Methodology

How Adsevon detects advertising problems

Adsevon’s analysis engine applies deterministic detectors to your campaign data. Each detector looks for one pattern, states the evidence it found, and is scored for confidence and priority. This page explains each concept in plain terms — what it detects, how to read it, and what it does not mean.

Relative, not universal

Detectors compare a campaign against its own history and account context. A “good” CPA or CTR in one business is a failure in another, so thresholds are never presented as industry truth.

Evidence before advice

Every finding ships with the numbers behind it. If the evidence is thin — too little spend, too few conversions — confidence says so instead of pretending certainty.

Data quality first

Broken tracking imitates almost every problem these detectors find. Rule out measurement issues before acting on any single signal.

Wasted spend

What it detects

Campaigns or ad sets that consume meaningful budget while producing no (or near-no) conversions over a window long enough for conversions to plausibly appear.

How to read it

Read it as "this spend has had its chance." Check the conversion definition and tracking before acting — a tracking break looks exactly like wasted spend.

What it does not mean

It does not mean the campaign can never work — prospecting and awareness activity may be doing a job conversions alone do not capture.

High CPA

What it detects

A cost per acquisition materially above the level the account itself sustains, or above a target the advertiser set, sustained over enough conversions to be more than noise.

How to read it

Compare against your break-even economics: the CPA a business can afford is set by margin and order value, not by what other advertisers pay.

What it does not mean

A high CPA is not automatically a bad campaign — high-value conversions legitimately cost more. Value per conversion decides.

Low CTR

What it detects

A click-through rate materially below the account’s own baseline for comparable placements, or falling steadily — the usual signature of weak ad–audience fit or creative fatigue.

How to read it

Judge CTR relative to the same campaign’s history and siblings, not against universal benchmarks. Placement and audience temperature change CTR by multiples.

What it does not mean

Low CTR alone is not a verdict. A narrow, high-intent audience can convert profitably at a modest CTR.

Budget concentration

What it detects

A single campaign (or a small few) absorbing a dominant share of total spend, so account performance hinges on one point of failure.

How to read it

Concentration is a risk signal, not an inefficiency: the question is whether the dominant campaign earns its share and what happens if it fatigues.

What it does not mean

It does not mean "spread budget evenly." Concentration behind a proven winner is rational; unnoticed concentration is the risk.

Scale candidates

What it detects

Campaigns with strong, stable returns and evidence of headroom — consistent ROAS above break-even, healthy conversion volume, and no deterioration trend — where additional budget is likeliest to keep performing.

How to read it

Scale gradually and watch marginal ROAS: the return on the next increment, not the campaign average, tells you when to stop.

What it does not mean

A scale signal is not a guarantee. Past efficiency at one budget does not promise the same efficiency at three times the budget.

Performance deterioration

What it detects

A sustained negative trend in the metrics that matter (CPA, ROAS, conversion rate) versus the campaign’s own recent baseline — distinguished from ordinary daily noise by persistence.

How to read it

Deterioration found early is cheap: check for external causes first (tracking, stock, seasonality, a landing-page change) before touching the campaign.

What it does not mean

One bad day is not deterioration. The detector requires a pattern, and even then it flags investigation, not automatic action.

Engagement deterioration

What it detects

Falling CTR and rising frequency together — the audience has seen the creative too often and is responding less. The classic creative-fatigue pattern.

How to read it

The fix is usually fresh creative or a rested audience, not a budget change. Compare frequency against the campaign’s own history.

What it does not mean

There is no universal frequency cap. Retargeting tolerates frequencies that would exhaust a cold audience.

Funnel signals

What it detects

Breaks in the chain between impressions, clicks, and conversions — for example healthy CTR with collapsing conversion rate, which points past the ad to the landing page or offer.

How to read it

Locating the break tells you where to work: creative (CTR), page (CVR), or economics (CPA vs margin). Each break has a different owner.

What it does not mean

A funnel signal identifies where the problem lives, not always what it is — the page, the price, and the tracking all deserve a look.

Confidence and priority

Two scores turn raw findings into a work list. Confidence reflects how much evidence supports the finding — data volume, window length, and consistency. Priority ranks what to do first by expected impact: how much spend or revenue the pattern touches, weighted by confidence. High priority plus high confidence is where an advertiser’s next hour is best spent.

Work an example yourself with the free ROAS calculator and CPA calculator, look up any term in the glossary, or follow the step-by-step wasted-spend guide.

Methodology maintained by Adsevon Editorial · Last reviewed October 23, 2026. Detector internals (exact thresholds and scoring weights) are product implementation details; the concepts, evidence model, and limits described here are the public contract.