Targeting precision that died with the cookie —
reborn as declarative wishlists.

The death of behavioral targeting.

It has been almost a decade in slow motion. Now it’s done.

In 2021 Apple shipped App Tracking Transparency, the dialog that ended IDFA as a usable identifier overnight. Opt-in rates settled in the low teens. Meta took a public, double-digit-billion-dollar revenue hit and told investors so on an earnings call. The industry didn’t recover; it adapted with weaker signal, more modeling, more guesses. In 2024 and 2025 Chrome moved through its Privacy Sandbox tracks and the broader market accepted what Safari and Firefox had already done: the third-party cookie is over as the substrate of programmatic targeting. Topics, FLEDGE/Protected Audiences, and the rest of the Sandbox are partial replacements, and they were designed to be partial. They are deliberately less precise than what came before.

What was actually lost wasn’t reach — the open web still has reach. What was lost was intent signal. Behavioral targeting was an inference machine: someone visited a sneaker page, clicked an athletic-wear ad, read a running blog. Conclusion: they are probably in-market for running shoes. Conclusion: show them a running-shoe ad. The entire industry was built on probabilistic guesses dressed up as audiences. When the cookie went, the guesses got worse, but more importantly, advertisers had to start paying attention to a question they had been allowed to ignore for fifteen years: was any of that inference ever accurate to begin with?

There is a well-documented body of research suggesting most behavioral targeting’s lift was a measurement artifact. Cookie-based attribution windows pre-credited the targeted ad for sales that would have happened anyway. The buyer was already going to buy. The ad just sat on the page. When researchers ran proper holdout studies, much of the lift disappeared. The dirty secret of the cookie era is that the industry was getting paid to find people who had already decided.

Gish exists on the other side of that admission. We don’t infer that someone might be interested. We surface the moment they actually said they were — with the variant, the budget, the occasion, and the deadline attached.

We don’t infer intent. We observe it.

At SKU + variant + occasion + budget granularity. Direct from the buyer.

A wish on Gish is a structured object. It includes the merchant, the SKU or product URL, the variant (size, color, configuration), the occasion (birthday, wedding, honeymoon, anniversary, baby shower), the deadline, the budget the buyer is willing to spend, and the social context (whose wishlist it’s on, when it was added, how many people have viewed it). Every one of those fields is something the buyer typed in, picked, or confirmed. None of it is reverse-engineered from behavioral exhaust. None of it requires a cross-site tracking cookie or a device graph.

That changes what an ad can be. A behavioral ad asks: which audience segment is this person closest to? A Gish ad asks: this person literally said they want the Samsung 77-inch S95F OLED. Do you sell that, or a substitute they’d accept? The first is a probabilistic match. The second is a deterministic one.

Eight ways declarative beats inferred.

In every row the advertiser is buying access to declared intent. Not a model’s guess from page-views. Not a lookalike audience trained on someone else’s purchases. The buyer named the thing.

Three ways to show up.

Every Gish ad placement is anchored to an actual declared wish or a category the visitor is browsing in. Nothing is interruptive.

Plus profiles are ad-free, forever.

The economics work because the wishlist owner decides whether their profile carries ads.

Most ad networks force monetization onto every surface, every user, every page. We don’t. The wishlist owner is the customer of the product, and the wishlist owner gets to decide. If they pay for Plus, the profile they share with their friends and family is permanently ad-free — for every visitor, in every session, in perpetuity. That is the deal.

The two surfaces that are monetized are Free profiles and the global explore feed. Both behave like classic ad inventory: high-volume, intent-rich, native-formatted. A visitor browsing the global explore is, by definition, looking for ideas. A visitor on a Free profile is, by definition, looking at someone who didn’t pay to remove ads. Neither is an interruption; both are matched to the wishes already in view.

