Meta just quietly 4x’d the maximum retention window on one of the most important audiences in your account — the Purchase event audience. The window went from 180 days to 730 days, and unless you actively opt out before May 18, 2026, every existing 180-day Purchase audience in your account auto-upgrades to two years (Jon Loomer Digital, May 7, 2026). That’s a bigger change than the headlines made it look. If you run DTC ads, the question isn’t whether to notice — it’s whether the new default actually serves your account.
TL;DR: On May 18, 2026, Meta raises the Purchase custom audience retention cap from 180 to 730 days. Existing 180-day audiences auto-update unless you opt out (Loomer, 2026). Extend if you sell supplements, beauty, CPG, or other repeat-purchase DTC. Opt out if you sell mattresses, furniture, single-purchase items, or you’re under 90 days old. Either way, your lookalike strategy needs a refresh.
What actually changed (and what didn’t)
Per Meta’s own notice, surfaced by Jon Loomer on May 7, 2026: “Starting May 18, 2026, you’ll be able to set audience retention to a maximum of 730 days when you select purchase events in your website and app activity custom audiences. Existing custom audiences with purchase events and 180 days audience retention will automatically update to 730 days. If you would prefer to keep your current audience retention settings, you can opt out before May 18, 2026” (Jon Loomer Digital, 2026).
The clean summary, for anyone skimming:
- Old cap: 180 days of retention for Purchase event-based custom audiences
- New cap: 730 days (two years) — a 4x expansion
- Effective date: May 18, 2026
- Default behavior: existing 180-day Purchase audiences auto-upgrade to 730 days
- Opt-out window: before May 18 — and per the rollout, the choice persists, so you don’t have to keep re-toggling it on new campaigns
- What didn’t change: non-purchase events (page views, add-to-cart, video views, lead form opens) still cap at 180 days for now
This is purely a retention-window change. The pixel still fires the same way, the event definitions are unchanged, and CAPI plumbing is untouched. What changes is how far back Meta will look to populate the audience — and that one knob ripples through almost every part of a structured DTC ad account.
Should you actually extend? The vertical-by-vertical decision matrix
Here’s the question Loomer didn’t quite answer: just because Meta will now let you keep a two-year Purchase audience, should you? The honest answer is “it depends on your repeat-purchase curve” — and most practitioner takes I’ve seen so far are skipping that nuance.
The way I think about it in the accounts I run: if your category’s repeat-purchase rate at 12-24 months is meaningful, a longer window gives Meta’s algorithm a richer signal for lookalikes, ASC seeding, and value optimization. If your category is one-and-done, you’re just feeding the algorithm stale signal. Meta’s Andromeda ML system already leans heavily on creative and event signal density — adding two years of low-relevance buyers doesn’t help it; it dilutes it.
A few non-obvious calls from this matrix worth flagging:
- Apparel is a “depends” — and the matrix puts it on the extend side. If you sell seasonally (swim, outerwear, holiday), a buyer from 14 months ago is genuinely valuable as a remarketing or seed audience. If you sell trend-driven SKUs where the actual product has moved on, the audience is noisier than useful. Default to extending only if your AOV-weighted repeat rate is real.
- “Brands <90 days old” matters more than people realize. A new pixel’s 180-day Purchase audience is already statistically thin and skewed by early adopters who may not represent your scaled buyer. Doubling that window to 730 days locks in that skew for two years before the algorithm gets to re-calibrate. Stay at 180 until you have a meaningful 6-month buyer base, then extend.
- Heavy promotional skew is a real risk. If your last 18 months included BFCM peaks that disproportionately weighted discount buyers, extending the window pulls all of that into your seed signal. The wider window only helps if the buyer behavior in it is representative of who you actually want more of.
This is the part Loomer’s announcement post didn’t unpack. The rollout headline says “more retention is better.” For most repeat-purchase DTC, that’s directionally right. For at least a third of the brands I see, it’s not.
What happens to your lookalike audiences when seed sizes 4x overnight?
This is the second-order effect almost no one is talking about. Most DTC accounts build Purchase-based lookalikes on top of their Purchase custom audience. When the seed source goes from 180-day Purchase to 730-day Purchase, the seed audience can grow 2-4x depending on your purchase cadence — and that materially changes what the lookalike actually looks like.
