The Playbook That Used to Work
If you built your Google Ads knowledge between 2014 and 2020, you probably learned some version of this:
- One keyword per ad group (SKAG structure) for maximum control
- Exact match keywords only to eliminate wasted spend
- Manual CPC bidding to control costs
- Separate campaigns for every product, service, or location
- Pause anything with a low Quality Score
This wasn't wrong. For the platform that existed at the time, it was genuinely good practice. The auction was simpler. Match types were stricter. Bidding was more predictable. Human judgment applied at the keyword level produced better results than trusting Google's automation, which was less developed and less reliable.
That platform no longer exists.
What Actually Changed
Exact match stopped meaning exact match. Google now matches exact match keywords to "close variants": different word orders, implied words, paraphrases, and semantically similar queries. An exact match keyword no longer means your ad only shows for that precise search. This has been gradually true since 2017 and is now firmly established. The implication: running a long list of exact match keywords to control coverage is less effective than it used to be, because match type alone no longer determines what queries trigger your ad.
Manual CPC is structurally disadvantaged. Google's smart bidding algorithms (Maximize Conversions, Target CPA, Target ROAS) have access to real-time signals at auction time that manual bidding cannot replicate: device, location, time of day, browsing behavior, search context, and dozens of other factors. A manual bid applies uniformly to every auction. A smart bid adjusts for every individual auction. For accounts with enough conversion data to train the algorithm, smart bidding consistently outperforms manual CPC, and the data threshold to get there has dropped.
Fragmented campaign structure starves the algorithm. When you run nine separate campaigns each with a small daily budget, each campaign runs on too few conversions per month to allow smart bidding to function properly. The general threshold is 30 to 50 conversions per month per campaign. Below that, the algorithm is guessing. This means a highly structured account with many campaigns and precise segmentation can actually underperform a consolidated account with fewer campaigns and more signal per campaign.
Broad match with smart bidding is no longer reckless. Broad match used to be a spend fire hose. It sent your ads to vaguely related queries you'd never approve manually. Combined with smart bidding and a strong conversion signal, it now functions differently: the algorithm uses real-time context to determine whether a broad match query is likely to convert, and bids accordingly. For accounts with solid offline conversion data, broad match with Target CPA or Target ROAS can find converting queries that exact match never would have captured.
The keyword is no longer the unit of strategy. The shift is from "which keywords should I bid on?" to "what signals am I feeding the algorithm?" Keywords are now one input among many. Audience signals, landing page quality, creative assets, conversion data quality, and offline conversion imports all shape what the algorithm does with your budget. An account with mediocre keywords and excellent conversion signals often outperforms an account with perfect keyword structure and a shallow conversion signal.
What the Updated Playbook Looks Like
Fewer campaigns, more signal per campaign. Consolidate where possible. Rather than separate campaigns for every product variation, build consolidated campaigns that can accumulate enough conversions for smart bidding to function. Use ad groups and asset groups to maintain relevance within a consolidated structure.
Offline conversions as the primary optimization target. If you're running lead generation, Google needs to know which form fills became customers, not just which ones fired a conversion tag. Importing qualified leads or closed sales as offline conversions gives the algorithm the signal it needs to optimize toward actual business outcomes rather than form submission volume.
Negative keywords as strategic filters, not a primary control mechanism. You still need negatives. But the goal of a negative keyword list is no longer to fence in exact match coverage. It's to filter out clear mismatches while allowing the algorithm latitude to find converting queries you didn't anticipate.
Quality Score as a diagnostic, not a goal. If Quality Score is low and conversions are strong, ignore it. If Quality Score is low and conversions are weak, use it as a clue that ad relevance or landing page alignment may be the problem. Then fix those things because they affect conversion rate, not because a higher number on a score is the goal.
Creative and assets as a lever. With RSAs and PMax, the creative you put into the system is one of the primary variables you control. Headline and description diversity, asset quality, and landing page relevance have more impact than they did when static ads were the default format.
What to Keep from the Old Playbook
Not everything changed. Some fundamentals remain true:
Conversion tracking that accurately reflects business value still determines everything. If your conversion setup is broken or misleading, no amount of updated strategy will fix it. The algorithm learns from your data. Garbage in, garbage out.
Negative keywords still matter. The platforms have more latitude than before, but they still benefit from explicit exclusions for irrelevant queries, competitor brand terms you don't want to spend on, and clearly misaligned intent.
Landing page quality and message match still determine whether clicks become customers. The auction has gotten smarter, but the page hasn't changed. If someone clicks and finds something that doesn't match what they expected, they leave.
The core question (what is a customer worth and what can I afford to pay to acquire one) is still the strategic foundation everything else rests on.