Google Smart Bidding sees your WooCommerce purchase event at checkout and then never hears from that order again. It doesn’t know when a customer refunds 14 days later. It doesn’t see the cancellation. It doesn’t catch the chargeback at 45 days. So it treats every checkout value as a successful outcome and keeps finding lookalike customers — which includes more of the ones who returned. Brands switching from Target CPA to Target ROAS see a median 14% increase in conversion value at similar ad spend (Launchcodex, 2026), but only when the value they report is net of refunds. When it isn’t, the lift evaporates.
The Temporal Signal Gap Between WooCommerce and Google Ads
Enhanced Conversions closed the identity gap — it tells Google Ads who the customer was, matching hashed email or phone from checkout against Google’s logged-in user graph. That recovers 5-15% of lost attribution on the median (SignalBridge benchmarks, 2026). Useful, but it answers a different question.
The temporal gap is about whether the conversion held. Checkout happens at time zero. The refund happens at t+7, t+21, t+45. Enhanced Conversions never enriches that later moment — it only enriches the original purchase event. So Google Ads stores the original $150 conversion and never learns that it ended up as a $0 conversion three weeks later.
Smart Bidding is an optimiser. It optimises toward the signals it receives. If the only post-click signal you send is “purchased,” it learns to find more purchasers — regardless of whether those purchases stuck.
The Symptom Most Stores Recognise But Don’t Name
The pattern is predictable. Ad spend scales, conversion volume climbs, dashboards look healthy — and refund rate quietly creeps up. Gross Target ROAS stays green; net ROAS (after returns hit the books three weeks later) softens. Customer service volume rises. Re-order rate from Smart Bidding-sourced customers is noticeably lower than from organic or direct traffic.
What’s happening is compounding: Smart Bidding has built a preference profile from whatever customers Google’s training data called “good,” and every new refunded-but-still-counted conversion reinforces that profile. Of $996 million in Google ad spend across 100 consumer brands in 2025, Maximize Conversions and Target CPA together accounted for ~43% of spend, while value-based approaches remained a minority (Launchcodex) — which means most stores are still optimising for event count, not validated revenue, making the temporal gap invisible by design.
This is the volume vs quality gap in Smart Bidding from a different angle. Ad blockers and cookie restrictions account for 30-50% missing conversions on the volume side (Optmyzee, 2026). Refund blindness compounds that loss from the quality side — the signals that do arrive are partially fiction.
What Google Ads Actually Needs: Offline Conversion Adjustments
Google Ads has a mechanism built for exactly this: offline conversion adjustments. The Google Ads API lets you retroactively modify or remove a previously reported conversion using the same gclid that identified the original click. You can reduce the value when a customer partially refunds, or retract the conversion entirely when an order is cancelled or charged back.
The acceptance window is 55 days from the original conversion. The signal has to arrive within that window or the adjustment is rejected. Once accepted, Smart Bidding incorporates the correction into its learning within a few days — and Google recommends a minimum 4-week ramp-up for Target ROAS anyway (Google Ads Help), so starting the refund feedback loop early matters.
The mechanism isn’t the problem. The problem is who has to build it.
Three Fixes, In Order of Effort
Fix 1: Flat value-based bidding on AOV (doesn’t help)
The easiest change is switching from Maximize Conversions to Maximize Conversion Value using a flat average order value. This is the “value-based bidding for beginners” option most agencies suggest first. It matters for margin weighting across product categories, but does absolutely nothing about refunds. You’re still reporting gross checkout value, still never correcting it. Refund rate behaviour is unchanged.
Fix 2: Server-side offline conversion adjustments via the Google Ads API (correct, but hard)
This is the textbook fix. Capture the gclid on landing, persist it against the WooCommerce order, listen for woocommerce_order_refunded and woocommerce_order_status_cancelled, and call the Google Ads API’s offline conversion adjustment endpoint with the gclid and adjusted value. It works. It’s what Google Ads official documentation recommends.
But the footprint is real: OAuth 2.0 against a Google Ads developer token, credential rotation, retry logic, rate-limiting against a Google API, logging for reconciliation. It assumes server-side GTM or custom cloud functions, which most SMB WooCommerce stores don’t have.
