Last week I attended Inside Google Ads Live in Toronto, hosted by Google Ads coach Jyll Saskin Gales. Across a full day of sessions from agency leads and in-house marketers, five tips and lessons came up again and again. They apply whether you run your own account or pay someone to run it, and each one comes back to the same point: Google’s automation is only as good as the direction you give it.
1. Control with automation, not instead of it
Jyll opened the event with a reframe. The old debate of control versus automation is over, and the useful question is how to steer the automation you already have. Smart Bidding has won the bidding war, so the system now makes most bid, match and ad decisions for you. It still follows your inputs almost literally.
In one example from her session, an advertisers’ consent banner change broke conversion tracking for a on April 1 (which I’ve seen happen too!). With a Target CPA bid strategy, the campaign saw almost no conversions and lowered its bids. The CPA on one campaign jumped from $77 to $563 signalling the issue in question. In another example, an advertiser switched a branded campaign from maximizing conversion value to a target ROAS and lifted its return from 3.1x to 7.4x. The automation was the same in both cases, but the instruction to tell Google was different.
Our job as Google Ad practitioners is to choose the goal, feed the system clean data and check the results. And also, high CPCs are not always the enemy. Low quality traffic is.
2. Chase qualified leads, not cheap ones
Cost per lead (CPL) is easy to report and easy to misread. Several speakers showed how the cheapest source of leads can be the most expensive source of customers. In one example presented, the lead type with the lowest CPL cost twice as much per qualified consultation as a pricier one. If we tell Google to optimize for form fills, it will happily buy more of those which may be the wrong signal entirely.
The fix for wasted lead gen spend is to implement a sales feedback loop according to Aldwin Keekon. Ask your sales team, or client, how the leads are going, and reconcile that with Google Ads campaign data weekly or bi-weekly against CRM pipeline stages (e.g. MQL, SQL, Opportunity, Closed-Won).
The next step is to send qualified sales stages (and hopefully revenue values) back to Google Ads through offline conversion imports to ensure Googles’ smart bidding models can optimize for qualified opportunities, rather than chasing lead volume. This feedback can also tell you what to change outside the ads. When sales says leads are ghosting them, it often means the ad earned attention but nothing useful followed, and a nurture sequence can do more than a new headline or optimization in-platform.
Another lead gen lesson sits on the creative side. Ads and landing pages should make it obvious who the offer is for, because clarity beats cleverness when you are trying to attract qualified buyers.
3. Give campaigns time, and measure over the right attribution window
I say this to clients all the time: you need to give campaigns time to learn.
Long sales cycles make short reports misleading. Jyll gave us an example of a B2B advertiser that spent $7,000 in one month on broad match search terms that looked irrelevant and produced a single conversion. Zoomed out to a longer period, the same campaign spent $22,000 for 11 conversions, which landed exactly on its $2,000 target cost per acquisition.
A software team testing video ads against Demand Gen saw a similar problem in the beginning. Its customers took weeks to reach the conversion that mattered, so the team extended the click window from 30 to 60 days, added strategic videos using problem/solution and call to actions of various durations, and left the campaign alone for five months. Cost per acquisition fell by 90%.
For practitioners, two reminders were emphasized. Low volume B2B accounts can take 6 to 18 months to reach statistical significance, so directional signals and a handful of session recordings from qualified visitors often teach you more than waiting for a perfect test. Secondly, you need to look beyond the Google Ads report when you judge a new campaign. The growth it creates can show up as direct or organic traffic in GA4, where Google Ads gets no credit.
If your sales cycle runs for weeks or months, judge your campaigns on a window that matches it.
4. Fix the foundations before chasing new features
New features work best on a solid base, and three foundations came up in nearly every session.
-
- Tracking. Audits can find overlapping conversion actions that count the same sale two or three times. One account gave every conversion a flat default value of $110, so a $50 purchase and a much larger order looked identical to the bidding system. If your tracking is broken, fix it so the algorithm doesn’t learn from bad data.
-
- Product data. For e-commerce, the shopping feed tells Google what you sell. Product titles are cut off at roughly 55 characters, so size, color and material belong at the front. One retailer rebuilt its feed by hand over three weeks and booked more business in a single month than in the previous three combined.
-
- Campaign structure. Give each campaign one job. An example that given was a domestic dealership that split one Performance Max campaign into separate new and used inventory campaigns, each with its own budget and conversion history. Over the next quarter its spend stayed flat at about $29,400, but website forms rose 93%, and cost per form fell from $143 to $74.
5. Use AI with guardrails and keep strategy human
AI is now built into the platform, and the results can be strong when the setup is right.
A healthcare recruitment advertiser tested AI Max only after two years of clean conversion data and clean search term reports. Over about 12 weeks, AI Max drove 60% of conversions and a 35% higher conversion rate, and cost per conversion on its expanded keywords fell 36%. The same test showed a risk with this new setting. Google’s suggested ad copy would have broken brand rules, so the team had to switch on text customization controls and review search terms twice a week instead of once per week.
For everyday work, the same logic applies.
To determine what to automate with AI, treat the model like a junior hire. As Dana DiTomaso puts it: “If you can’t explain it to a junior, you can’t explain it to an AI model.”
She also suggested:
-
- Only automate tasks that are performed at least 3x per month. If it’s a one-off or quarterly task, it’s not worth the effort to build and refine the prompt reliably.
-
- Don’t automate high-stakes levers like budgets, targets, account pauses, or client-facing communications.
-
- AI should be used to provide alerts, recommendations, but the human should still be “in the loop” to review all outputs and key decision-making.
The final reminder is that judgment is the most important skill in managing Google Ads, and no setting will rescue a flawed strategy.
What this means for your Google Ads account
Google’s automation is powerful and oftentimes obedient, so your results depend on the direction you give it. Before your next budget increase, check what the account is optimizing for, whether your tracking reflects real revenue, and whether your reporting window fits your sales cycle.
If you need help tracking the right results for your business, and hiring for human strategy over AI, let’s talk.
