Building an acquisition system that held at 4x sale-day budgets
Sales broke the algorithm exactly when the business needed it most
App installs, first purchases on app and web, and retention all ran inside shared accounts. When budgets surged for a major sale, three competing objectives sent mixed signals into the algorithm at once. It couldn’t tell a loyal customer from a new one, so it chased the easiest conversion. Learning reset, and days of optimization were lost exactly when the business needed performance.
About 70% of repeat buyers bought again within 30 days because of the sales calendar. Beyond that, most came back through organic and direct. Paid had to focus on first purchases.
Meta’s algorithm learns from the whole account’s history, so splitting objectives by campaign inside one account still blended the signals.
A sale is six weeks away. What do you change?
Make your call to see mine.
I chose four dedicated accounts: app installs, first purchase on app, first purchase on web, and retention.
- CXOs, worried about rebuilding mid-growth
- Two agency partners, who preferred one account they controlled
- Meta’s own reps, whose standard advice was to consolidate
- About two slower weeks while the structure was built
- A higher acquisition-cost ceiling for genuine new buyers
- Heavier reporting across four accounts
Four accounts, each learning one thing
One shared account, four objectives
Installs, first purchases and retention competed for the same learning. On sale days, the signals tangled.
Four accounts, no audience overlap
Each account learned one objective. Existing high-intent buyers were excluded from acquisition.
Acquisition scales to 4x a normal day
The three acquisition accounts could take four times their daily budget during sales without resetting. Retention stayed steady by design.
A leaner product feed
With ROI Hunter, about 40% of products came out of acquisition: unprofitable items, high-traffic items that never sold, and low-margin brands. Conversion rate rose about 20%.
The structure held where it used to break
- One I Heart Beauty sale delivered 3x sales, 2x profit and 105% more new users than comparable past sales.
- In daily data, the old single-account setup had started to break at about 2x budget. Recovery after volatile periods fell from 2–3 weeks to 2–3 days.
- The same structure carried seven in-house brand launches, each measured separately.
- Recognized with Purplle’s Superstar award (April 2021) and a 2021 MCube Awards nomination.
I’d automate reporting from day one
Four accounts made the monthly CXO reviews heavy to build by hand. That time would have been better spent on analysis.
Separate intent before you optimize for efficiency. Scale doesn’t break good structures; it exposes weak ones.
More work