EL.KHAWAGA-UAE

Case Studies / 01

Retail · United Arab Emirates

Atlas Retail —
247% ROAS in 90 days.

We took a multi-store retail chain off a generic paid playbook and rebuilt their entire performance stack — feeds, creative, audience architecture, and measurement. The ROAS curve never went back.

  • 247%

    ROAS lift

    Vs. 90-day rolling baseline

  • −38%

    CAC drop

    Across paid channels

  • +12%

    AOV

    Driven by bundling logic

  • 90 days

    Timeline

    Strategy → results

01

The brief.

Atlas Retail came in mid-quarter with a flat top-line, rising CAC, and four agencies layered on top of each other across paid social, paid search, programmatic, and influencer. Each agency reported double-digit ROAS in isolation. The P&L was telling a very different story.

The brief was simple on paper and brutal in practice: replace the entire performance stack inside one quarter without dropping revenue. No experimentation budget. No grace period. No excuses if the curve dipped.

02

State of play.

We ran a 10-day audit. Three things were broken at once:

  • The product feed had three SKU duplicates per item and a feed taxonomy that bore no relation to how customers searched.
  • Creative was being produced as a service line by the social agency, then handed to the paid agency cold. There was no performance loop on the creative pipeline.
  • Every agency was measuring with last-click attribution against their own platform. Stacked together, they were claiming 312% of the actual revenue.

03

Strategy.

We collapsed the four agencies into one operating model under Elkhawaga, with three principles:

  • One feed, one taxonomy. Rebuilt from scratch against actual search-term data, not legacy product catalog schemas.
  • Creative as a paid asset, not a deliverable. Concepts written against funnel position, then split-tested weekly against a control. Cuts that lost two weeks were killed.
  • One source of truth. Server-side conversion tracking, GA4 as the system of record, weekly incrementality checks on the largest channels.

04

Execution.

Weeks 1–3: rebuilt the feed, deduped SKUs, restructured the Google and Meta campaign architecture around four buyer segments instead of seventeen.

Weeks 4–7: shipped 31 creative variants tied to funnel stage, rotated weekly against a thompson-sampling style allocation. The top 6 by week 7 were doing 80% of the spend.

Weeks 8–12: turned on incrementality testing on Meta and YouTube, discovered programmatic display was credit-laundering, killed $42k/month of it without revenue impact, redeployed into refreshed creative and search expansion.

05

Outcomes.

At day 90, blended ROAS was 4.8x against a 90-day rolling baseline of 1.4x. Spend was held flat. Revenue tracked +89% YoY for the quarter. CAC dropped 38%. Average order value rose 12% from tightening the bundling logic on the highest-margin SKUs.

Twelve months on, the framework is still running. The team that built it now staffs Atlas' in-house performance group — we rolled off paid retainer at the end of year one and handed over a working machine.

Client

They came in, replaced four agencies, didn't flinch, and gave us back the performance number we'd stopped believing in.

— Head of Growth, Atlas Retail (engagement 2025–present)

Next case

Northstar Hospitality —
+58% direct bookings.

See all case studies