Skip to content

Lookalike Audience Build and Test on X Ads

DD
New1 order in queue
1

About this service

Lookalike audiences sound magical in theory and disappointing in practice. You upload a seed list of your best customers, click the "create lookalike" button, and X hands you back an audience that - in the dashboard - looks perfect. Then you run ads against it for two weeks, the CPA comes in worse than your broad targeting, and you conclude that lookalikes just do not work on X. They do work. What does not work is the default workflow: a single seed list, a single 1% audience, no test structure, and no validation against which slice of the lookalike is actually converting.

This engagement builds and tests lookalike audiences on X Ads the way they should have been built the first time. You bring the active X Ads account, the seed list (or the data to build one), and a small test budget. I build the seed architecture, generate multiple lookalike slices, configure a structured test, and deliver a written audience report telling you which slice to scale and which to retire.

This is a one-time setup engagement. It is not ongoing management. You keep the account, the seed lists, the lookalike audiences, and the report - everything stays in your account, owned by you.

1. What you actually get

The build covers the complete lookalike lifecycle: seed → slices → test → verdict.

  • Seed-list architecture. Most advertisers use one seed list and call it done. We build 3-5 seed lists, each scored for quality, size, and projected reach. Typical seeds: your top-decile customers by LTV, recent purchasers, engaged email subscribers, and pixel-fired high-intent visitors. Each seed has a written rationale.
  • Follower-Look-alikes alongside pixel-based lookalikes. X offers both. We configure both where the data supports it, with separate naming conventions so the test results stay clean.
  • Multiple lookalike slices, not just the 1%. Typically 3 slices per seed - a tight slice (top 1-2%), a medium slice (1-5%), and a broad slice (1-10%). The slice size is the single biggest lever on lookalike performance, and most advertisers never test it.
  • A structured A/B test across slices. Each slice gets its own ad group with the same creative, the same bid strategy, and the same budget. Test runs for 5-7 days at a spend level you set (typically $300-$1,500 total test budget, paid by you directly to X). The structure isolates the slice variable so the verdict is clean.
  • Pixel and conversion API verification before the test starts. If the measurement layer is broken, the test results are broken. We verify the events that matter against your backend before the test launches - same approach as a measurement setup engagement, scoped down to the test events.
  • A written audience report at the end. Slice-by-slice performance, CPA by slice, reach by slice, frequency by slice, and a clear scale-or-retire recommendation per slice. The report is the deliverable you keep.

2. How the four days run

  1. Day 1: Intake and seed build. You grant manager-level access to the X Ads account and share your seed source - a customer CSV, an email list, a pixel audience, or a follower list. I review the seed, score it for quality and projected reach, and confirm the 3-5 seed lists in writing. If your seed is too small (under 500 records for pixel-based, under 500 followers for Follower-Look-alikes), we will tell you before the test launches.
  2. Day 2: Slice generation and test structure. The 3-5 seeds generate 9-15 lookalike slices (3 slices per seed). Each slice becomes its own ad group with shared creative, shared bid, shared budget. Pixel and conversion API are verified against your backend for the test events.
  3. Day 3: Test launch. The structured A/B test goes live with your spend. I monitor for the first 24 hours to confirm pacing, exit the learning phase, and flag any obvious creative issues.
  4. Day 4: First-read report. A first-read report is delivered at the 24-48 hour mark - early signal on which slices are outperforming, which are underperforming, and any pacing corrections. The full slice-by-slice verdict is delivered at day 7 (a follow-up message - the engagement fee covers the four-day setup plus the day-7 verdict message).

3. Who this is for - and who it is not

A strong fit if:

  • You are already running X Ads with a working prospecting layer and want to add a tested lookalike strategy on top.
  • You have a real seed source: at least 500 customers, 500 pixel-fired high-intent visitors, or 500 engaged email subscribers. Smaller seeds produce lookalikes that are statistically indistinguishable from broad targeting.
  • You have a test budget of $300-$1,500 to run the structured A/B test (paid by you directly to X).
  • You have a working conversion event on your site (purchase, lead, sign-up) so the test has a clear winner metric.

A poor fit if:

  • You have not started running X Ads yet. Lookalikes are a refinement on top of prospecting, not a substitute for it.
  • Your seed source is under 500 records. We will tell you this before you order - small seeds produce noisy lookalikes and the test verdict will not be trustworthy.
  • You do not have a working conversion event. Without a clear winner metric, the test cannot produce a scale-or-retire verdict.
  • You want ongoing management. This is a one-time setup; ongoing management is a separate engagement.

4. You bring / I run - the division of responsibility

This is a management-only engagement, and the split is deliberate.

You bring:

  • The active X Ads account, with billing you control.
  • The test budget ($300-$1,500 typically), paid by you directly to X - I never bill or fund media.
  • The seed source: a customer CSV, an email list, a pixel audience, or a follower list.
  • A working conversion event on your site, with a backend source we can use to validate the test events.
  • A 24-hour response habit on intake questions - important for the four-day delivery window.

I run:

  • Seed-list architecture, slice generation, and the structured A/B test design.
  • Pixel and conversion API verification scoped to the test events.
  • Test launch, 24-hour monitoring, and the day-7 verdict.
  • The written audience report and the scale-or-retire recommendations.

Ad spend stays on your card with X. Nothing routes through me.

5. Reporting and handover

Two deliverables. The first-read report at the 24-48 hour mark: early signal on which slices are outperforming, which need pacing corrections, and any creative issues. The full audience report at day 7: slice-by-slice performance, CPA by slice, reach by slice, frequency by slice, and a clear scale-or-retire recommendation per slice. The report is structured so a future vendor (or your in-house team) can pick it up and execute without re-doing the analysis. A 20-minute handover call walks through the report.

6. Why this engagement

Three differences worth naming plainly. First, the engagement tests multiple slices per seed - typically 3 slices (tight, medium, broad) - instead of generating one 1% lookalike and hoping for the best. Slice size is the biggest lever on lookalike performance, and most advertisers never test it. Second, the seed architecture uses 3-5 seeds, not one. Different seeds produce different lookalikes, and the only way to know which seed produces the best lookalike is to test them against each other. Third, the test structure isolates the slice variable - same creative, same bid, same budget across slices - so the verdict is clean. Most lookalike tests fail because they change three variables at once and the advertiser cannot tell which one moved the CPA.

What is included

  • Base X Ads campaign · Your ad account, your spend
  • Revisions · 1 round included
  • Delivery · 4 to 6 days

Frequently asked questions

Industries served

  • Food and Drink
DD
@ddangVietnamMember since Sep 2026
Online

Bing and Pinterest management, no lock-in

Completed orders
0
Rating
No reviews yet
Active listings
7
Ad platforms
7

More from Duc Dang

View storefront