Est. reading time: 7 minutes

Pinterest almost always enters the media plan as a social line item. The team that runs Meta takes it on, the playbook that works there comes along (broad audiences, hook-first creative, a verdict by day seven), and six weeks later the channel is written off as a test that didn’t pan out.

When those accounts get a second look, the channel usually wasn’t the problem. Pinterest’s delivery system matches ads to queries typed by people planning purchases weeks or months out, which makes it search inventory that got bought and judged as feed inventory. The Meta habits themselves are sound, they overlap heavily with what we look for when we audit a Meta ad account, but they were built for a different delivery system.

The fix is planning, not spend. Keyword structure, launch timing, creative, and measurement all change once the channel gets filed correctly.

Pinterest ads run on a keyword auction. Build the account around queries.

Keyword targeting on Pinterest works the way a search advertiser would expect. You attach keywords to an ad group, and your ads become eligible when people search those terms.

Match types carry over too. Broad match is the default, quotation marks set phrase match, brackets set exact match, and negative phrase and negative exact keep the wrong searches out. An ad group holds up to 20,000 keywords, and Pinterest recommends at least 25 per group.

The queries are the reason to bother. Pinterest reports that 97% of its top searches are unbranded, meaning the person typing “postpartum workout plan” or “small backyard shade ideas” hasn’t chosen a brand yet. On Google, the commercial terms in most categories carry years of entrenched competition. On Pinterest, the same intent sits in category language, and it’s winnable by brands nobody has searched for by name.

The structure that follows is a search structure. Build one ad group per query theme, tight enough that a single creative concept can honestly answer every keyword in it, and review keyword-level results on a schedule, moving proven converters onto exact match and adding negatives where broad match drags in the wrong intent. When “linen duvet cover” and “coastal bedroom ideas” share an ad group, neither the bid nor the creative can be right for both, and the reporting can’t tell you which intent is producing buyers.

Interest targeting still exists, and it’s the piece that feels familiar from paid social. Use it as a discovery layer around the keyword core rather than the core itself. An interest describes who someone tends to be. A keyword captures what that person is trying to get done, and that’s the signal worth structuring the account around.

Seasonal demand shows up months early. Launch on the searcher’s calendar.

Pinterest documents its own planning behavior. Searches for big seasonal moments begin months before the events, and the platform’s guidance is to have content live four to six weeks ahead of a moment at minimum. The purchase happens near the event. The searching, saving, and shortlisting happen well before it, and that early window is where a brand enters the consideration set.

The reflex from other channels is to launch the seasonal campaign when the promotion starts. Run that timing here and you arrive after the boards are built, spending to change decisions instead of shape them. Pinterest Trends shows the query curves directly, including when a term started climbing last year and when it peaked, filterable by age, interest, and location. Read the curve for your season and start spending while it’s still rising, not when the calendar says the event is close, with budget weighted the same way, heavier on the climb and tapering after the peak.

Starting early also solves a delivery problem. Campaigns need conversion volume to optimize, and volume accumulates over weeks, so a campaign launched at the peak burns its learning period during the only days that mattered. Gift categories get the obvious version of this calendar. The quieter version applies to anything people plan (weddings, renovations, travel, back to school), and the query data will show you when that planning starts.

Design the Pin to answer the query it appears beside

On the search surface, your ad renders inside a grid of results for the query, most of them organic. Nothing about that placement rewards interruption. Someone asked for “small entryway storage ideas” and every tile on screen is competing to be the best answer, so a repurposed Meta ad (a talking-head hook, a trend format, a hard pattern interrupt) shows up as the one tile that isn’t answering it.

The build is closer to a search listing than a social ad. Pinterest’s own creative guidance starts with a vertical 2:3 ratio (1,000 by 1,500 pixels) so the tile isn’t cropped or shrunk on mobile, where most Pinterest use happens, and it recommends visible branding on every ad. Add a subject that reads at thumbnail size and a short text overlay written in the query’s own language, “entryway storage for renters” rather than a tagline. Saves raise the stakes on that clarity, because a saved Pin resurfaces later, far from the campaign that paid for the first impression.

Match the creative to the ad group the way search advertisers match ad copy to keywords. If the ad group is built around “coastal bedroom ideas,” show the product inside a coastal bedroom, not on a white sweep. This is the practical reason to keep keyword themes tight, since loose themes make query-matched creative impossible, and query-matched creative is what earns the click.

Set the measurement for purchases that arrive well after the click

Pinterest’s own campaign documentation is direct about the tempo. It describes usage patterns as deeper and less frequent than on social and messaging platforms, and notes that its delivery system typically needs 50 to 200 conversions per week to learn who to show ads to. Both points argue against the standard social verdict, a week of spend, a last-click glance at analytics, and a shutoff.

Reporting windows deserve a deliberate choice. Pinterest reports conversions on adjustable click and view windows and defaults to a 30-day click with a 1-day view when you haven’t chosen one. Set the windows to fit your consideration cycle, and hold the verdict until at least one full planning-to-purchase cycle has run. Expect your analytics platform to disagree along the way, because a last-click model hands credit to whichever touch sat closest to checkout, and on a channel where the save precedes the purchase by weeks, that’s rarely Pinterest.

The discipline is the same one we laid out in how to read Google Ads data without fooling yourself. Know what each number can and cannot see, and pick the report you’ll trust before the campaign launches, not after the numbers disappoint someone.

Give the delivery system enough volume to learn

That 50 to 200 weekly range also dictates structure. Ten ad groups splitting forty weekly purchases will sit in learning indefinitely, so consolidate until each ad group can plausibly clear the range, then split as volume grows. And feed the system cleanly, with the Pinterest tag on site and its server-side conversions API passing events, because delivery optimization is only as good as the conversion signal that reaches it.

None of this asks the team to learn a new craft. It asks the plan to file the channel correctly before the first dollar moves, because a channel filed under paid social inherits paid social assumptions at every step, budget, timeline, creative, and verdict. Give Pinterest a keyword structure, a launch calendar built off query curves, and a verdict window matched to how its users buy, and the line item that kept disappointing as social finally gets judged on the mechanics it runs on.