× more conversions than promo codes or vanity links. Codes capture roughly one sale in five; the rest arrive later, through a brand search or a direct visit. And when an assistant answers in-chat, eMarketer notes, the tracked link that triggers a commission never occurs at all.
Podscribe is a measurement vendor and its pixel figures include modelled conversions; the code multiple is the stable part of its data across quarters.
Read in full
6 of 7 channels read at the sourceChatGPT Ads
The only large paid surface where the person is mid-question rather than mid-scroll, and comparison questions are the most expensive inventory precisely because they are the most valuable. That is the exact moment a storefront is built for, which makes this the sharpest available test of whether the storefront outconverts a raw link on cold traffic.
Anything needing a real-time read. Reporting on the managed side arrived weekly or biweekly through an account team, and the self-serve dashboard is a beta.
Google Ads (Search)
Buying a named intent. Someone searching a category or a comparison has already decided to decide, which is the same visitor the storefront is written for.
Bidding on the advertiser's own brand terms — most affiliate programmes forbid it outright, and it is the fastest way to lose a programme rather than a campaign.
Meta Ads (Facebook, Instagram)
Volume and cheap reach, and the only place on this list where an audience can be built from a list we already hold rather than found by intent.
Exactly what a storefront is for. The reader arrived to be entertained, not to decide, and a page of researched comparison asks for attention a scroller has not offered. If the storefront thesis is right anywhere, it is right last here — which makes this a useful control rather than a first test.
Reddit Ads
Reaching a niche that has already gathered itself, at the lowest cost per click on this list. Cheap clicks change the arithmetic completely: where a $12 ChatGPT click needs a strong conversion rate to survive a first-month commission, a $1 Reddit click gives an experiment room to be wrong.
Anything that reads as marketing. The audience is hostile to advertising in a way no other platform's is, and a storefront's honesty is either the thing that saves it here or irrelevant — which makes this the most interesting second test after ChatGPT.
beehiiv Ad Network
Borrowed trust. A recommendation inside a newsletter someone already chose to receive is the closest paid analogue to the creator endorsement the storefront is designed to follow — the same shape of attention, bought rather than earned.
Testing anything cheaply, or controlling where the ad lands. The floor is $15,000 and beehiiv places the campaign.
EthicalAds
Reaching developers in the place they are already reading documentation, on a page whose contextual relevance is doing the targeting work. The audience overlaps the densest part of the corpus, and the format — a small unit with text, no interruption — is the one the storefront's own design language came from.
A fast test. There is a human approval step, a four-figure minimum, and no conversion tracking, so the loop from spend to answer is long.
Queued for review
1 listed, none pricedListed so the gaps are visible. Each one gets a full record only once its rates and minimums have been checked at the source.
Microsoft Advertising (Bing)
SearchingPaid results on Bing and its syndicated partners, plus Microsoft's own surfaces. Smaller than Google by an order of magnitude, historically cheaper per click, and with an audience skewing older and more desktop-bound. Long treated as the most affiliate-tolerant of the search platforms, which is the reputation this record exists to check.
No rates are shown above, because none have been checked. Guessing them would make this page no better than the directories it exists to correct.