Text version of this lessonExpand
Google Ads automation depends on conversion and value signals. Many accounts do not fail because bidding tools are missing. They fail because purchase value, currency, transaction ID, refunds, consent status, or enhanced-conversion fields are not trustworthy. Enhanced conversions can improve matching quality, but they do not define business value for you. This lesson gives you an enhanced conversions and value QA sheet plus a value fault router for deciding what can be trusted by tROAS or Maximize conversion value.
Lesson output: enhanced conversions and value QA sheet
This sheet puts technical implementation and business value into the same evidence table. Google Ads can learn from the conversion value you send. If that value is pre-discount revenue, mixed currency, duplicate orders, or pre-refund gross revenue, automation learns the wrong target more consistently.
| Check layer | Inspect first | Pass condition | If failed |
|---|---|---|---|
| Match fields | Where email, phone, name, and address come from; whether they appear reliably and respect consent | Field source, format, and consent status are explainable | Do not upload dirty fields or bypass user choice |
| Deduplication | Transaction ID, order ID, duplicate purchase, confirmation refresh, and multiple tags | One order counts once and reconciles across Ads / GA4 / Shopify | Do not trust value automation while duplicate orders remain |
| Value definition | Product revenue, tax, shipping, discounts, pre-refund / post-refund revenue have fixed rules | Purchase value and business review definitions can explain differences | Do not switch to aggressive tROAS before value definition is fixed |
| Currency and market | Currency, market currency, conversion logic, and Google Ads billing currency | Cross-market values can be compared and currency code is stable | Do not make cross-market scale calls while currency drifts |
| Privacy and consent | Consent Mode, cookie banner, privacy policy, tag behavior, and data-use explanation | User choice and platform sending behavior are aligned | Do not sacrifice consent boundaries for match rate |
Define the terms before using the sheet
Enhanced conversions use hashed first-party customer data to improve conversion matching. They are not a new conversion event. Email, phone, name, or address fields can help Google match an existing conversion back to an ad interaction, but they cannot fix a wrong purchase event, duplicate order, bad value, bad currency, or refund definition.
Hashed first-party data is customer information you collected directly, transformed by SHA256 or a similar one-way hashing method so the original text is not directly readable. Source, formatting, and consent boundaries still need to be explainable.
Transaction ID is the unique order or lead identifier used to help prevent duplicate conversion counting. Without a stable transaction ID, confirmation-page refreshes, multiple tags, or duplicate scripts can count one order more than once.
Purchase value is the order value sent to the ad system. You need to define whether it includes product revenue, tax, shipping, discounts, refunds, and canceled orders. If value is wrong, automation becomes more consistently wrong.
Currency is the currency code attached to value, usually an ISO 4217 three-letter code such as USD, CAD, or EUR. Cross-market accounts can easily distort ROAS, tROAS, and budget decisions when currency is missing or mixed.
Consent status is the user choice about advertising, analytics, or personalization data use. Tags and platforms adjust collection, modeling, and sending behavior based on consent state. Match rate should not override user choice.
Consent Mode is Google's mechanism for adjusting tag collection and modeling behavior according to user consent state. You see it around the cookie banner, Google tag, GTM, and privacy settings. If consent state and tag behavior disagree, enhanced-conversion match quality and privacy boundaries both become unreliable.
checkout is the path where a shopper moves from cart to payment, address entry, shipping selection, and completed purchase. It is where purchase value, transaction ID, customer fields, and consent state often meet. If checkout fields are missing or fire twice, value QA becomes dirty at the source.
Enhanced conversions solve recognition, not business definition
| Layer | Solves | Does not solve |
|---|---|---|
| Enhanced conversions | Improves the stability of recognizing and matching existing conversions | Does not define net revenue, refunds, margin, new-customer weight, or low-quality orders |
| transaction ID | Helps one order avoid duplicate counting | Does not guarantee order value, currency, or refund definition is correct |
| value-based bidding | Lets the system optimize toward conversion value or ROAS goals | Does not know margin, cash recovery, or refund risk unless value definition is designed clearly |
Sample at least 20 recent orders on one row across Ads, GA4, Shopify, and finance
The most dangerous state is not empty value. It is value that looks mostly normal while the business definition has drifted. Your sample table should include order ID, Ads value, GA4 purchase revenue, Shopify gross / net sales, discount, tax, shipping, refund, chargeback, currency, and product group.
