The Human Judgment Layer
What The Algorithm Doesn't Know
Google Smart Bidding is optimising your account right now. It's using click data, auction signals, device types, and time of day. It's sophisticated. It's fast. And it's missing the information that actually determines whether you make money.
The algorithm sees revenue. It does not see profit. That gap is where most ecommerce brands lose money without realising it.
8 Things Smart Bidding Cannot See
Each of these information gaps creates a systematic bias in how your budget is allocated. The algorithm isn't broken - it's blind.
Returns Rates by SKU
What Google Sees
A sale worth £89.
What Actually Happened
The customer returned it 11 days later. Net revenue: £0. You paid £12 to acquire nothing.
Commercial impact: Fashion brands with 30%+ return rates can show 4x ROAS while losing money on every third transaction. Smart Bidding optimises to generate more of these 'phantom sales.'
Cost of Goods Sold (COGS)
What Google Sees
Revenue of £120 from a conversion.
What Actually Happened
COGS was £78. Fulfilment was £9. Payment processing was £4. Your real margin was £29 - before ad spend.
Commercial impact: A product with 5x ROAS but 15% margin is barely breaking even. A product with 2.5x ROAS and 60% margin is highly profitable. The algorithm treats them identically.
Stock Levels & Availability
What Google Sees
A high-converting product worth pushing harder.
What Actually Happened
You have 12 units left. The next shipment is 6 weeks away. Every sale brings you closer to a stockout and disappointed customers.
Commercial impact: Aggressive bidding on low-stock items creates demand you cannot fulfil. Worse, it trains customers to expect availability you cannot maintain.
Supplier Price Increases
What Google Sees
Historical conversion data suggesting this product converts well at £45 CPA.
What Actually Happened
Your supplier raised prices 12% last month. The margin that made £45 CPA profitable no longer exists.
Commercial impact: The algorithm continues bidding based on outdated economics. By the time performance 'looks wrong' in the platform, you have burned through weeks of unprofitable spend.
Cash Flow Constraints
What Google Sees
An opportunity to scale - more budget would mean more conversions.
What Actually Happened
You are 14 days from a VAT payment. Your supplier invoice is due in 7. The cash needed for ad spend is committed elsewhere.
Commercial impact: Smart Bidding has no concept of working capital. It will recommend increasing spend at exactly the moment your business cannot afford it.
Fulfilment Costs & Thresholds
What Google Sees
Two orders, both worth £50.
What Actually Happened
Order A is a single lightweight item shipped for £3.50. Order B is three heavy items shipped for £14.80. Same revenue, vastly different profit.
Commercial impact: Without fulfilment cost data, the algorithm cannot distinguish between a profitable order and a loss-making one. It optimises for revenue volume, not commercial viability.
Lifetime Value & Repeat Purchase Rates
What Google Sees
A first-order worth £35 - low value, reduce bids.
What Actually Happened
That customer reorders every 6 weeks. Their 12-month LTV is £420. The £35 first order was a gateway to a high-value relationship.
Commercial impact: Subscription and consumable brands are systematically under-bid by algorithms that only see the first transaction. Your best customers look like your worst ones.
Discount & Promotion Context
What Google Sees
Conversion rate doubled last week - this is working.
What Actually Happened
You ran a 30% off promotion. Margins were 8% instead of 38%. The 'best week' was actually the least profitable.
Commercial impact: The algorithm learns that promotional periods are 'good' and bids accordingly. It cannot distinguish between genuine demand signals and artificially inflated conversion rates from discounting.
The Compound Effect
Any one of these blind spots is manageable. The problem is they compound. A high-return product with thin margins, low stock, and expensive fulfilment looks identical to your best-performing SKU - from the algorithm's perspective.
Smart Bidding doesn't make bad decisions. It makes uninformed decisions. It optimises perfectly for the wrong objective because it lacks the commercial context that determines whether a sale is actually worth having.
This is why ROAS can look healthy while your P&L deteriorates. The metric and the outcome have become disconnected - and the algorithm is the last to know.
The Human Judgment Layer
We don't turn off automation. We constrain it with commercial reality.
Before Smart Bidding makes a single decision, we feed it margin-adjusted conversion values. We segment SKUs by their commercial role - scale, profit, or cash recovery. We apply stock-aware bid modifiers. We set portfolio-level constraints that reflect your actual P&L targets, not platform vanity metrics.
The algorithm handles the auction. We handle the economics. That's the division of labour that actually works.
COGS & margin data
Fed into conversion values so bidding reflects profit, not revenue
Returns intelligence
SKU-level return rates adjust expected value before the algorithm bids
Stock constraints
Low-stock SKUs are de-prioritised before demand outstrips supply
Cash flow awareness
Spend pacing aligned to actual business cash position, not platform recommendations
Fulfilment economics
Heavy, bulky, or expensive-to-ship products weighted accordingly
LTV signals
Gateway products bid higher based on downstream customer value