Birinchi Shop

Technical checklist for evaluating a CPA network

You can check a network's technical side yourself before you start — and that takes public docs, public numbers and working links, not marketing claims. Below are 17 neutral criteria: why each matters, how to check it, and where Birinchi Shop stands — with an evidence link. What we lack or have only partly is stated openly too.

Sign up free
1 minute · no card or documents
30,000+UZS per delivered order
14 minmedian payout time (last 90 days)
350+active products
0 UZSto start
Data as of 2026-11-21. Payout term: within 1 day; 14 min is the actual median over the last 90 days. How the numbers are counted

Criteria table

The criteria apply to any CPA network. The "Birinchi Shop" column is our own assessment: "Yes" — works and can be verified, "Partial" — with a limitation (described), "No" — not available yet. Every row has an evidence link you can open yourself.

Each criterion: why it matters, how to check, Birinchi Shop

1. Public API and documentation Yes

Why it matters. You need an API with public docs to automate traffic, submit orders from your own CRM or site and pull reports programmatically. A closed API cannot be evaluated in advance.

How to check. Do the docs open without signing up; is there a machine-readable spec (OpenAPI); are error codes, scopes and rate limits documented?

At Birinchi Shop. REST API v1: offers, links, orders, stats, an event feed and webhooks. Docs, the OpenAPI spec and a single-file guide for AI coding assistants are public without a key. Keys get only the scopes you select. Order submission via API unlocks after 3 delivered orders or with admin approval.

Evidence: API docs · OpenAPI spec · AI-ready guide (guide.md) · Reference (statuses, scopes)

2. Webhooks and S2S postbacks Yes

Why it matters. If a conversion never reaches your tracker, your ads optimize on wrong data. Postbacks should fire on every status change, be verifiable as genuine and be retried when your server does not answer.

How to check. Which statuses trigger it; the macro list; is there a signature (HMAC); the retry schedule; can you send a test event?

At Birinchi Shop. On every status change: an S2S postback (GET, 20 macros: click_id, sub1–sub5, status, payout, payout_earned, etc.) or a JSON webhook signed with HMAC-SHA256. Without a 2xx answer it is retried up to 8 times (1, 5, 15, 60 minutes and so on, about 2 days). A "Test" button in the panel sends a test event. Signature verification examples are on GitHub (Node, Python, PHP).

Evidence: API docs · GitHub: webhook signature check · API & postback guide

3. Recovering missed events Yes

Why it matters. If your server is down for a few hours, webhook retries may run out. You then need a way to fetch the events yourself later.

How to check. Is there a cursor-based event feed, and how long are events kept?

At Birinchi Shop. GET /events — a cursor-based event feed (order created, status changed); events are kept for 90 days. Every event has an evt_ ID for de-duplication.

Evidence: API docs · AI-ready guide (guide.md)

4. Tracker integration (click_id, sub1–sub5) Yes

Why it matters. Media buyers track every click in their own tracker (Keitaro, Binom, Voluum); the network must keep click_id and sub parameters through to the order and return them in the postback.

How to check. Append ?click_id=…&sub1=… to a link, place an order and check the values come back in the postback; is there a ready setup guide?

At Birinchi Shop. click_id and sub1–sub5 are stored with the order and returned in postbacks/webhooks; API orders can also carry your own ID (external_id). Any tracker connects via the universal S2S postback; there is no separate native integration (preset template). Keitaro and Binom guides are on GitHub; for revenue in your tracker use the {payout_earned} macro (amount only on delivery, otherwise 0).

Evidence: GitHub: Keitaro setup · GitHub: Binom setup · API & postback guide

5. Real time in the panel Yes

Why it matters. Seeing new orders and status changes without reloading the page makes daily work faster.

How to check. Does a new order appear by itself while the panel is open, or do you have to reload?

At Birinchi Shop. An open panel shows a new order, status change, balance change or notification within 1–2 seconds (Server-Sent Events).

Evidence: Technical capabilities

6. Independent uptime monitoring and a public status page Yes

Why it matters. A self-declared "99.9%" cannot be checked. A public status page run by an independent service shows outages to everyone.

How to check. Is there a public status page, is it run by an external service rather than the network itself, and which parts (site, API, panel) are checked?

At Birinchi Shop. Since 2026-11-18 an external monitoring service checks the site, API and panel every 5 minutes; results are public at /status. The history starts on that date. We do not publish an uptime percentage ourselves.

Evidence: Status page (/status)

7. Public technical metrics Yes

Why it matters. Claims about API speed, server errors and postback delivery are only checkable when there are numbers and a stated method.

How to check. Are numbers published; is the method described; is there a machine-readable version; are unmeasured values left blank rather than estimated?

At Birinchi Shop. API latency (P50/P95), 5xx rate, postback delivery, duplicate rate, status freshness, payout speed and site speed — with the methodology, on a page and as JSON. The platform measures these itself; they are not externally audited; rows without enough data say "measuring".

