How does offer analytics work in CPA systems

1) What is offer analytics and where does it live

In CPA systems, "offer" is a product configuration (brand/geo/land/payout model/validity rules) for which the platform counts conversion, cost, and quality. Offer analytics is the collection and processing of events for each offer, comparison of sources/creatives, indexing rates and making decisions about scaling/stop.

Key objectives:

See efficacy: CTR, CR (click→reg→KYC→FTD), CPA, ROAS/ROI.
Quality control: D7/D30 retention, 2nd-dep rate, chargeback/refund, NGR.
Manage the economy: payout braces, caps/pacing, freezes/disputes.
Protect: deduplication, anti-fraud, FTD validity rules.

2) Offer data scheme (minimum contract)

Offer: `offer_id`, `brand`, `geo`, `device`, `landing_id`, `payout_model (CPL/CPA/Hybrid/RS)`, `payout_rules`, `cap/day`, `compliance_guides`.

Traffic touchpoints: `click_id`, `sub_id/aff_id`, `utm_`, `creative_id`, `placement`, `device/os`, `ip/asn`.

Events (S2S): `registration`, `kyc_approved`, `deposit_success {amount,currency,is_ftd}`, `second_deposit`, `refund/chargeback`.

Finance: `payout_calc`, `hold`, `schedule`, `currency`, `fx_rate(date)`.

Quality: `risk_flags`, `fraud_score`, `cohort_metrics`.

💡 All timestamps - UTC, amount currency - ISO-4217 + showcase with conversion to "report currency."

3) Attribution, deduplication, and validity

Attribution key: 'click _ id' (created by the network/partner redirector).

Event idempotence: unique 'event _ id' → ≡ retry "already accounted for."

FTD deduplication: by '(account_id OR payment_fingerprint) + offer + window 30-90 days)'.

FTD validity rules (in the offer): minimum deposit, prohibition of bonus/zero replenishment, KYC status.

Source intersections: with a few clicks before FTD - policy (last click inside the attribution window) or data-driven on the tracker side.

4) Offer analytics metrics (operating layer)

Funnel:

`CTR = Clicks / Impressions`
`CR1 = Reg / Clicks`
`CR2 = KYC / Reg`
'CR3 = FTD/Reg '(or' FTD/KYC'for strict funnel)

Economics:

`CPA = Spend / FTD`
`ARPU_Dn = NGR_Dn / FTD`
`Payback = min{n: Cum_ARPU_Dn ≥ CPA}`
`ROAS = NGR / Spend`, `ROI = (NGR − Spend − Direct_Opex) / Spend`

Quality:

`2nd_dep_rate = Users_with_2nd_dep / FTD`
`Retention_D7/D30`, `Chargeback_rate`, `Refund_rate`

5) Pipeline events: from click to pay

1. Click: redirector creates' click _ id ', normalizes UTM, writes log.

2. Transition to land: propagation 'click _ id '/cipher; front without sensitive data.

3. Server events from operator: 'registration/kyc/deposit/...' on S2S network endpoint → → handler queued.

4. Offer rules: validation, attribution window, deduplication.

5. BI showcases: FTD (D1/D7/D30), NGR, Payback cohorts, write-offs/disputes.

6. Billing and payments: model calculation (CPA/Hybrid/RS), hold, reconciliation certificates.

6) Payout indexing and quality braces

CPA payment by offer is often associated with the quality of the cohort. Example of rules:

Base bracket: 'CPA = $120' if '2nd _ dep ≥ 25%' and 'ARPU _ D30 ≥ $90'.
Downgrade: 'CPA = $100' if '2nd _ dep <20%' or 'Chargeback> 3%'.
Increase: 'CPA = $140' if'ARPU _ D30 ≥ $110'and'D7 _ ret ≥ 40%'.
Indexing is recalculated by cohorts and sources. Decisions are recorded in the additional agreement to the offer.

7) SmartLink and auto-rooting of offers

SmartLink distributes the flow between offers within the class (one GEO/vertical). Offer analytics gives feedback to the router:

Signals: eCPA, CR to FTD, ARPU proxy, share of refusals/disputes, compliance flags.
Management: white/black-list offers, manual "pins," caps, pacing, priorities.
Goal: maximize Payback/quality while adhering to validity rules.

8) Anti-fraud and quality at the offer level

Click fraud: IP/ASN scoring, velocity rules, device fingerprinting, source lists.

Reg-fraud: document matches/CCM, packs of night registrations, low engagement.

FTD fraud: "bonus" deposits, returns, "carousel" of payments.

