StopChasingROAS:HowtoWrestleBackControlofYourGoogleAds

The 31% Hallucination

A finance director at a £40M home-and-garden DTC sat in a Monday board meeting last spring, looked at his marketing dashboard, and told the room: "This Google Ads account is the most profitable thing we do."

He was wrong. The platform showed a 6.2× blended ROAS. The bank balance showed a deficit. The gap between what the dashboard said and what the P&L said was 31% - a hallucination the entire leadership team had been making decisions on for nine months.

This is the gap most UK ecommerce brands live inside without realising. Google Ads tells one story. Contribution margin tells another. When you let the platform's scorecard set your strategy, you optimise for what Google measures - which is not what your business needs to survive.

This piece is about wrestling back control. Not by abandoning Google Ads - that's not the answer for ecommerce brands spending £15k+/month - but by running it on your scorecard, not Google's. The framework is BOI® (Bid On Intent). The operational discipline is what we call the Five Rounds: a weekly rhythm that turns commercial intent into the signal Smart Bidding actually optimises against.

If your dashboard is healthy and your bank balance isn't, this is the playbook.

Why Google's scorecard isn't your scorecard

The Google Ads UI is built around what Google needs to measure to run an auction. The metrics it surfaces - ROAS, conversions, impression share, Quality Score - are designed to keep its auction densely populated and its inventory liquid. They're useful inputs. They aren't your business outcomes.

The mismatch isn't malicious. Google isn't trying to make you unprofitable. It's that the platform can only optimise against the signal you feed it - and if the signal is gross revenue (which is the default), the bidder will scale spend on whichever SKUs convert at the highest reported revenue. For ecommerce brands with variable product margins, that's usually your lowest-margin lines.

Revenue goes up. Contribution margin goes down. The account looks healthier on the surface and worse underneath. That's the 31% Hallucination, structurally explained.

The three myths that keep brands stuck

Myth 1: Blended ROAS is the right top-line metric

Blended ROAS averages your high-performing champions with your loss-making clearance lines. A 5× blended ROAS can be made up of 12× on flagship SKUs and 1.8× on clearance - or it can be a flat 5× across the catalogue. The two scenarios produce wildly different P&L outcomes and require completely different management interventions.

You don't manage a sales floor by averaging your top earner with your worst performer and calling that the team's productivity. You don't manage a Google Ads account that way either. The right top-line metric is POAS at SKU level (here's the full read on why), with the blended view as a sanity check, not a target.

Myth 2: Every SKU should be optimised the same way

Treating a 60%-margin hero product the same as a 12%-margin clearance line is not just inefficient - it's a guaranteed contribution margin drag. Smart Bidding configured against blended targets will allocate budget toward the SKUs that hit the target cheapest, which structurally favours the lowest-margin products in your catalogue.

The fix isn't more sophisticated automation. It's per-SKU commercial intent fed into the bidder. Every SKU has a job. Custom labels in the product feed are how you tell Google what that job is.

Myth 3: Smart Bidding works on auto-pilot

Smart Bidding does work - but only when it's fed accurate, current, commercial signal. The "set it and forget it" framing isn't wrong because automation is bad. It's wrong because your commercial intent isn't static. Stock levels change. Margins compress. New SKUs launch. Returns rates shift. Suppliers raise prices.

If the data feeding Smart Bidding is six months stale, the algorithm is optimising against a business that no longer exists. The discipline isn't to fight automation - it's to make sure automation has the right signal every single week.

What BOI® actually is

BOI® - Bid On Intent - is JudeLuxe's framework for running Google Ads against commercial intent rather than platform metrics. Every SKU in your catalogue is assigned one of five commercial jobs (Scale, Profit, Protect, Recovery, or Gateway) calculated from margin, stock, cash impact, and customer-acquisition role. The bidder is fed the SKU's current job as the signal it optimises against - not blended ROAS.

The full methodology is on the BOI® page. The job taxonomy is covered in the SKU jobs framework article. What this piece adds is the operational rhythm: how you maintain BOI® week to week in a live account.

