Methodology
How Bach actually audits your account.
No black box, no borrowed benchmark, no surveillance. Bach reads your own account history, sizes every leak in the money it costs you, and changes nothing until you approve. This is exactly how it reasons — and a sample audit you can read end to end.
In one paragraph
Bach.ai audits a Meta ad account by reading its own 12-month history across structure, audience, creative, budget and tracking, scoring each signal against that account’s baselines rather than a generic benchmark. It sizes every issue in recoverable spend — wasted budget on fatigued, overlapping or over-target delivery — ranks them by money, grounds the judgement in MER and margin rather than platform-reported ROAS, respects the learning phase, and proposes each fix for your approval before anything changes.
What Bach reads — and how it reasons.
A good audit is not a list of metrics. It is a judgement about what is actually wrong, grounded in how Meta’s delivery and D2C economics really work. Four principles govern every read.
Your own account, not a generic benchmark
Bach scores every campaign, ad set and creative against your account’s own 12-month history — your baselines, your seasonality, your audiences. A 2.4× ROAS is a win for one brand and a leak for another; the benchmark that matters is your own trend, not an industry average.
Margin and MER, not vanity ROAS
Platform-reported ROAS can over-credit and flatter — counting the same conversions across reporting layers. Bach reads spend against blended performance (MER) and your margin, so a "7× ROAS" campaign that is really cannibalising organic, or selling at a loss after COGS, gets flagged — not celebrated.
Respects the learning phase
A new ad set in learning, or one that just had its budget moved, is not a loser — it is a campaign without enough data yet. Bach knows the difference between under-performance and under-information, and waits for signal before it calls anything a leak.
Health before blame
Before Bach blames a creative or an audience, it checks the plumbing: pixel and Conversions API health, attribution windows, billing and delivery gaps. A drop is often a broken signal or a delivery outage, not a bad ad — and a tool that misreads that erodes your trust fast.
How a leak is sized
Every leak is sized in the money it costs you.
Not a severity colour, not a 1–10 score. Bach estimates the recoverable spend behind each issue — the budget currently going to delivery that is fatigued, overlapping, or paying above your target — and surfaces it. The biggest leak is simply the one costing you the most.
- Wasted spend = budget on segments past their efficient frequency, overlapping audiences bidding against each other, or delivery above your target CPA.
- Sized against your margin and blended return, so the number is what Bach estimates is recoverable — not a platform-reported figure.
- Ranked largest-first, with the expected before → after attached to the fix.
Synthetic sample account · illustrative figures, localised to your region. Your numbers come only from your own connected account once you run an audit.
100+ checks, every day — surfaced as a ranked list, not a dashboard.
Bach runs more than a hundred automated checks across campaign structure, audiences, creative, budget & bidding, tracking & signal, and profitability. You never have to read them. They collapse into one ordered list: the leak costing you the most, then the next, each with a one-tap fix and its expected impact.
The hard stop
It proposes. You approve. Then it executes.
The audit is read-only. Every fix is a proposal with its expected impact attached — nothing touches your Meta account until you tap approve, and every change Bach makes is logged and reversible in one click. The methodology only matters if you can trust the hand on the controls.
The standards behind every audit.
Proof, not promises. Each of these is true the moment you connect.
Read-only by default
Bach connects to look first. Nothing on your account changes until you approve a specific proposal.
Logged & reversible
Every action is timestamped, explained in plain language, and can be rolled back in one click.
Never trains models
Your account data runs your audit and nothing else. It is never used to train models and never resold.
Encrypted & least-privilege
Tokens are encrypted at rest and used on a least-privilege basis. Disconnect any time, in one click.
Benchmark sources & how we use them
Where the numbers come from — and how honest they are.
Our claims are the foundation of visitor trust before any review exists. Every recurring figure on this site is registered here with its provenance and an honest label. Ranges labelled illustrative are planning aids, not promises. The only real numbers are the ones that come from your own connected account once an audit runs.
Documented in official platform guidance. Behaviour is deterministic and reproducible.
A fact about Bach's own audit engine, verifiable from the product itself.
Reported in independent or platform materials. Direction is credible; your result will vary.
A representative range drawn from operating experience and public benchmarks. Used to frame the order of magnitude, not to promise an outcome.
A working rule we apply internally, derived from operating accounts at scale. Varies by spend level, audience size, and creative type.
| Figure the site uses | Honesty status | Source basis |
|---|---|---|
| Learning phase exits at ~50 optimisation events / 7 days per ad set | Verified platform mechanic | Meta Business Help Center — About the learning phase. Platform-documented; Bach uses this to distinguish under-performance from under-information. |
| ~19% more reported conversions with full Conversions API implementation | Directional benchmark | Independent CAPI studies and Meta advertiser materials cite a 10–40% range in reported conversions. We use ~19% as a mid-range directional figure. This is reported uplift, not incremental or guaranteed revenue lift. |
| 20–40% of Meta spend commonly wasted | Illustrative planning range | Our operating experience across accounts, supplemented by directional industry benchmarks. The range frames the order of magnitude of typical inefficiency — your actual recoverable spend comes only from your own audit. |
| ~2.8× median e-commerce Meta ROAS; strong brands 4–5× | Illustrative planning range | Aggregated public benchmark ranges. Varies widely by vertical, margin, geography, and attribution window. Bach deliberately does not use industry ROAS as a target — it uses your own account trend instead. |
| ~14-day winning-creative fatigue window | Operating-experience heuristic | A working threshold applied in Bach's creative-fatigue checks, derived from our operating experience. The actual window varies materially by daily spend, audience size, and creative type. |
| 30–45% typical prospecting / retargeting audience overlap | Operating-experience heuristic | Our operating experience. Used to set the overlap-threshold flag inside Bach's audience checks. Your account's actual overlap is calculated directly from your Meta data when you run an audit. |
| 100+ automated checks per audit | Verified product fact | Bach's own audit engine. The count reflects distinct checks across campaign structure, audiences, creative, budget & bidding, tracking & signal, and profitability — verifiable from the full checklist. |
This register is updated when figures change on the site. Figures labelled illustrative are labelled that way wherever they appear. The only numbers that are truly yours come from your own connected account.