Here’s the tradeoff. A bigger seed isn’t automatically a better seed. Lookalike quality is a function of seed homogeneity. If your 180-day buyer cluster was tight — repeat customers, similar AOV, similar acquisition channel mix — and your 730-day cluster is broader and noisier, the 1% lookalike off the new seed is genuinely different from the 1% lookalike off the old seed. Sometimes that’s good (more scale, better fill rates). Sometimes it’s bad (lower-quality matches, lower ROAS at the same audience size).
The practical playbook I’m using in the accounts I run:
- Don’t rebuild every lookalike on May 18. Let the existing LLAs continue to run for 14-21 days, then read the impact in the data — ROAS, CPA, AOV, frequency. If your post-extension LLAs are performing within 5-10% of pre-extension, the bigger seed is helping. If they degrade meaningfully, the seed needs to be tightened.
- Build a “controlled-seed” Purchase audience. Create a manual Purchase audience locked at 180 days (yes, you can still do this — the cap rose, but you can set retention to whatever value you want below it). Use it as a controlled seed for one set of lookalikes while letting the 730-day seed feed another. A/B them by ad set.
- Watch for seed-size LLA degradation curves. If a 730-day seed pushes you above the audience size where Meta’s LLA model materially degrades (typically when seed quality variance gets too high relative to size), you’ll see it in lower frequency on the LLA ad sets and rising CPM as Meta delivers to less concentrated pockets.
For the broader argument that lookalike audiences are dead in their classic form, this update doesn’t change the trajectory — but it does change how you should think about the seeds you do still build. Bigger seed, tighter monitoring.
How does this change your Advantage+ Shopping audience architecture?
Advantage+ Shopping campaigns are where this update has the most underrated impact. ASC uses your Purchase audience as a critical signal layer — both for the “existing customers” budget cap (the slider that controls how much of your ASC spend reaches existing buyers) and for the broader optimization signal that feeds Meta’s bidding model.
When that Purchase audience definition extends to 730 days, two things shift simultaneously. First, your “existing customer” pool inside ASC quadruples in scope. The 25% existing-customer budget cap that was carving out roughly 25% of spend toward your 180-day buyer pool is now carving out 25% toward a much larger and older buyer pool. Second, the value-rules layer Meta rolled out a couple of weeks ago interacts with this directly.
Value rules for audiences let you bid more aggressively for high-value audience segments. If one of your value rules targets “past purchasers,” the rule’s footprint just grew 4x. That can be exactly what you want if a 24-month-old buyer is genuinely worth more than a cold prospect — or it can blow up your bid efficiency if the rule is now bidding up against a much more diluted pool. Audit every value rule that references the Purchase audience and make sure the bidding logic still maps to the new definition.
The architecture move I’m making: in the Advantage+ Shopping campaigns I run, I’m explicitly setting the “existing customer” definition to a custom 90-180 day Purchase audience rather than relying on the default. That keeps the existing-customer carveout tight and lets me use the bigger 730-day audience as an exclusion or value-rule input separately. Same logic for accounts running broad targeting with Advantage+ audience layering — the layering decisions assume a specific buyer definition. Don’t let Meta silently redefine that assumption.
Existing-customer exclusions just got materially more complex
If you’ve built your prospecting test architecture around excluding “past purchasers” — and most structured DTC accounts have — extend that mental model. A 180-day exclusion was meaningful but narrow. Buyers from 7 months ago could still see your prospecting ads. A 730-day exclusion is wildly different: it now removes anyone who has bought from you in the last two years from your prospecting reach.
For high-frequency repeat brands, that’s actually a problem. Your “prospect” pool shrinks because Meta is now suppressing two years of buyers from prospecting delivery. For mature DTC brands, this can hide your scale ceiling — you’ll see prospecting CPMs rise and incremental reach plateau, and the cause won’t be Meta delivery; it’ll be that your exclusion just doubled.
The fix is one of two patterns:
- Tiered exclusion architecture. Instead of one “all past purchasers” exclusion, build three: 0-90 days (suppress hard from prospecting), 91-365 days (suppress from prospecting but eligible for win-back), 366-730 days (eligible for both prospecting and reactivation). This lets you treat lapsed buyers as different from active buyers — which they statistically are.
- Time-decayed value approach. Use the 730-day audience as a value-rule audience rather than an exclusion. A buyer from 18 months ago can still see prospecting-style creative, just bid down to reflect their lower marginal value. This works better in accounts where buyer reactivation contributes meaningfully to net revenue.
