Broad match and Smart Bidding: the default architecture that works in 2026
By Tom Goodwin, Founder of GAMEPLAN.
The default Google Ads architecture that works in 2026 is broad match keywords feeding Smart Bidding, governed by clean conversion data and a target that reflects your margin. Broad match supplies the query breadth the algorithm needs to find demand you would never have keyword-listed yourself. Smart Bidding then bids per auction on predicted value, so the breadth stops being a leak and becomes coverage. The catch, and it is the whole game, is that this only works on top of reliable measurement and a disciplined negative keyword list. Bolt broad match onto bad tracking and you have built a machine for spending money quickly.
I have run this architecture across £20m+ of managed media, including the accounts that earned my team Google Premier Partner status in February 2024 and one that grew 60% year on year. The growth did not come from broad match being magic. It came from feeding a good algorithm good data and getting out of its way.
Why does broad match work now when it failed before?
Because the bidding underneath it changed. For years broad match ran on manual bids: it matched your keyword to a wide set of queries, then bid the same amount whether the searcher was ready to buy or idly browsing. That is why it earned its reputation as a budget incinerator. Smart Bidding sets a different bid for every auction based on the predicted value of that specific query, device, time and audience. So the same breadth that used to waste money now finds value, because the bid scales down on weak intent and up on strong intent automatically.
The two technologies are designed for each other. Broad match without Smart Bidding is reckless. Exact match without breadth starves the algorithm of the signals it needs. Together, fed properly, they outperform the legacy exact-match sprawl most accounts still carry.
What is the legacy architecture you should move away from?
The old best practice was the single-keyword ad group: hundreds or thousands of tightly themed ad groups, each on exact and phrase match, each with bespoke ads. It made sense when humans set bids, because tight structure let you control spend precisely. It makes no sense now.
| Dimension | Legacy SKAG architecture | 2026 default architecture |
|---|---|---|
| Match type | Exact and phrase, tightly controlled | Broad, with strong negatives |
| Bidding | Manual or basic automated | Smart Bidding on value or CPA |
| Structure | Hundreds of single-keyword ad groups | Few, well-fed campaigns |
| Data per campaign | Thin, fragmented | Consolidated, enough to learn |
| Maintenance | Constant manual bid edits | Target setting and negatives |
| Failure mode | Misses demand you did not list | Wastes spend without negatives |
The legacy approach fragments your conversion data across so many segments that Smart Bidding can never learn. Consolidation is the unlock: fewer campaigns, each with enough conversions, each bidding on clean value.
What do you need in place before you switch?
Four things, and the architecture amplifies the absence of any of them.
- Reliable conversion tracking. This is non-negotiable. Broad match plus Smart Bidding bids on predicted conversion value, so if your conversions are wrong, every bid is wrong, at scale, fast.
- Enough conversions per campaign. Aim for 30+ a month per campaign so the algorithm has signal to learn from. If you cannot reach that, consolidate campaigns until you can.
- A margin-based target. Set target ROAS or target CPA against real economics, not a number that feels safe. The algorithm optimises to whatever you tell it, so tell it the truth about your margins.
- A maintained negative keyword list. Broad match will find queries you do not want. Negatives are how you teach it the edges of your business. This is the one job that stays manual.
If those four are not in place, fix them before you touch match types. I always rebuild measurement before I touch bidding, because the order matters.
How do you set the right target?
Start from contribution margin, not gross revenue. If a sale carries 40% margin and you want to keep half of that, your acceptable cost of acquisition is clear, and your target ROAS follows from it. Set the target, then leave it alone long enough for Smart Bidding to learn, usually two to three weeks. The most common self-inflicted wound I see is owners changing targets every few days, resetting the learning each time, then blaming the platform for volatility they created.
When you do adjust, move targets in small steps, 10 to 15% at a time, and let each change settle. Smart Bidding rewards patience and punishes fidgeting.
How do you keep broad match from wasting spend?
Negatives, monitoring, and structure. Three habits:
- Review the search terms report weekly at first, then fortnightly. Add negatives for irrelevant queries and intent mismatches. This is the maintenance the architecture requires in exchange for its reach.
- Keep brand separate. Never let broad match prospecting campaigns absorb your brand traffic, or your blended ROAS will look healthy while your prospecting quietly fails.
- Watch the query mix, not just the metrics. If the report fills with informational queries, the campaign is prospecting too early in the journey, and no bid strategy fixes a demand-stage mismatch.
Done well, the weekly negative review takes 20 minutes and pays for itself many times over.
When is this architecture the wrong choice?
It struggles in three situations. Very low volume accounts that cannot feed any campaign enough conversions are better served by tighter match types until volume grows. Businesses with weak or untrustworthy conversion tracking should fix measurement first, full stop. And highly regulated sectors, such as private healthcare, need extra care because broad match can surface queries that trigger policy issues, so negatives and monitoring matter even more there.
For most businesses spending real money, though, broad match feeding Smart Bidding on clean data is now the default that works, and the legacy alternative is a maintenance burden that performs worse.
If your account is still running legacy exact-match sprawl and underperforming, the fix is usually consolidation and clean measurement, not more keywords. I rebuild paid media architecture around the data and targets that actually move the business. Tell me your current structure and the number you are trying to hit. Start here: /work-with-me/paid-media-strategy.