Trove Boutique Curated Intelligence
Piece I · Trove Curate · Issue №07

Trove Curate.

Editorial-grade category pages, edits and lookbooks — generated from your catalogue and your point of view. Imitates the voice of your buyers, not the voice of an AI. Paginated like a magazine spread; measured by scroll-depth as well as conversion.

Apply for an atelier slot Read the method
The argument

A filter grid surfaces products.
An edit surfaces an argument.

Argumentless commerce is high-friction and low-trust. The boutiques whose category pages read like a magazine spread convert better than the ones whose category pages read like an inventory dump — not by a small margin, and not by accident.

01 · Voice
Imitates your buyers, not the model.
Trove Curate is conditioned on a small set of curator-authored reference edits. The generated voice matches the shop, the season and the audience. There is no detectable house style underneath; nothing reads as a machine spoke first.
02 · Layout
Paginated like a magazine spread.
Product placements respect the reading rhythm — the first piece is given room to argue for itself, the next four sit in a quieter grid, the closer earns full width again. We measure scroll-depth, not just clicks.
03 · Grounding
Edits cite your catalogue and your context.
Every claim is grounded in the SKU description, the buyer's notes or a verifiable supplier statement. Nothing is invented. Where the catalogue is silent, the edit is silent; we will not write around an absence with vague prose.
How it runs

From catalogue intake to first published edit, in seventeen days.

WeekPhaseWhat the desk doesWhat you receive
W0IntakeCatalogue ingest. Voice calibration against six to twelve reference edits authored by your buyers.Voice calibration report.
W1First sketchThree category drafts, one lookbook draft. Reviewed against the calibration set; rewritten where the voice drifted.Four drafts with line-by-line provenance trail.
W2PolishBuyer mark-up addressed. Layout pagination tested on three viewports.Press-ready edits and a documented style sheet.
W2 · Day 17PublishEdits go live on the shop. Scroll-depth and conversion telemetry begin streaming back.First published edit; live telemetry dashboard.
W3+SteadyWeekly fresh edits. Quarterly voice re-calibration as the buyers' point of view evolves.One to three new edits per week.
Pilot result · independent design boutique · Hackney Seven weeks in, the boutique's "Objects for the table" category saw average scroll-depth move from 38% to 71% of page height; units-per-transaction on the category lifted by 14.3%. The buyer described the published voice as "exactly what we'd write if we had time to write it." No edit was published unreviewed; the reviewing time was, in her phrase, "twenty minutes per piece, against the day it used to take."
Where it works · where it doesn't

Trove Curate is exact for some shops, wrong for others.

Works best
Independent retailers with a point of view and a catalogue that benefits from context — design objects, specialty food and drink, fashion, lifestyle, books, archival prints, homewares. Catalogues in the three-hundred-to-thirty-thousand SKU range.
Wrong for
Pure-utility commerce. If a category sells fasteners by the millimetre, you do not need an edit; you need a fast filter grid. We will tell you so during the application call, and we will refer you to a better-fitting tool.

Apply for an atelier slot