Google keeps saying it: "move your keywords to broad with Smart Bidding and you will get more conversions". And it has the data to say so — on average. The problem is that your account is not the average. This lesson teaches you how to check it in your own account with a clean experiment, and how to tell when broad is a real gain and when it is Google buying reach with your money.
Why modern broad is different
Today's broad does not match on words: it uses auction signals (the full query, the intent, the user's history, the landing page, the context) to decide whether your ad fits. And with Smart Bidding, the algorithm bids differently for each term: high on the ones it estimates will convert, next to nothing on the rest. That pairing is what makes it work: broad opens the door and Smart Bidding decides whose click to pay for. Without Smart Bidding, broad is the same old broad: reach with no judgement.
The consequence: broad needs data. In a campaign with few conversions, the algorithm has nothing to learn from and broad buys noise. Minimum requirement: a campaign stable on its target CPA/ROAS with dozens of conversions a month.
The clean experiment
Do not switch the whole campaign over "to see what happens". Google Ads has custom experiments: a copy of the campaign with the change, traffic split, comparison with statistical significance.
- Candidate campaign: Smart Bidding with a target, 30+ conversions a month, negatives kept up to date, clean search terms.
- Experiment: a copy of the campaign; in the copy, move only one ad group's keywords to broad (or a few ad groups with clear intent), not the whole campaign. Keep the same target.
- 50/50 traffic split, 4-6 weeks minimum (or two complete conversion cycles).
- Negative seatbelt on both versions: the universal list, the roots from your mining and, if you have one, the white list. Broad with no negatives is a test rigged against itself.
- Do not touch anything else during the experiment.
What to measure
| Metric | What you are looking for |
|---|---|
| Total conversions and CPA/ROAS | More conversions at the same cost, or the same conversions more cheaply? That is the question |
| New search terms with conversions ("incremental queries") | Broad is finding demand you did not have: the argument in favour |
| New terms with no conversions and their cost | The noise you are buying: the argument against |
| Conversion rate | If it drops a lot, broad is bringing in less decided people |
| Lost IS (budget) | If broad eats the budget, the comparison is distorted |
Google flags significance in the interface: if it is not significant, there is no result, there is noise. Do not declare a winner before that.
When it wins and when it does not
It tends to win when: the business has demand you cannot express in keywords (broad services, varied vocabulary), the measurement is rich (values, enhanced conversions), and the budget is not capped.
It tends to lose when: the product is very specific (exact already covers it), the account has few conversions, the landing page converts badly (broad brings in colder people and the landing does not convince them), or the budget is capped (broad spends it on the new stuff at the expense of what already worked).
And one frequent result: it wins on volume, it loses on CPA. At that point the decision is a business one: do you want more customers at a slightly higher price? Sometimes you do.
The account pattern that integrates it
If broad wins, do not replace everything. A healthy pattern:
- Exact + phrase for the core (control, low CPC, brand).
- Broad in its own ad group by intent, with Smart Bidding, with the negative seatbelt and with a weekly review of terms: broad is the one that generates the most negative-keyword work.
- The white list (previous lesson) is its natural partner: broad to reach, white list to stay on topic.
💡 Ninja trick: broad multiplies the new search terms every day, and that multiplies the negative-keyword workload. It is the reason why anyone using broad seriously ends up automating: SQONS cleans up by performance and relevance every night, Negative Keywords keeps the topic in line with the white list, and Smart Bidding (SBNS) watches that the CPA of the broad campaign does not drift away from the business target. Broad, yes — with the machinery in place.
What you should remember
- Modern broad works because of Smart Bidding's signals: without it, do not use it.
- Test it with a 50/50 experiment, in one ad group, 4-6 weeks, with negatives on both sides and nothing else touched.
- Measure total conversions and CPA/ROAS, incremental queries with conversions and the noise you buy; respect significance.
- If it wins: its own ad group, negative seatbelt and a weekly review.