How Plum Guide rebuilt retargeting into a measured, brand-worthy growth channel

Plum Guide is a luxury travel platform that curates the world's best holiday homes. Every home on the site has passed a rigorous vetting process. The brand is the foundation of the business, and every touchpoint — including a retargeting ad in a social feed — has to live up to it.

By early 2026, the team saw room to take paid social further. Industry-wide signal loss had narrowed the addressable audience and pushed frequency higher than anyone wanted. Dynamic ad formats made it hard to present the portfolio with the brand's polish. And the question every sophisticated advertiser eventually asks — how many of these bookings are truly incremental? — deserved a properly engineered answer rather than platform-reported attribution.

Plum Guide partnered with Parrot Partners to build on that foundation: to re-engineer the data layer, the creative system, and the measurement framework as one programme, and to prove incrementality before scaling further.

Engineering the data foundation

The first weeks were an engineering project, not a media project. Working in a shared channel with Plum Guide's engineers, we designed a complete event taxonomy — property views, date selection, checkout steps, purchases — flowing through both browser pixel and server-side Conversions API, with deduplication and advanced matching handled correctly from day one. Two dynamic catalogues were built and enriched: the full portfolio of vetted homes with pricing, location, and property attributes refreshed daily across four currencies, and a second catalogue of 700 curated destinations, so that someone browsing Paris could be retargeted with Paris itself, not just a single home.

On top of the clean signals, the campaign architecture follows how people actually plan travel. Destination-led dynamic campaigns re-engage people at the city and region level; funnel-depth campaigns follow individual intent — homes viewed, dates selected, checkouts abandoned — each with its own message, window, and bid logic. Dynamic landing pages take every click to the exact search or listing it came from, and frequency is a deliberate design constraint: capped, monitored, and reviewed with the client's analytics team on a shared dashboard built on a custom UTM taxonomy.

Plum Guide's engineers shipped feed and pixel changes; we validated them in the ad platform within hours; the loop repeated. Work that often takes agencies and clients a quarter to coordinate was live within weeks.

Creative at catalogue scale, on brand

Dynamic product ads rarely flatter a luxury brand: raw catalogue photos, with no voice. We built our own rendering platform to address this. The system takes Plum Guide's live feed, applies overlay templates designed in a browser editor with the client's brand team — typography, price, location, and property attributes composed onto each photograph — and renders every image in four aspect ratios, over 125,000 on-brand images per template publish, served back to Meta through rewritten feeds.

The result is catalogue advertising that looks art-directed at the scale of the entire portfolio: every home, every placement, every format carrying the brand, with a new template deployable across the full catalogue in hours rather than weeks of production.

Measurement from two independent angles

Incrementality was designed in before the first pound was spent, and approached from two independent angles so the answers could be triangulated. At the user level, a holdout audience was silently excluded from all retargeting delivery so booking behaviour could be compared against an identical group that saw the ads. At the market level, a geographic experiment was designed with Plum Guide's data science team: 161 US media markets split into treatment and control, advertising switched off entirely in control regions for ten weeks, and results evaluated with permutation testing against more than a year of baseline data.

The analysis was deliberately conservative. Early weeks that flattered the result were excluded and the test re-run. A structural identification bias between the groups was found and corrected. Planning figures were taken from the most conservative bound the experiment produced, not the headline. The result held throughout: retargeted users booked at roughly three times the rate of the withheld group — a lift of around +200%, significant at p < 0.0001 — with seven figures of incremental sales identified in the base case, at an incremental cost of sale of roughly 3%: around a 30× return on every pound of retargeting spend.

The geo experiment pointed the same direction: markets with ads on out-booked ads-off markets by around +8% and out-sold them by around +20% over the ten weeks, with the permutation analysis confirming a positive signal approaching — though not yet clearing — formal significance at Plum Guide's booking density. The test was extended to accumulate the volume a market-level read requires.

Proven incremental. Retargeted users book at roughly three times the rate of those who see no ads — a +200% uplift, statistically significant — generating seven figures of incremental sales at around a 30× incremental return on spend. Leadership now plans on the most conservative figure the experiment produced.

Directionally supported at market level. Ads-on markets out-sold ads-off markets by around 20% in the geographic experiment — an early, directional signal pointing the same way as the user-level result, with the test extended to reach a conclusive read.

Scaled with discipline. Retargeting now reaches a multiple of its previous audience at a fraction of the frequency, re-engaging the full funnel with intent-matched, on-brand creative. The holdout was deliberately enlarged to nearly a third of the audience, so incrementality is a permanent, always-on property of the channel rather than a one-off test.

The infrastructure compounds beyond one channel: the event taxonomy, catalogues, and rendering platform now serve as reusable acquisition machinery, already extending into new audiences, new markets, and creative programmes co-designed with the brand team.

Parrot Partners has worked with Plum Guide since February 2026. The engagement combines hands-on Meta campaign management with deep technical collaboration: a shared working channel with Plum Guide's marketing, engineering, analytics, data science, and brand teams, ad-hoc syncs whenever a decision or build calls for one, and joint decision reviews at each stage gate.

We contribute strategy, campaign operations, tracking architecture, and our proprietary creative rendering platform; Plum Guide contributes engineering execution, first-party data, experiment analysis, and brand direction. The programme was designed so each side ships what it is best at.