Meta Andromeda Signals: What to Feed the Algorithm Now That Targeting Is Dead
TL;DR
With targeting precision fading, Meta Andromeda runs on signal quality. Here is exactly what data to feed it: events, value, and identifiers, ranked by impact.
Once a client accepts that manual targeting is not the lever it used to be, the next question is always the same: "so what do we actually control?" The honest answer is signal - what you tell Andromeda about who converts and how valuable they are. I am Naman Khetawat, and here is the exact hierarchy of signal we prioritise when we take over a Meta account in 2026.
The citable answer: Meta Andromeda's retrieval quality depends on the signal you feed it, ranked by impact: server-side purchase events with hashed customer identifiers first, value-based optimisation data second, and engagement/view-content events third - accounts with event match quality above 7/10 consistently retrieve and convert better than identical accounts with weak signal. Here is what that means for setup.
Signal Is the New Targeting
Under the old model, you told Meta who to show ads to. Under Andromeda, you tell Meta what a good outcome looks like, and it finds who to show ads to on its own, at a scale and precision no manual targeting rule could match. That shift means the quality of what you feed it - not your audience-building skill - is now the primary lever available to you.
The Signal Priority Stack
| Priority | Signal type | Why it ranks here |
|---|---|---|
| 1 | Server-side purchase events (Conversions API) with hashed email/phone | Highest-value, most reliable event; identifiers let Meta match to real people despite browser blocking |
| 2 | Purchase value passed with each event | Lets Andromeda optimise toward high-value customers, not just any conversion |
| 3 | Add-to-cart / initiate-checkout events | Useful mid-funnel signal, but should never substitute for purchase data |
| 4 | View-content / engagement events | Weakest signal; useful for audience seeding, not core optimisation |
Event Match Quality Is the Metric to Watch
Meta scores how confidently it can match your server-side events to a real user profile, on a scale visible in Events Manager. We treat anything below 6/10 as an active problem, because it means a meaningful share of your purchase events are not being matched to a person Andromeda can learn from - effectively wasted signal. Getting this score above 7-8 usually requires passing more than just email: phone number, first/last name, and city/zip, all hashed, meaningfully improve match rate over email alone.
This is the single most common gap we find auditing new Meta ads accounts - businesses that set up Conversions API once, years ago, and never revisited whether it is passing complete identifier data.
Value-Based Optimisation, Not Just Conversion Counting
Passing purchase value with every event lets Andromeda optimise toward customers who spend more, not just anyone who converts. Without value data, the system treats a ₹300 order and a ₹15,000 order identically, which means it has no reason to prioritise finding more of the second type. Brands with a wide range of order values see the largest gains from turning this on, because it lets the retrieval engine chase the customers that actually move revenue.
Don't Substitute Weak Signal for Strong Signal
A common mistake is optimising toward add-to-cart or view-content events because purchase volume feels too low to optimise against directly. This can work short-term for very new accounts, but it should be treated as a temporary bridge, not a destination - the moment purchase volume supports it, optimisation should move back to purchase events, because weaker signal always retrieves a lower-quality audience than genuine purchase intent does.
A Real Example
A jewelry D2C brand had event match quality stuck at 5/10 for over a year, despite Conversions API technically being "live." Auditing the implementation showed only email was being passed, unhashed inconsistently across devices. We rebuilt the event payload to include hashed email, phone, and location data, plus purchase value on every event. Match quality rose to 8/10 within two weeks of sufficient volume, and cost per purchase dropped 21% with zero changes to targeting, budget, or creative - purely from Andromeda having better signal to retrieve against.
FAQ
What signal does Meta Andromeda actually use?
Primarily server-side purchase events via Conversions API, with hashed customer identifiers (email, phone, name, location) attached, plus purchase value. Weaker engagement signals like view-content events matter far less and should not be the primary optimisation target once purchase volume supports better data.
What is a good event match quality score?
Aim for 7 or above out of 10. Scores below 6 typically indicate incomplete identifier data being passed with events, which materially limits how well Andromeda can match your ads to real converting customers.
Does passing purchase value actually change delivery?
Yes. Value-based optimisation lets the algorithm prioritise finding customers who resemble your highest-value buyers, not just anyone who converts at any amount. Brands with wide order-value ranges typically see the largest improvement from enabling this.
Fix Your Signal Before Changing Anything Else
If your Meta account's targeting and creative both look fine but performance still lags, the underlying signal quality is the most likely culprit worth checking first. Book a call with Balistro and we will audit your Conversions API setup and event match quality.
