We took Cristy Cali from $360K a year to a 7-figure-a-year business.

Cristy Cali is a founder-led artisan jewellery brand in New Orleans with a private showroom, trunk shows and a bracelet subscription. The engine had stalled. In the year to 11 July 2025 the ads spent $97,281 and returned $319,000, a 2.5x, against $94,225 and $281,000, a 2.9x, the year before. Gross profit from ads was down $5,000.

Pressure came from outside the account too. Silver rose 140% in six months and tariffs of 30% to 50% hit a handmade line. Acquisition had fallen two years running while revenue held up on bigger orders: 3,248 new customers in 2024, 2,759 in 2025, about 1,823 on the 2026 pace. 68% of buyers order once and never return. Working the existing base was being mistaken for a strategy.

Inside the account was a mechanical fault nobody had seen. The December 2025 prospecting campaign excluded past purchasers using pixel audiences that had collapsed to about 20 people each, so it excluded almost nobody from January to June 2026. $22,229.68 of cold budget landed on 8,108 existing customers at a frequency of 137.05. The cause was not a broken rule: only about 6% of the account’s Purchase and PageView events carry a browser cookie, so website audiences never fill. June came in at 1.79x against a 1.82x break-even.

Step #1 Rebuild the exclusions on data Meta can match.

On 7 July we retired that campaign and built separate retention and retargeting campaigns on customer lists uploaded from Shopify. On 11 August we uploaded the all-time buyer list pulled live from the Shopify admin, 15,796 customers with 0 invalid records and $5.16M of lifetime value, and repointed the account-level existing-customers segment at it. Retention above retargeting above prospecting was locked the same day, with 27 exclusions on prospecting.

Step #2 Give the corrected audiences something worth watching.

Four founder and brand videos went live on 16 July and returned 21.55x on about $60 in the first week, then 11.58x on $245.22 across the month. Eight lifestyle films cut into 24 ads launched on 5 August and returned $5,838.32 on $413.37, still 6.46x with the outlier stripped out.

Step #3 Answer the audit in the tool it came from.

In August the founder commissioned an independent review of the account. We reproduced all five of the reviewer’s headline numbers before disputing any of them, conceded the exclusion leak harder than the review had stated it, and disclosed a finding it had missed: cold prospecting had spent $82,630.47 for 824 purchases at a $100.28 cost per order and $137,964.78 back, a 1.67x at or below break-even. Warm revenue had been subsidising acquisition rather than the reverse. One claim in our own first draft was wrong. Simon pushed back, we opened the interface, the number was there, and the draft was replaced.

Step #4 Report the number that survives scrutiny.

Meta’s claimed revenue for July was up 245%. The monthly report leads with store sales per dollar of paid spend instead, $8.32 in June against $10.54 in July, and says in writing that it is not a 245% gain and we will not report it as one.