It’s also upstream of a second problem: if you’re issuing refunds outside WooCommerce — directly in the Stripe or PayPal dashboard — the woocommerce_order_refunded hook never fires at all, which we unpacked in refunds issued inside WooCommerce break GA4 in one way. Any offline-adjustment architecture needs to handle both refund paths to be complete.
Fix 3: Route refund hooks directly to Google Ads via a server-side pipeline
Transmute Engine™ is a first-party Node.js server that runs on your subdomain — for example, data.yourstore.com. The inPIPE WordPress plugin captures not just purchase events but the full WooCommerce hook surface, including woocommerce_order_refunded and woocommerce_order_status_cancelled, and routes them through the same pipeline that handled the original checkout. Sending an offline conversion adjustment to Google Ads is native to the architecture, not a bolt-on — and because the server persists the gclid against the order at checkout, the adjustment has the identifier it needs when the refund fires.
Your Diagnostic Check (No Changes Required)
Before deciding whether to fix this, measure it. Pull 90 days of WooCommerce data and segment by utm_source:
- Refund rate on Smart Bidding-sourced traffic (Google / cpc) over the window.
- Refund rate on organic and direct traffic over the same window.
- The ratio. If Smart Bidding’s refund rate runs materially higher than organic — 1.5× or more is common — the training signal is already distorted. The gap widens month on month.
Then open your Google Ads conversion action. If “Include in Conversions” is Yes and no offline adjustment source is configured, that is the gap, confirmed on Google’s side.
Key Takeaways
- Smart Bidding sees the checkout, not the refund. Without offline conversion adjustments, post-purchase corrections never reach the algorithm.
- Enhanced Conversions fixes identity, not truth. It recovers 5-15% of lost attribution but doesn’t correct conversions that later reverse.
- The acceptance window is 55 days. Refund adjustments outside that window are rejected by the Google Ads API.
- Flat value-based bidding on AOV does not help. You’re still reporting gross checkout value; Target ROAS optimises toward it either way.
- The diagnostic is a ratio, not a number. Compare refund rate on paid Google traffic against organic — the gap is the size of the distortion.
Frequently Asked Questions
Because Smart Bidding only receives the purchase conversion at checkout. WooCommerce refunds issued days or weeks later never reach Google Ads unless you actively send an offline conversion adjustment back using the original gclid. The algorithm treats every refunded checkout as a successful conversion, so it keeps finding lookalike customers with the same return-prone behaviour.
Use Google Ads offline conversion adjustments. Whenever the woocommerce_order_refunded or woocommerce_order_status_cancelled hook fires, your server sends an adjustment to the Google Ads API referencing the original gclid — either retracting the conversion entirely or reducing its value. That requires capturing the gclid at click time, storing it against the order, and having a server-side component that can call the Google Ads API when the hook fires.
No. Enhanced Conversions solves the identity gap — matching a checkout to a logged-in Google user via hashed email and phone. That recovers 5-15% of lost attribution on the median. It does nothing about the temporal gap — the 7 to 45 day window between checkout and refund — because Enhanced Conversions only enriches the original purchase event. Post-purchase corrections require offline conversion adjustments.
Usually because the value being reported is gross revenue at checkout, not net of refunds. Target ROAS optimises toward whatever value you tell it matters, so if you send $100 for an order that gets refunded two weeks later, the algorithm learns that customer profile is worth $100 and bids accordingly. Until refund adjustments reach Google Ads, value-based bidding is optimising toward a number that doesn’t land in the bank.
Adjustments must be uploaded within 55 days of the original conversion for Google to accept them, and Smart Bidding incorporates the corrected signal into its learning within a few days of the adjustment reaching the account. Google recommends a 4-week ramp-up period or three complete conversion cycles before activating Target ROAS, so the earlier refund signals start flowing, the faster the algorithm converges on profitable customers.
Run the diagnostic first. The ratio will tell you whether this is your problem before you change any configuration. See how Transmute Engine routes refund hooks into Google Ads.