For local lead-generation advertisers on Meta
Rehearse before you spend.
Advert Lumen shows which of your ad variants deserve a live test, how big that test has to be, and what your own account's history already says about what works. It never pretends to know more than the evidence allows.
How the saving happens, in order of certainty
- 1 · Ships first
Fewer variants reach the live test
A structured review of each creative, and a comparison grounded in your account's own history, remove the variants that history already argues against. Where the evidence cannot separate two variants, Advert Lumen says they are too close to distinguish.
- 2 · Ships first
The test you do run is sized correctly
The test planner turns your own CPM and link CTR into a budget per variant, a duration and a stop rule. When the budget cannot answer the question, it says what difference it could detect instead.
- 3 · After validation
Forecasts, only where they have earned it
Numeric forecasts appear only for objectives where a leakage-free evaluation showed they beat a plain historical baseline, labelled as model estimates with their track record beside them.
We will publish the savings we observe with pilots, with their sample sizes, once we have measured them. Until then we make no claim about how much you will save.
What you get
Results keep three things apart: what was measured, what was estimated, and what was suggested.
Account context
Observed: Measured in this account's own history.Link CTR, CPC, CPM and cost per lead by placement, age and gender, each with its impression count.
What has worked for you
Observed: Measured in this account's own history.Creative features against your results, with counts and intervals; your waste share and learning spend.
Creative review
Generated: Written by a language model. No real people were surveyed.Hook, clarity, offer, proof, call to action and placement fit, each with a suggested change you can apply and re-run.
Scorecard
Observed: Measured in this account's own history.Every eligible prediction is compared with what actually happened, kept whether it went well or not.
Three ways in. Pick the one you are comfortable with.
Connect Meta
One read-only permission, ads_read: “let this app view your ad reports”. Advert Lumen cannot change an ad.
Import an export
Upload an Ads Manager CSV or XLSX export and map its columns. No access granted.
Describe a campaign
Four short steps for people who will do neither. Results use industry benchmarks, clearly labelled.
Honest about accuracy
Synthetic-audience vendors advertise accuracy figures they grade themselves. We publish our own record of predictions against actual results instead: coverage with its interval, the interval score next to a plain baseline, and the sample size. It starts empty and fills as campaigns finish.