Conversion Tracking Setup

A conversion tracking setup begins with an event definition that a sales, ecommerce, app, or lead-generation team can recognize outside the advertising dashboard. “Form submitted” may be a useful technical event, but it is not automatically the business event. A stronger definition states whether the lead was accepted, whether the order was paid, whether the app user completed a valuable action, or whether the subscription remained active. Write the event name, owner, value rule, rejection rule, and expected delay in plain language. That document becomes the reference when pixels, postbacks, analytics, and CRM totals disagree.

Separate the primary optimization event from supporting events. A purchase or qualified lead may be the primary event, while product views, button clicks, registrations, or checkout starts help diagnose the journey. Do not let a high-volume supporting event replace the business goal simply because it gives an algorithm more data. The useful hierarchy is diagnostic event, intent event, conversion event, and validated outcome. Every campaign report should make that hierarchy visible so the media team knows which signal is safe for troubleshooting and which signal can support a budget decision.

For a lead campaign, write down the difference between a submitted form and an accepted lead before installing the event. Decide which state you can report reliably and how the later qualification result will be reviewed. A form submission can help diagnose the journey while qualification is pending. We recommend keeping that provisional signal separate from the accepted result that justifies a larger FroggyAds allocation.

Draw the conversion path from ad interaction to final record. Include the campaign ID, creative ID, source or zone ID, click ID, session or device identifier, event ID, order or lead ID, timestamp, and final status. Mark where each identifier is created, transformed, stored, or dropped. Redirects, cross-domain journeys, payment providers, app stores, CRM imports, and privacy settings are common breakpoints. A simple map often reveals that the advertising platform is not missing a conversion; the connection between two systems is missing the key needed to match it.

Where the integration supports an event or transaction identifier, keep it stable across signals describing the same action. Otherwise follow the receiver's documented counting or priority method and retain an independent reconciliation record. Test a browser signal, a server notification and an allowed repeat to see what the receiver actually records. Internal duplicate detection alone does not establish how an external platform counts the event or represents a later refund.

Test a repeated request and a separate legitimate action. An order identifier should allow your supported receiver or reconciliation process to distinguish another notification about the same order from a second real purchase. Keep the original click association where available, but do not use one click as the universal duplicate key: a customer can make more than one legitimate purchase after a visit. Verify the integration's actual counting rules before relying on that distinction.

A browser pixel is often the fastest way to measure an on-page action. It can read page context, work with a tag manager, and support quick QA in the browser. Its weakness is dependence on browser execution, storage, consent, network conditions, and the user remaining on the page long enough for the request to complete. Use it when the event truly occurs in the browser and the loss risk is acceptable. Do not assume that a successful thank-you page view proves a qualified lead or paid order.

Server-side postbacks suit actions confirmed in your backend, including outcomes approved after the visitor has left. Send only the fields supported by the actual receiver, preserving the required click reference and business-event identity. Follow its documented delivery, counting and security rules, and verify that browser and backend signals for one order do not appear as separate sales.

For a postback, test what happens when the response is missing or delayed. The receiver may have processed the event even if its acknowledgement did not arrive. Inspect the receiving record where possible and follow its documented repeat rules before sending again. Preserve event identity across an allowed retry, and keep unsupported status or correction cases visible. An invented parameter does not create duplicate protection or cancellation support.

Create test cases for a normal conversion, duplicate submission, invalid form, refunded purchase, delayed approval, cross-device return, blocked storage, consent denial, and abandoned journey. Each case needs an expected result in the advertising platform and in the business system. A test is not complete when a tag debugger shows a request. It is complete when the event arrives with the right identifiers, value, currency, timestamp, and validation status. Save screenshots or log references so another person can reproduce the result after a future site change.

Run the matrix again after changes to checkout, forms, tag management, consent tools, domains, app releases, CRM fields, or payment providers. Conversion tracking often fails quietly: the page still loads and campaigns still spend, but values disappear or every event becomes unattributed. Add a small monitoring alert for a sudden drop in match rate, a spike in duplicates, missing values, or a large difference between platform and backend totals. Fast detection protects both optimization and reporting credibility.

