The scenario
This playbook applies to a direct-to-consumer home goods brand selling through its own Shopify store that has built most of its growth on Meta advertising. For years the formula works: product photography, broad targeting, and a return on ad spend that clears the margin threshold. Then it reverses. Cost per order rises each quarter, the founder no longer trusts Ads Manager, and the in-house team is producing creative without knowing which of it works.
A brand in this position typically asks Axoria to take over paid social with a clear priority: profitable new customers, not just orders.
The challenge
Accounts like this usually suffer from three overlapping issues. Browser-side pixel tracking has degraded after platform privacy changes, so the algorithm is learning from an incomplete picture. Campaigns with no customer exclusions spend heavily on people who have already bought and would return through email anyway, which flatters ROAS while hiding the true cost of a new customer. And creative is judged on gut feel rather than any consistent test design. The question is how much of the reported performance is real new-customer acquisition, and what it takes to grow that profitably.
Research and analysis
We start by reconciling Ads Manager against Shopify order data by customer email and order timestamp. This shows what share of attributed purchases come from existing customers and what share is attributed on view-through alone. We then segment historical creative by format, hook and message and match it to first-order outcomes. In this scenario a small number of concepts, usually problem-led video showing the product solving something, account for most genuinely new customers, while product-shot statics, which make up most of the output, mainly reach returning buyers. Post-click, social traffic typically lands on product pages designed for search visitors, with no brand context and no reason to trust the brand on a first visit.
Strategy
Fix the signal first
Implement the Conversions API through server-side tagging alongside the pixel, deduplicate events, and pass hashed customer data so the platform can distinguish new from returning buyers and optimise accordingly.
Then restructure for new customers
Separate prospecting from retention with strict customer-list exclusions, set targets on cost per new customer rather than blended ROAS, and move retention spend to a small, capped campaign that complements rather than replaces email.
Creative is reframed as a testing programme. Every concept enters the account with a hypothesis, a fixed budget and a decision rule, and the production calendar is reorganised around what the tests show.
Execution
Optimization
Once the new-customer signal is reliable, weekly reviews centre on cost per new customer by campaign and the proportion of tested concepts that beat the control. The creative programme iterates on winning hooks rather than winning ads, so a successful problem-led video spawns variants in new formats and openings. We introduce TikTok only once a creative style that works for cold audiences has been established, and scale budget in increments tied to whether cost per new customer holds.
How results are measured
We measure in Shopify, not in Ads Manager. Platform-reported ROAS stays on the dashboard for reference, but every decision uses order data with a new-versus-returning flag, compared against the period before the programme started and adjusted for seasonality where the category demands it.
- Cost per order: total paid social spend divided by Shopify orders attributed to the channel, reported blended and for new customers separately. A programme on track shows blended cost per order stabilising and then falling as spend on existing customers is removed and creative efficiency improves.
- New-customer share: the proportion of paid-social orders from first-time buyers, taken from Shopify customer records. This is usually the figure the founder cares about most, and success means it rising steadily as exclusions and prospecting structure take effect.
- Creative win rate: the share of new concepts in each weekly cohort that beat the control on cost per new customer. A healthy testing loop shows this improving over successive cohorts as the team learns which hooks and formats work with cold audiences.
What we have learned
- Blended ROAS hides the problem. Separating new from returning customers changes every decision in the account. It is the first thing we check in any e-commerce paid social audit.
- Tracking quality is a performance lever. Better event data improves delivery before a single ad changes.
- Test concepts, not executions. Judging creative at the hook level produces more reusable learning than judging individual ads.
- Paid and owned channels need one plan. Coordinating with email removes wasted spend and is a core part of our e-commerce marketing approach.