Evidence: Technical metrics · Metrics (JSON)

8. Backups and recovery Partial

Why it matters. If a server fails, orders, balances and stats must not be lost and the platform must come back quickly.

How to check. How often are backups taken; are copies stored separately; is there a standby server; is switching to it automatic?

At Birinchi Shop. A live copy of the database runs on a separate standby server (a few seconds behind); an encrypted database backup is made every hour, with daily copies stored in separate locations; files, code and settings are copied to the standby server hourly; an alert fires if a backup is missed or the live copy disconnects. On 2026-11-15 a full restore drill was carried out — the database was restored from backup in a separate environment and verified to work. Switching to the standby server is manual — there is no automatic failover, hence "Partial".

Evidence: Technical capabilities

9. Security practices Yes

Why it matters. The network holds your account, API keys, payout documents and customer data.

How to check. HTTPS only; are passwords and keys hashed; do keys have scopes; do AI assistants connect via OAuth; can partners see customer phones; has there been an external audit?

At Birinchi Shop. HTTPS only (via Cloudflare), Content-Security-Policy; passwords hashed with bcrypt, API keys stored only as hashes; scoped keys; a password reset revokes keys and AI connections; AI assistants connect via OAuth (PKCE); payout documents are encrypted; partners see only the last 4 digits of a customer phone. No external security audit or penetration test has been done.

Evidence: Data protection

10. A way to report vulnerabilities Yes

Why it matters. If a researcher does not know where to report a hole, it stays open or goes public.

How to check. Is there a /.well-known/security.txt (RFC 9116) and a policy with scope and rules?

At Birinchi Shop. /.well-known/security.txt (contact, languages, policy link) and a public policy: scope, what not to do, how we respond. There is no paid bug bounty.

Evidence: security.txt · Vulnerability disclosure policy

11. AI and MCP integration Yes

Why it matters. Partners work with AI assistants such as ChatGPT and Claude: picking offers, creating links and checking stats from a chat saves time.

How to check. Is there an MCP server; secure connection (OAuth); docs written for AI; can the AI spend money without confirmation?

At Birinchi Shop. Partner MCP server (/api/v1/mcp): ChatGPT or Claude connect to a partner account via OAuth — offers, links, orders, stats. Resuming ads or changing budgets requires the partner's confirmation. A separate shop MCP server serves shoppers. Servers are listed in /.well-known/mcp.json.

Evidence: mcp.json · AI-ready guide (guide.md) · Shop MCP server

12. Click log and tracking accuracy Yes

Why it matters. CR and ad performance are only right if clicks are counted right: bots and link previews must be filtered out, and the share of lost orders must be known.

How to check. Are clicks counted server-side; are bots filtered; are there unique clicks; is the share of orders linked to a click measured?

At Birinchi Shop. Since 2026-11-18 every click is logged: bots, link previews and prefetches are separated, unique clicks (30-minute window), CR by period; raw IPs are not stored. API /stats returns clicks and unique_clicks. "Tracking accuracy" (share of orders with a recorded click) shows "measuring" until 30 orders are collected.

Evidence: Technical metrics

13. Duplicate and retry protection Yes

Why it matters. If one customer is counted twice, or a request is resent after a network error, stats and payouts get distorted.

How to check. Is there an Idempotency-Key; how does the one-phone-one-product rule work; is the double-counted share published?

At Birinchi Shop. API Idempotency-Key (a retried request returns the existing order); one phone + one product = one order per 24 hours; for part of the catalog a new order is refused while the customer's earlier order for that product is in progress; a simultaneous double submit creates one order; Meta lead IDs are unique. The duplicate rate is on the Technical metrics page.

Evidence: Technical metrics · API docs

14. Anti-fraud Yes

Why it matters. Fake orders raise call-center and delivery costs and drag down the product sale rate for honest partners.

How to check. Which checks exist: self-orders, abnormal flows, bot clicks; is there ML-based scoring?

At Birinchi Shop. Rule-based: self-order ban (panel, API, AI chat and link form), holding orders for staff review when a partner's confirmation rate is very low, burst protection for external lead flows, rate limits, bot click filtering. In addition, an ML model trained on our own order history scores the sale probability of every new order (advisory; it never rejects an order by itself).

Evidence: Technical capabilities

15. SDKs and code samples Yes

Why it matters. Ready code cuts an integration from hours to minutes.

How to check. Is there an official SDK, for which languages, and is it published on package registries (npm, PyPI); are there working samples and a Postman collection in a public repo; do they match the OpenAPI spec?

