Lookalike refresh schedule

Quick Answer

For most accounts, refresh your Facebook lookalike seeds every 60 days, and rebuild the lookalike itself every 90 days or whenever CPA drifts more than 25% off baseline. High-velocity ecom accounts (spending £1K+/day) may need to refresh every 30 to 45 days. Low-velocity accounts can stretch to 120 days. The right schedule is the one that keeps CPA within target — data, not calendar, should trigger refreshes.

Why you need a refresh schedule at all

Lookalikes drift for four reasons:

1. Seed ageing. Your source audience gets older every day. Customers from 12 months ago may not match today's ICP.
2. Audience saturation. At high spend, you burn through the lookalike pool. Frequency climbs, new conversions slow.
3. Creative fatigue. Running the same ads too long exhausts the audience even when the lookalike is still fresh.
4. Platform drift. Meta updates its signals and models. An audience that was optimal last quarter may be suboptimal now.

A refresh schedule is how you stay ahead of all four without waiting until performance collapses.

The baseline schedule

Every 30 days: Audit account-level metrics. Look at frequency, CPA trend by audience, overlap between audiences.

Every 60 days: Refresh the seed source. Pull a fresh customer list, rebuild the custom audience with the latest data, save with a new version number.

Every 90 days: Rebuild the lookalike from the refreshed seed. Launch it in a new ad set. Run for 7 days in parallel with the old one. Pause the loser.

Every 14 to 21 days: Refresh creative. This is separate from audience refresh but the two need to happen together — creative fatigue often masquerades as audience fatigue.

Step-by-step refresh routine

1. Set a calendar reminder. Every 60 days: "Refresh lookalike seeds." Every 90 days: "Rebuild lookalikes." Every 30 days: "Audit frequency and overlap."
2. Export fresh data. Pull your latest customer list, filtered to the last 90 days or 180 days depending on your refresh window.
3. Rebuild the custom audience. Upload the new file. Do not edit the old audience — save as a new version (e.g., Customers_180D_v2).
4. Build the new lookalike. Audiences > Lookalike > pick the new custom audience > same country, same size as before > save as LAL_1%_Customers_v2.
5. Create a parallel ad set. Clone your existing ad set and swap the audience to the new lookalike. Keep creative, budget, and placements identical.
6. Run for 7 days. Monitor CPA, CTR, and frequency.
7. Decision time. If the new version beats the old, pause the old. If not, keep the old running and investigate why the new version underperformed (seed issue? creative fatigue?).

Real examples

DTC apparel brand, £2,000/day spend. Established a 60-day refresh cadence. Before the cadence, CPA would drift from £18 to £26 over two months. After the cadence, CPA stayed within £18 to £21. The brand saved roughly 15% on ad spend simply by refreshing on schedule.

B2B SaaS, slower pipeline. 90-day cadence was too aggressive — sales feedback loop was long and the team did not have enough new data to meaningfully refresh the seed every 3 months. Switched to 120 days. CPL remained stable.

Local services business. Refresh every 45 days because the customer list was small and every new booking materially changed the seed. The faster cadence kept the lookalike tightly aligned with recent behaviour.

Signals it is time to refresh (regardless of calendar)

• CPA has drifted 25%+ off baseline in the last 14 days
• Frequency has crossed 3.5 and continues climbing
• CTR has dropped 30%+ vs the first week of the ad set
• Conversion rate from click to purchase has dropped even though CTR is steady
• You just shipped a major product change (new category, new price point) that changes your ICP
• You launched in a new country and the old lookalike is diluted

Signals to hold off on refreshing

• CPA is still inside target — do not fix what is not broken
• The ad set is still in learning phase — refresh interrupts learning
• You just refreshed — give the new audience at least 14 days
• Creative was just rotated — changes from creative refresh can mask audience issues

FAQs

Does refreshing the seed destroy my learning data?
Partially. When you launch a new ad set with the new lookalike, it will need to go through learning phase (50 conversions). Keep the old ad set running in parallel until the new one has proven itself.

Can I skip refreshing if I use Advantage+ Audience?
Refreshing is less critical with Advantage+ because Meta continuously expands beyond your seed. But the seed itself still ages — refresh every 90 to 120 days even with Advantage+.

Should I refresh all lookalikes at once or stagger them?
Stagger. Refreshing everything simultaneously puts multiple ad sets into learning phase at once, which is disruptive. Spread refreshes across a few weeks.

What if my refreshed lookalike performs worse?
Investigate: is the new seed narrower or broader? Has creative fatigued? Is there overlap with another new audience? Do not immediately blame the refresh — check the other variables.

How do I know if my refresh cadence is right?
Monitor CPA trend. If it stays stable, your cadence is working. If it drifts upward despite refreshing, refresh more often. If refreshing never improves anything, refresh less often.

Refresh creative at the same pace as audiences

A disciplined audience refresh schedule is only half the system. The other half is creative refresh — without it, your fresh audiences will still see tired ads. Pix-Vu tests new creative variants against your audiences so every refresh cycle ships winners, not experiments. Try Pix-Vu free at https://pix-vu.com.