If you’ve built your account on the structured DTC test architecture I’ve written about — distinct prospecting, retargeting, and reactivation campaign structures — the 730-day default messes with prospecting most. Audit those exclusions first.
How do you actually opt out before May 18?
The tactical piece is short. Inside Ads Manager, go to Audiences, filter to website and app activity custom audiences with purchase events, and for each one you don’t want auto-extended, edit the audience and set retention back down to 180 (or whatever value you actually want). Meta has confirmed that the opt-out preference persists — once you set it, it sticks on new campaigns and rebuilds without you needing to re-toggle every time.
Two notes from running through this in real accounts:
- Don’t bulk-opt-out reflexively. The lazy move is to opt out of everything because “the default changed.” The smarter move is to evaluate each audience against the decision matrix above and only opt out of the ones where the decision actually matters.
- Document what you change. Especially in agency accounts or multi-stakeholder setups — if you opt out of some audiences and not others, the next person looking at the account in three months needs to understand why. A short Notion doc or audience-naming convention (e.g., suffix audiences with “_180d” when manually capped) saves a lot of confusion.
Frequently Asked Questions
Does the 730-day Purchase audience window apply automatically to my existing audiences?
Yes. Per Meta’s notice via Jon Loomer (May 7, 2026), existing custom audiences with purchase events and 180-day retention will automatically update to 730 days on May 18, 2026, unless you actively opt out beforehand. The change is the new default — keeping your existing 180-day setting requires explicit action before the deadline.
What’s the risk of letting a brand-new DTC account auto-extend to 730 days?
The biggest risk is sample bias. A pixel under 90 days old has a thin, statistically skewed Purchase audience dominated by early adopters and friends-and-family buyers (Meta Engineering Andromeda 2024 notes that signal density matters more than volume for ML targeting). Extending that window doesn’t fix the bias — it locks it in for two years. Stay at 180 until you have a representative 6-month buyer base.
Should I rebuild all my Purchase-based lookalike audiences on May 18?
No — at least not immediately. Let existing lookalikes run for 14-21 days after the auto-extension, then read ROAS, CPA, and AOV impact. If post-extension performance is within 5-10% of pre-extension, the bigger seed is helping. If it degrades meaningfully, build a controlled 180-day seed and A/B test it against the 730-day seed by ad set. Don’t rebuild reflexively without data.
Does this affect non-purchase events like page views or add-to-cart audiences?
No. The cap expansion applies specifically to website and app activity custom audiences built on purchase events. Page views, add-to-cart, lead form opens, video view audiences, and other non-purchase event audiences still cap at 180 days under current Meta documentation. If Meta expands the cap to additional event types later, expect a similar opt-out window.
For a deeper dive, see my guide on meta advantage+ shopping campaign structure: account architecture and budget best practices for dtc in 2026.
The Bottom Line
Meta’s 730-day Purchase audience window isn’t a tactic — it’s a default change that quietly redefines what “past purchaser” means inside your account. For repeat-purchase DTC (supplements, beauty, CPG, pet, apparel), the bigger window is mostly a gift: better lookalike seeds, richer ASC signal, more granular value-rule audiences. For single-purchase categories, brand-new pixels, and privacy-sensitive verticals, it’s a quiet downgrade if you let it auto-apply. Make the decision actively before May 18, 2026 — and then audit your lookalike refresh strategy, ASC existing-customer caps, and exclusion architecture so the new default doesn’t silently change behavior you’ve spent months tuning.
Sources
- Jon Loomer Digital. “ChatGPT Ads Go Self-Serve, Purchase Retention Expands, and More.” May 7, 2026. https://www.jonloomer.com/chatgpt-ads-self-serve/
- Jon Loomer Digital. “Meta Value Rules for Audiences Explained.” 2026. https://www.jonloomer.com/meta-value-rules-audiences/
- Jon Loomer Digital. “A Guide to Meta Ads Targeting in 2026.” 2026. https://www.jonloomer.com/meta-ads-targeting-2026/
- Meta Business Help Center. “How Long Do People Stay in a Custom Audience From My Website?” https://www.facebook.com/business/help/419552341510847
- Meta Engineering. “Meta Andromeda: Supercharging Advantage+ automation with the next-gen personalized ads retrieval engine.” December 2, 2024. https://engineering.fb.com/2024/12/02/production-engineering/meta-andromeda-advantage-automation-next-gen-personalized-ads-retrieval-engine/