Because here's the thing nobody talks about: assigning SKU jobs once is easy. Maintaining accurate jobs week-on-week, as your business reality shifts, is the discipline that separates frameworks that work from frameworks that decorate slides.

The discipline is what we call the Five Rounds.

The Five Rounds: your weekly operating rhythm for BOI®

Every working day, one round. No skipping. Friday's decisions inform next Monday's recon.

The five jobs of an ad SKU

For completeness - here's the five-job taxonomy that BOI® assigns to every SKU. Full version with worked examples is on the SKU Jobs Framework page.

• ScaleHealthy margin, full stock, auction headroom. Bid aggressively for growth.
• ProfitDefault state. Bid for maximum contribution margin per pound spent.
• ProtectContested auction. Defend position without chasing volume.
• RecoveryStock low, margin compressed, returns climbing. Pull back, protect cash.
• GatewayCustomer-acquisition product. Accept lower per-unit margin for new-customer LTV.

The taxonomy is the structure. The Five Rounds is the operational discipline that keeps the taxonomy current.

The product feed is where the real power lives

Most marketers spend their time in the Google Ads UI. The UI is where Google wants you to play - it surfaces the levers Google wants you to pull.

The product feed is where the real control lives. It's the only file Google doesn't write for you. Custom labels in the feed are how you communicate SKU-level commercial intent to the bidder. Title structure is how you control which queries you match. Attribute completeness is how you avoid impression suppression. Image quality is how you compete for click-through.

Without the feed, BOI® is just a vibe.

For ecommerce brands on Shopify, the default Google Shopping app feed covers 60-70% of attribute completeness. For Magento brands, the default export typically covers 40-55%. For high-SKU accounts, the gap widens further. Feed engineering is the dominant work of catalogue-scale PPC (see our high-SKU positioning for the full read).

Feed engineering rule: the SKU's current BOI® job lives in custom_label_0. Anything more sophisticated than that lives in custom_label_1 through custom_label_4 with explicit documentation. Never trust the feed to "look after itself."

The paradox: dashboard worse, business better

When you implement BOI® properly, the first 90 days produce a result that scares most teams.

The case study (£40M home-and-garden DTC, Q2 2025):

• Blended ROAS moved from 6.2× → 5.4× (looks worse)
• Bottom-tier SKU segment (running at 1.8× ROAS, structurally loss-making) was excluded from active bidding entirely
• Top-tier SKU segment (now running at 11.0×) received the reallocated spend
• Contribution margin rose 31% quarter on quarter
• Cash position improved by £180k in the same window

The blended ROAS slipped because the loss-making SKUs being cut had been propping up the vanity metric. They were converting cheaply at high reported revenue, which inflated the blended figure while quietly destroying contribution margin. Cutting them was visible in the dashboard before the lift was visible in the bank account - which is exactly the experience that makes nervous teams reverse the work in week three.

Excluding "conversions" that cost more than they earn isn't a failure. It's a move of extreme strength.

This is the paradox of running on commercial intent rather than platform metrics: dashboards can look worse before they look better. The brands that survive the 90-day window are the ones who measure their work in contribution margin, not blended ROAS.

Three things to verify in your own account this week

If this article describes the gap between your dashboard and your P&L, here's where to start. Not a full audit - just three checks you can run yourself before deciding whether you need outside help.

If any of those three checks fail, the work to fix is structural - not a tactic, not a campaign optimisation, but a foundational rebuild. We do this as a free Profit Audit for qualifying brands. Five to seven days, written PDF report, quantified findings in pounds. Yours to keep whether you work with us or not.

Stop answering to Google's scorecard

Every Monday morning, ask one question:

"Is this working for the business, or am I just working for the platform?"

If the answer is uncomfortable, the problem isn't your effort. It's the scorecard you're keeping. Google's scorecard rewards spend density. Your P&L's scorecard rewards contribution margin. Until your bidder is fed your scorecard's signal, you'll keep producing the 31% Hallucination.

Fix the model. Then scale it.

Frequently asked questions

— Chris Avery, Co-founder, JudeLuxe