| Field | How to compare | Risk signal |
|---|---|---|
| Google Ads value | Compare row by row with GA4 purchase, Shopify gross / net sales, and finance net revenue | Ads revenue rises while backend net sales do not |
| currency | Check USD / CAD / EUR codes and converted values by market | Cross-market orders mix currencies and distort ROAS |
| discount / shipping / tax | Confirm discounts are deducted and tax/shipping inclusion follows fixed rules | Value looks normal but the business definition is too optimistic |
| refund / chargeback | Connect refunds, chargebacks, high-risk SKUs, and net revenue review | High-refund orders keep training the system on gross revenue |
Value fault router: locate the fault before scaling it
Value quality problems rarely appear in one dashboard only. Route the symptom first, then decide whether to fix the tag, fix currency, fix deduplication, add refund review, or pause automation trust.
| Symptom | Likely fault | First evidence | Do not do |
|---|---|---|---|
| Ads value above Shopify net sales | Pre-discount value, tax/shipping-heavy gross value, or refunded orders are still training at original value | Sample the latest 20 orders and compare Ads, GA4, Shopify, refunds, and discounts row by row | Do not raise tROAS trust while net revenue does not reconcile |
| Currency drifts across markets | Currency is missing, the code is wrong, conversion logic drifts, or account currency and order currency are blended | Sample USD / CAD / EUR orders by market and check value, currency, and converted value | Do not use mixed-currency blended ROAS to decide cross-market scaling |
| Duplicate purchase for one order | Thank-you page refreshes, multiple tags, duplicate client/server reporting, or unstable transaction ID | Pick 10 real orders and compare transaction ID across Ads, GA4, Shopify, and GTM / tag firing records | Do not treat duplicate purchase events as conversion-rate improvement |
| Refund and margin missing from value review | The system learns pre-refund gross value and does not know refund rate, margin, chargebacks, or support cost | Sample by SKU or product group and connect Ads value, refunds, chargebacks, margin tier, and net revenue | Do not let high-refund SKUs keep training automation at full value |
| Consent and match fields unclear | Fields appear only in some payment paths, formatting is inconsistent, or Consent Mode / cookie banner behavior does not match tag behavior | Use real checkout paths to inspect field source, pre-hash formatting, consent state, privacy policy, and tag behavior | Do not bypass user choice or upload dirty fields for match rate |
20oz value trust lab: choose the order issue, then choose the repair action
In a real account, value quality is not one switch. It is a set of order-level evidence. The same 20oz tumbler can expose discount-value drift, Canada currency drift, duplicate purchase after a thank-you-page refresh, and high-refund SKU risk. Do not let better-looking ROAS in Google Ads become the only reason to trust tROAS.
| Order scenario | First evidence | Repair first | Write back to the sheet |
|---|---|---|---|
| 20oz tumbler discount order | Shopify collected $72.80, Ads value received 89.00, and GA4 purchase revenue is 72.80 | Fix value definition across discounts, tax, shipping, and pre-refund / post-refund revenue | Value gate failed; do not raise tROAS trust |
| Canada order currency drift | Shopify shows CAD 96, Ads value is 96, but currency is missing and the account reads it as USD | Fix currency passing and conversion rules | Currency gate failed; pause cross-market scale decisions |
| Thank-you page duplicate purchase | The same transaction ID appears twice in GTM preview and GA4 DebugView, while Shopify has one order | Fix transaction ID dedupe, GTM triggers, and duplicate client/server reporting | Dedupe failed; current value cannot enter automation trust |
| High-refund SKU keeps spending | The 20oz gift set has stable ROAS, but 14-day refund rate is 28% and most tickets mention leaking lids | Add post-refund revenue, chargebacks, margin tier, and SKU quality review | Post-refund revenue gate failed; fix product quality or review definition first |
During the drill, do not default to raise tROAS trust. Only test value-based automation after field source, dedupe, value, currency, post-refund review, and consent boundaries stay stable. Otherwise you are not scaling good orders; you are scaling a bad value definition.
