Official certified partners of Google & Meta
The client came to us with a situation typical of Instagram businesses: there's a product, there's demand, but there's no system. Previously, ads had been run only with the “Boost” button right in the app, without any analytics — no one tracked how much a lead cost, where buyers came from, what worked and what didn't. Sales were closed manually in Direct messages, and the brand had no website at all.
Our goal was to build a measurable sales system from scratch: connect a proper ad account, teach the client's team to record leads and conversions, and, based on this data, gradually scale results — lowering the cost per purchase and growing the number of orders.
The first thing we ran into, even before launching the first campaign, was that the client physically didn't have their own Meta Business Manager ad account: the Instagram account was registered to an individual with no way to add partners. Instead of waiting for this to be resolved, we launched campaigns through the agency's ad account, arranging for the budget to be topped up directly, and in parallel started preparing the move to the client's own account.
The second problem was just as critical: without a CRM and without traffic source tagging, it was impossible to tell which buyer came from ads and which from organic. We introduced a simple but effective system: “lead” / “paid” labels right in Direct conversations and a shared tracking spreadsheet synced with the client's team. Later, when basic conversion data wasn't enough for analysis, we reworked the spreadsheet to cover the full sales funnel, so we could see exactly at which stage inquiries were being lost.
Then the main work began — systematic creative testing. We ran each product line (leopard and ribbed sets, lace sets, tights, seasonal turtlenecks) through a sequence of formats: static → video → UGC, constantly replacing weak ad creatives with new ones. In parallel, we also tested a discount hypothesis — bundles like “2 packs for a fixed price.” The result: discount offers consistently lost to regular creatives, confirming the client's own previous experience. We didn't push the idea further; we quickly closed this hypothesis and redirected resources to testing new categories.
The hardest period was spring: from March to May, the cost per purchase crept up — from $2.83 to a peak of $8 — despite a stable budget. A weak season in the niche coincided with the existing creatives having run their course. We needed to keep the work steady and systematic: testing new categories, switching off and retesting weak items (tights were relaunched several times) and searching for a new way to present the product.
The turning point came in July, when we refreshed the creative for the seasonal “turtlenecks” category with a voiced-over video format. The effect was visible to the naked eye: in literally one week (July 13–19), purchases grew from 37 to 59, and the cost per purchase dropped from $3.00 to $1.90.
Building on this success, in August 2026 we helped the client take another important step — launching their own website on the Horoshop platform, setting up GA4, Google Tag Manager and the Meta Pixel, and starting to test ads not only for messages but also driving traffic directly to the website. This is a second sales channel that opens the brand's path beyond Instagram Direct.
The numbers speak for themselves. The cost per purchase, which reached $8 in spring, dropped to $1.11–1.45 by August — a decrease of more than 3.5x.
ROAS in the ad account grew from 240–550% in spring to 1,200–2,100% in August, and if you count total ROAS including sales outside the ad account, in the best weeks of August it reached a fantastic 3,700–4,100%.
The strongest week of the entire period was August 11–17: 102 purchases in the ad account, a cost per purchase of $1.11 and a total ROAS of 4,113%.
This case shows that a systematic result is almost never built on a single successful creative or a lucky launch. What worked here was a combination of factors: honestly working within constraints that can't be removed in a day (no ad account, manual tracking without a CRM, selling without a website), a willingness to quickly drop hypotheses not supported by data (discounts) and, most importantly, the ability to react to stagnation in time rather than keeping a dead creative running just in case. It was replacing one stagnant video in July that became the point from which the cost per purchase started falling.
Scaling to the brand's own website at the end of the period is the next level of growth on top of what we built inside Instagram Direct during the first six months of our collaboration.