• Plus owner pays $4.99/mo → their profile is ad-free for every visitor, forever.
• Free owner doesn’t pay → their profile shows one matched sponsored card, capped, native-format. — the cap is one because scarcity is what holds the $100–$250 CPM; more slots would collapse each one to global-explore rates, dilute conversion intent, and make the page read as ad-supported rather than as a wishlist with one relevant suggestion. One slot is also the right pressure for Plus: enough that ad-free has value, gentle enough that Free doesn’t feel punishing.
• Global explore is always monetized — that’s the discovery surface, not a personal wishlist.
• No retargeting cookies. No cross-site pixels. No device graph. Targeting is the declared wish, period.
• Owners never see ads on their own profile, Plus or Free. Self-view is always clean — the owner declared the wish, so there’s nothing to sell them; they’re often editing, which an ad slot would disrupt; and the owner can see their own private wishes, so excluding self-view at the data layer keeps the “private wishes are never advertised against” guarantee architecturally hard, not just policy.

We show you what their friends already see.

Nothing is inferred from private behavior. Nothing is extracted from a hidden tracking pixel.

A wish on Gish is, by default, something the user has already shared with their network — their family, their close friends, their gift-givers. That’s the entire reason it exists. When an advertiser bids on that intent, they’re buying access to the same signal the wisher’s mom can see. The privacy posture is symmetric: no party gets more information than the wisher chose to share.

Private wishes — wishes the user has marked surprise-only, private-to-self, or unshared — are never advertised against. Ever. The fence is hard. The bid model can’t see them; the targeting graph can’t see them; the auction can’t see them. They’re excluded at the data layer, not the policy layer.

There is no IDFA equivalent on Gish. There is no third-party cookie. There is no cross-site graph. We don’t need any of it — the buyer told us what they want. Everything else is theatre.

What it costs to show up.

Indicative bands by category and placement. Final auction prices are dynamic; these are the floors.

These are floors, not ceilings. Brand-week category takeovers and exclusive editorial co-treatment price separately, by sponsorship contract. Developers can hit the same inventory programmatically via the Gish API once advertiser auth is live.

Self-serve dashboard waitlist.

For brands ready to run sponsored explore and profile placements without a managed-service team. Waitlist open; rolling access in cohorts.

Brand-week + managed service.

For category takeovers, multi-market campaigns, and editorial co-treatment. Reach our enterprise team directly.

Closed-loop attribution, finally.

Because a wish is a structured object with a known purchase moment, attribution is deterministic.

The cookie era turned attribution into a forensics exercise. Multi-touch models, view-through windows, MMM regressions, post-iOS-14 modeling — every brand built a Frankenstein stack to answer a question that should have been trivial: did this ad cause this sale? On Gish, the wish is the sale candidate. We know when it was added, when it was fulfilled, by whom, through which retailer link. If your ad was the surface that drove the wish to a contribution, a gift purchase, a group-buy split, or a self-purchase routed through Smart Savings, the loop is closed with a server-side event, not a client-side pixel. No probabilistic attribution. No cookie windows that lapse silently.

We support brand-side conversion endpoints so a retailer’s commerce system can confirm the redemption. We also expose a query API for incrementality testing: hold out a percentage of matched impressions and measure the conversion delta directly, in production, against your own audience. If the lift isn’t there, you stop paying. That is a stronger guarantee than anything the behavioral era offered.

• Server-side conversion events tie the original wish to the eventual purchase, contribution, or group-buy redemption.
• Built-in holdout testing runs an incrementality experiment against your own matched audience — no third-party measurement vendor required.
• SKU-level reporting because the wish is keyed to the SKU. You see exactly which variants you reached.
• Occasion-level reporting slice by wedding, birthday, honeymoon, anniversary, holiday. Which occasions does your category convert in?
• No cookie windows, no view-through guessing — the wish lives until it’s fulfilled or removed. That’s your attribution window.

Side by side with what you’re buying today.

Programmatic display, Meta interest targeting, retail-media networks, and Gish — on the things that actually matter.

This is not a claim that Gish replaces every line of an advertiser’s media plan. Programmatic still has reach. Meta still has scale. Retail-media still owns post-search intent inside the walls of Amazon, Target, Walmart. What Gish offers is something none of those four can: pre-purchase, cross-retailer, variant-level intent — declared, not inferred. That is a new category of inventory, and it didn’t exist last cycle.

Things advertisers ask us.

The five questions that come up in every introductory call.