Smart Bidding (AI Bidding)

Smart Bidding is the group of automated bid strategies in Google Ads that use machine learning to set a separate bid for every auction, aiming for the most conversions or conversion value within your budget or target. Instead of one fixed bid per keyword, it adjusts the bid for each search using signals about the person, device, place and time.

• Target CPA: Aims for an average cost per conversion that you set.
• Target ROAS: Aims for an average return on ad spend, meaning revenue divided by cost.
• Maximize conversions: Gets as many conversions as possible from the budget.
• Maximize value: Maximize conversion value gets the most revenue from the budget, with an optional ROAS target.
• Fuel: It learns from the conversions you track, so bad tracking means bad bidding.

This lesson follows one example: a D2C saree brand from Surat that sells silk, georgette and cotton sarees online, with prices from about Rs 1,500 to Rs 25,000. It runs Search, Shopping and Performance Max campaigns and wants profitable revenue, not just more orders.

Key Characteristics of Smart Bidding

• Auction time: A new bid is set for every single auction, which no person could do by hand.
• Signals: It uses many signals, such as device, location, time of day, language, browser, audience lists and the search itself.
• Goals: It optimises towards the conversion goals marked as primary, so those goals must be the right ones.
• Targets: Targets are averages, so some clicks will cost more and some less than the target.
• Learning: Each new strategy or big change needs time and conversions before results settle.

How Smart Bidding Works

1. Choose the goal: The saree brand tracks purchases with order value, set up in Google Ads conversion tracking.
2. Choose the strategy: Because order values vary widely, it chooses Maximize conversion value.
3. Learn first: For the first weeks it runs without a target so the system can learn what a normal return looks like.
4. Add a target: Once results are stable, the brand sets a target ROAS close to what the campaign already achieves, not a big jump.
5. Bid per auction: For each search, the system predicts the chance of a sale and its value, then sets a bid to meet the target.
6. Feed back: Every recorded sale, and its value, updates the model, which is why accurate values matter so much.

Example: Smart Bidding for a Surat Saree Brand

• Target: The brand first works out its break-even ROAS from product margins, returns and shipping, then sets a target above it.
• Launch: It starts Maximize conversion value without a target, then adds a target ROAS near the level already achieved after the results settle.
• Diwali: For a short festive sale, it uses a seasonality adjustment so bids react to the expected rise in sales from day one.
• Outage: When a checkout update breaks tracking for two days, it adds a data exclusion for those days.
• Returns: Because many returns happen after a sale is counted, the brand later uploads adjusted values, covered in enhanced conversions and offline conversion import.

Benefits of Smart Bidding

• Scale: Bids for thousands of auctions a day, each with its own signals.
• Signals: Uses information advertisers cannot see or bid on by hand.
• Efficiency: Often gets more conversions or value for the same budget once it has learned. Measure this in your own account rather than assuming it.
• Time: Frees the team from daily bid changes to focus on offers, pages and creative.

Limitations of Smart Bidding

• Needs data: Low conversion volume makes learning slow and results unstable.
• Garbage in: Wrong conversion goals or values lead it straight to the wrong outcome.
• Less transparency: You see results, not the reason for each bid.
• Patience: Frequent target or budget changes restart learning.
• Credit: It may bid heavily on people who would have bought anyway, which incrementality testing helps check.

How AI Changes Smart Bidding

What AI Automates Now

Smart Bidding is itself AI: it predicts the chance and value of a conversion for each auction and sets the bid. It also works with broad keyword match types and Performance Max to decide which searches and placements are worth paying for.

What Still Needs a Human

People still choose the goal, set a target based on real margins, send correct order values, and decide when a sale or a stock problem needs a manual adjustment. The model does not know that a silk collection is out of stock unless the feed and pages say so.

Risk to Watch

A target based on revenue rather than profit can push spend to low margin products. Watch profit, not only ROAS, and use Google Ads scripts or alerts to catch sudden spend spikes.

Do It with AI

Use this prompt to pick a strategy and a starting target. It works in ChatGPT, Claude or Gemini.

1. Export the last 30 days of cost, conversions and value per campaign.
2. Run the prompt and check the break-even calculation yourself.
3. Switch one campaign first and leave it alone through the learning period.
4. Compare profit, not only ROAS, before switching the rest.

Check Before You Use It

• Facts: Recalculate margins and ROAS by hand; AI tools can make arithmetic mistakes.
• Brand fit: Make sure the strategy supports the products the brand most wants to sell, not just the cheapest ones.
• Compliance: Send only allowed data with conversions, and never pass customer names or phone numbers in plain text.