Green diagnostics do not prove the business value definition is correct
Diagnostics can reveal matching, coverage, and implementation issues. They do not tell you whether value should include tax, subtract refunds, or weight new customers. Use diagnostics as a severity signal, not as proof that the business definition is correct.
| Severity | Signal | Action |
|---|---|---|
| Low | Diagnostics look normal, but business value definition is still under review | Keep sampling and do not raise automation trust yet |
| Medium | Match coverage or implementation status fluctuates, while value mostly reconciles | Pause optimistic assumptions and inspect fields, tags, and paths |
| High | Order value, currency, dedupe, or refund logic clearly drifts | Do not trust tROAS, Maximize conversion value, or weighted value until fixed |
Value automation needs trust levels
| Trust level | Condition | Allowed action |
|---|---|---|
| Do not trust | Purchase, value, currency, transaction ID, or consent boundary is unclear | Fix tracking only; no value automation |
| Low trust | Events fire, but sample is low and value definition is still being reconciled | Observe conservatively; do not raise tROAS trust |
| Observation trust | Sample orders reconcile and diagnostics show no High risk | Keep weekly net-revenue / refund review |
| Testable trust | Fields, dedupe, value, currency, refund, and sampling remain stable | Carefully test tROAS or value-based automation with rollback lines |
30-minute value QA review: put tech, operations, and finance on one sheet
This lesson should not end with knowing what enhanced conversions are. The usable output is a value QA review that the ad lead, developer, operator, and finance lead can all read. The meeting does not debate whether to trust Google Ads. It checks whether the amount, currency, order identity, and refund treatment being learned by the system match the business facts.
| Time | Question to confirm | Evidence to leave |
|---|---|---|
| 0-5 min | Date window, market, campaign, conversion action, and order sample scope | 20-order sample sheet with Ads / GA4 / Shopify / finance sources labeled |
| 5-12 min | Whether transaction ID is stable and one order appears once across thank-you page, GTM, and server-side paths | Transaction ID screenshots or exports for 10 real orders |
| 12-20 min | Whether purchase value and currency reconcile row by row across discounts, tax, shipping, and pre-refund / post-refund revenue | Value difference notes that separate valid definition differences from implementation errors |
| 20-26 min | Whether high-refund SKUs, chargebacks, or low-margin products still train the system at full value | Post-refund revenue or SKU quality layer, plus any exclusion, segmentation, or value-weighting decision |
| 26-30 min | Whether this week's tROAS state is stop, low trust, observation trust, or testable trust | Responsible lead, repair action, review time, and rollback line |
If the meeting ends with the data looks okay, the lesson is not finished. A useful closeout is one sentence: because these orders prove value, currency, dedupe, and post-refund revenue are stable, this action is allowed this week; or because this gate failed, the team fixes tracking only and does not hand value to tROAS yet.
The common beginner mistake is treating higher conversion value in Google Ads as business growth, treating enhanced-conversion recording status as a profit signal, or treating one or two days of ROAS movement as permission to scale. Start with three questions: is this the same order value, is this the correct currency, and is this order still worth learning from after refunds?
Copyable lesson notes: turn value QA into next-week accountability
This lesson is easy to misread as a technical switch. Do not only write "enhanced conversions are on." Copy a value trust record instead: whether the system can learn from value, which gate is blocking trust, who fixes it, and when the sample is checked again.
Google Ads value QA copyable lesson notes
Current pressure: Do not raise tROAS trust just because enhanced conversions are recorded, Diagnostics looks green, or ROAS improved.
First proof: Sample the latest 20 orders and put Google Ads value, GA4 purchase revenue, Shopify gross / net sales, discounts, tax, shipping, refunds, currency, and transaction ID on one row.
Current gate: If match fields, dedupe, value definition, currency, post-refund revenue, Consent Mode, or checkout field source fails, fix signals only and do not scale.
This week's action: Choose one repair move: fix value definition, fix currency, fix transaction ID dedupe, add post-refund revenue, or keep the account in low-trust observation.
Stop action: Do not let Maximize conversion value or tROAS scale this value set while net revenue, currency, duplicate purchase, post-refund revenue, or consent boundary fails reconciliation.
Review window: Re-sample on the same 7-day basis next week. Do not replace order-level evidence with dashboard screenshots.
Next route: If the value gate passes, move to when and how to scale. If it fails, return to conversion tracking setup and metric reading to complete the evidence.
Stop / Go rules: do not let tROAS scale wrong value
| Stop | Go | Proof needed |
|---|---|---|
| Order value cannot reconcile to Ads value | Sample orders explain Ads / GA4 / Shopify / finance differences | Last 20 orders reconciliation sheet |
| Currency drifts, purchase duplicates, or transaction ID is unstable | Currency and transaction ID are stable in real order paths | Cross-market order samples and tag test records |
| High-refund orders continue training the system at full gross value | Refund, chargeback, and net-revenue review are part of the value definition | Post-refund revenue or finance review layer |
| Diagnostics and business checks repeatedly conflict | Diagnostic severity, responsible person, fix action, and review window are written | Value-signal stop/go record |