Reaction: auto-freezing of payments for the offer/sabu, investigation, reconciliation acts, payout adjustments.

9) Dashboards of offer analytics (mandatory recruitment)

1. Funnel by offer: kliki→reg→KYC→FTD by 'source/creative/placement/device'.

2. Quality/cohorts: ARPU D1/D7/D30, 2nd-dep, Retention, Payback.

3. Write-offs/disputes: deviations, chargeback/refund for reasons and sources.

4. Tracking stability: delay of postbacks, share of duplicates, EMQ/share of events without 'click _ id'.

5. Payout economics: actual CPA/Hybrid/RS vs braces, cash flow forecast (hold, schedule).

10) Alerts and SLAs for the offer

Track SLA: delay S2S> 15 minutes, error-rate> 1%, proportion of duplicates> 0.5%.

Business SLA: drop in'CR (reg→FTD)' by X σ, surge in chargeback, ARPU drawdown.

Compliance: deviations of creatives/lands, complaints, brand-bidding - auto-escalation.

Reaction: auto-stop/cap, manager notification, investigation launch.

11) Disputes and reconciliation acts

Standard procedure:

1. Reconciliation of volumes by 'click _ id/event _ id' for the period (UTC).

2. Comparison of FTD validity by offer rules.

3. Unloading of "border" cases (rejected/chargeback).

4. Final act, payout/RS adjustments, update of quality brackets.

12) Formulas and mini-example

Source: for 30 days by offer 'BR-Android'

Spend = 25 000; Clicks = 50 000; Reg = 4 000; KYC = 2 600; FTD = 600

GGR_D30/FTD = 130; bonuses = 15%; game providers = 10% GGR; payments = 3% of deposits; chargeback = 1% of deposits

Contractual CPA = $110 (basic bracket)

NGR_D30 calculation (per 1 FTD):

GGR = 130 → minus bonuses 19.5 (15%) → minus providers 13 (10%) → minus payment orders ~ 3.6 → chargeback minus ~ 1.2 ⇒ NGR ≈ 92.7
ARPU_D30 = 92,7; CPA actual = 25,000/600 = 41.7 (for network)
Payback (crude): average daily ARPU ≈ 92.7/30 = 3.09 ⇒ 41.7/3.09 ≈ 14 days
Quality bracket: if 2nd-dep_rate ≥ 25% and ARPU_D30 ≥ 90 → CPA indexing to $120.

13) Frequent errors in offer analytics

1. Count according to GGR, excluding bonuses/commissions → false ROAS.

2. No idempotency → FTD duplicates in retreats.

3. Mixing geo/devices in one offer → "average temperature," bad decisions.

4. Lack of chargeback/refund events → overstated ARPU.

5. Fuzzy FTD validity rules → eternal disputes, payment freezes.

6. Only EPC without cohorts and 2nd-dep → a fragile economy.

7. There are no alerts for delays S2S → data gaps, billing disruption.

14) Checklists

14. 1. Before the launch of the offer

Described'payout _ model', 'payout _ rules', attribution window, FTD validity
s2s-схема: `registration/KYC/FTD/2nd_dep/refund/chargeback`
UTC, currencies, fx table; idempotency по `event_id`
Antifraud threshold, white/black source list
Dashboards: funnel, cohorts, write-offs, SLA metrics
Alerts: delay> 15 min, CR anomalies, chargeback bursts

14. 2. Weekly routine

Reconciliation of "operator↔set" by events and amounts
Quality bracket update (by cohort)
Retro by creative/land/sources by offer
Update white/black-lists, analysis of controversial cases

15) 30-60-90 plan for the implementation of offer analytics

0-30 days - Frame and hygiene

Standardize the offer model and S2S events; Enable idempotency, UTC, currencies.

Raise the basic dashboards: funnel, D7/D30 cohorts, SLA tracking.

Enter the minimum payout brackets and anti-fraud thresholds.

31-60 days - Quality and economics

Add chargeback/refund, 2nd-dep, Retention to reports.

Set up CPA indexing by cohort quality, automate mouthguards and pacing.

Connect SmartLink signals (white/black list, priorities).

61-90 days - Sustainability and audit

Implement logs with the correlation 'click _ id/event _ id', DLQ and retray.

Perform S2S stress tests and reconciliation with BI/financials (NGR consistency).

Formalize playbooks of controversial cases and quarterly audits of offers.

Offer analytics is not a "lead table," but a system: a reliable S2S circuit, strict validity rules, a cohort economy and transparent payment braces. When each offer is described as an object with events, quality and finances, you quickly distinguish "noise" from growth, protect margins and scale only those bundles that really give Payback and long LTV.