At Birinchi Shop. Official SDKs, version 0.1.0 (beta): Node.js (18+, zero dependencies, TypeScript types, webhook signature verification) and Python (3.8+, standard library only, type hints). Install: npm install birinchishop and pip install birinchishop (npm and PyPI); source code on GitHub. The examples repo has curl, Node.js, Python and PHP samples, webhook signature verification, a Postman collection, Keitaro and Binom guides, MCP setup and a copy of the OpenAPI spec.

Evidence: npm: birinchishop · PyPI: birinchishop · Node.js SDK (GitHub) · Python SDK (GitHub) · Postman collection

16. Test environment (sandbox) Yes

Why it matters. You need to test an integration without creating real orders — in a COD network the call center phones every order.

How to check. Is there a separate test key or mode; are its orders kept away from the call center; can you test webhooks and postbacks with status changes?

At Birinchi Shop. Since 2026-11-18 a test key (bsk_test_…) can be created in the panel. Read requests return real data; POST /orders and POST /links run the same checks as with a live key, but no order or link is created and nobody is called — the response has "test": true and an X-Birinchishop-Mode: test header. A signed test webhook (order.created) follows a test order; POST /api/v1/sandbox/orders/{id}/status simulates status changes (webhook and postback, {test} macro = 1). Test objects are kept for 24 hours (cleared if the server restarts); a test key cannot manage ads; MCP works with a test key too.

Evidence: AI-ready guide (guide.md) · API docs · OpenAPI spec

17. Changelog and API versioning policy Yes

Why it matters. If an API changes silently, integrations break. A dated changelog and a versioning rule tell you what to expect.

How to check. Is there a dated changelog; how is the version indicated; how are breaking changes announced?

At Birinchi Shop. A dated changelog (API, MCP, tracking, infrastructure) and a versioning policy are public. The version is in the path (/api/v1) and in the X-API-Version header of every response. Since 2026-10-22 no endpoint or field has been removed from v1 — changes have been additive only.

Evidence: Changelog · OpenAPI spec

Start today

Sign up, pick a product and get your first tracking link.

Become a partner

A 10-minute hands-on check

You can check any network in this order. Birinchi Shop links are in brackets.

  1. Open the docs. Do the API docs and OpenAPI spec open without signing up (/api/v1/docs, /api/v1/openapi.json)?
  2. Find the status page. Is it run by an external service (/status)?
  3. Ask for numbers. Are API speed, errors and postback delivery published with a method (Technical metrics, JSON)?
  4. Check security.txt. Is there a /.well-known/security.txt and a disclosure policy (security.txt, policy)?
  5. Run a code sample. Are there samples, an official SDK and a Postman collection in a public repo (examples, Node.js SDK, Python SDK)?
  6. Test a postback. Open an account, add your tracker's postback URL and send a test event — did it arrive, is the signature valid? Is there a test key for a full run without real orders (guide.md, "Test mode")?
  7. Read the change history. Is there a dated changelog and a versioning rule (Changelog)?

What Birinchi Shop does not have

  • Automatic failover. A standby server is kept ready, but switching to it is manual.
  • An external security audit. No penetration test or certification; the technical metrics are our own measurements too.
  • Native tracker presets. Trackers connect via the universal postback.

Full list: Technical capabilities.

About this page

We chose the criteria, but they are not tailored to any network: ask any network the same questions. The "At Birinchi Shop" part is our own assessment, not an external audit — which is why every claim comes with a link you can check. The page is updated when things change; dated changes are in the changelog.

Updated: 2026-11-18. Measured values: Technical metrics; all capabilities: Technical capabilities.

FAQ

How do you evaluate a CPA network technically?

Look at what can be verified, not at promises: public API docs and an OpenAPI spec, signed webhooks and S2S postbacks with retries, a way to recover missed events, independent uptime monitoring, public technical metrics, backups, security practices and security.txt, tracker and AI/MCP integration, code samples, a test environment and a changelog.

How can I verify Birinchi Shop's technical claims?

Every criterion has an evidence link: API docs (/api/v1/docs), the OpenAPI spec, code samples on GitHub, an independent status page (/status), technical metrics (with JSON), security.txt and the changelog. All of them open without signing up.

What does Birinchi Shop not have?

Automatic failover, an external security audit and native tracker presets.

Where is the uptime percentage?

We do not publish an uptime percentage ourselves. Since 2026-11-18 an external monitoring service checks the site, API and panel every 5 minutes; results are public at birinchishop.uz/status.

Can I check postbacks without signing up?

Macros, the signature and the retry schedule are in the public docs and the GitHub samples. To send a real test event to your tracker you need an account: add the postback URL and press "Test", or send a test order with a test key (bsk_test_) and simulate status changes — no real order is created.

Is this assessment independent?

No — the "At Birinchi Shop" part is our own assessment. That is why every claim has a verification link and what we lack is stated openly.

Earn on every delivered order

Warehouse, call center and delivery are on us. You bring the traffic.

Sign up free

More in English

All English pages →