Include an outcome whose final state changes after the first notification. For example, a submitted order may later be paid, rejected or refunded. Verify which states the supported integration can represent, which remain in your backend and when the accepted count becomes suitable for a buying decision. Keep the original event time and later status history so a delayed approval is not mistaken for a new customer action.

Maintain a daily or weekly reconciliation table with platform conversions, unique business records, accepted records, rejected records, revenue, refunds, and the number of records that could not be matched. Differences do not automatically mean one system is wrong. Attribution windows, time zones, view-through credit, modeled events, deduplication, and late status changes can all create legitimate gaps. The goal is to explain the gap well enough that the media team knows which number should guide bidding and which number belongs only in channel reporting.

Measure match rate separately from conversion rate. A strong conversion rate with a poor match rate can hide lost attribution. A high match rate with weak lead quality can make the technical setup look healthy while the campaign fails commercially. Review event volume, duplicate rate, unmatched rate, conversion delay, qualified rate, revenue, and cost per validated outcome together. When those metrics move in different directions, investigate the event definition and source mix before changing bids.

Send only the information required for attribution and measurement. Click IDs, event IDs, timestamps, values, and status codes are usually more useful than personal details. Avoid putting email addresses, names, phone numbers, or sensitive information in URLs, query strings, or ad-platform parameters. Define retention, access, and deletion rules for the reconciliation data. Consent and privacy requirements vary by market and implementation, so the tracking design should be reviewed with the organization’s legal and privacy responsibilities in mind.

Treat missing identifiers as a diagnostic category, not as proof of fraud or low quality. Privacy settings, browser restrictions, app-to-web transitions, and legitimate cross-device behavior can break a deterministic match. Separate “unmatched,” “invalid,” and “not yet mature.” That classification prevents the media team from blocking useful traffic because of a technical gap. It also makes the real tracking problem visible instead of hiding it inside a generic failed-conversion bucket.

Create campaign cells that preserve the variables you intend to optimize. If source-level decisions matter, source IDs must survive into the final outcome table. If device or browser experience matters, those dimensions must be available in both media and conversion records. Start with one stable conversion definition and avoid changing the event, landing page, bid logic, and targeting at the same time. A clean change log should state what changed, why it changed, the expected effect, and the review date.

Scale only after the setup survives normal traffic. Test volume can pass QA while real users expose redirects, slow pages, unusual devices, duplicate retries, and delayed business updates. Increase spend gradually, monitor match and duplicate rates, and keep a control campaign visible. If platform conversions rise while validated outcomes fall, pause the expansion and inspect source-level records. Tracking is valuable when it improves a decision, not when it merely increases the amount of data collected.

Before launch, confirm one primary event, supporting events, event ownership, value and currency rules, click-ID capture, event-ID deduplication, attribution window, time zone, consent behavior, negative tests, delayed outcomes, and the reconciliation process. Confirm that every internal stakeholder uses the same event names. A media buyer, developer, analyst, and sales team should be able to point to the same definition and explain when the event becomes final.

After launch, review event match rate, duplicate rate, unmatched records, delay distribution, qualified rate, revenue or value, source-level economics, and recent implementation changes. Keep the checklist short enough to use regularly. The best setup is not the one with the most tags. It is the one that produces a stable, auditable connection between paid media and the business result the advertiser actually values.

Use our FroggyAds formats and available targeting settings to define the traffic you want to evaluate. Campaign budgets and source controls make the first purchase manageable; device, browser or GEO information can support a more detailed comparison where your setup retains it. Confirm the fields available to the chosen format rather than assuming every dimension reaches every tracker or business record.

Adscore signals and our traffic controls help you review the traffic purchase. Your own order, lead or affiliate-approval records establish which outcomes are accepted. Keep those jobs connected without treating either one as a substitute for the other. A technically delivered event can still be provisional, while a real business outcome can remain unmatched to a campaign if an identifier was lost.

Our published minimum deposit is USD 50, which funds the account rather than defining the required spend for every campaign. Current published starting bids are USD 0.003 CPC for Push and Native, USD 0.10 CPM for Display and USD 0.0001 CPC for Pop. Use the starting bid, selected buying model and your own spending ceiling as separate planning inputs. Actual auction costs and accepted-outcome economics must come from the campaign you run.