Most underperforming Google Ads accounts don’t have a bidding problem. They have a data problem that looks like a bidding problem. Smart Bidding gets blamed for chasing the wrong users, when the real issue is that someone told it to treat a button click and a completed sale as equally valuable. The fix is not a new bid strategy — it’s primary vs secondary conversions, a framework that controls exactly what Google’s machine learning is allowed to learn from.

The setup that quietly wrecks Smart Bidding

Picture an account where every trackable action — form fills, key clicks, page views, cart adds, checkout starts, and actual purchases — flows into the same “Conversions” column, weighted identically. Reports look great. Conversion volume is high, ROAS looks efficient, and the account appears healthy on paper. Then the business checks its bank balance and none of it lines up.

Here’s a realistic illustration. A Performance Max campaign generates 4,000 clicks and reports a 62% conversion rate. Dig into the composition and roughly 90% of those “conversions” are button clicks, form interactions, and abandoned checkouts — only 37 are actual purchases. That’s a true purchase rate of 0.9%, not 62%, and a signal-to-noise ratio of roughly 9:1 working against the algorithm. On $5,400 of spend against $11,000 of revenue, effective ROAS lands at 2.04 — well short of the 4.0+ that category would typically expect.

Smart Bidding isn’t broken here. It’s doing exactly what it was told to do.

Why Smart Bidding chases the wrong signal

Smart Bidding is a pattern-matching engine, not a keyword tool. Google’s own documentation confirms it evaluates audience data, query sequences before and after the click, device, time of day, landing page behavior, and a wide range of other signals to find more users who resemble your best conversions. Every action logged as a primary conversion becomes part of that “ideal customer” definition.

The trouble starts when high-intent actions (a purchase) and low-intent micro-actions (a cart add) sit in the same primary pool. The algorithm loses contrast — it can no longer tell a buyer’s pattern from a browser’s pattern, because the account told it those two outcomes are equivalent. And because button clicks and cart adds are far easier to generate than completed purchases, the algorithm takes the path of least resistance and optimizes toward volume instead of revenue, unless a human draws a clearer line.

Primary vs secondary conversions: two settings, two jobs

The fix reframes conversion tracking as an algorithmic training decision, not a reporting preference. In Google Ads, primary and secondary status do two distinct jobs:

• Primary (optimization): Populates the “Conversions” column. Smart Bidding actively trains on these — this is the algorithm’s curriculum.
• Secondary (observation): Populates the “All conversions” column. The bidding strategy ignores these for optimization — this is your diagnostic layer.

Treat this as data architecture, not tag management. Everything you mark primary shapes what the model learns to chase. Everything secondary gets stored for human review but stays out of the training loop — at least in principle, which is worth unpacking further below.

What actually belongs in the primary pool

Only true macro goals qualify: a completed purchase, a submitted lead form, a booked consultation — actions that map directly to revenue, not actions that merely correlate with it. If you can’t draw a straight line from the action to a dollar of pipeline, it doesn’t belong in primary.

Secondary conversions can carry as many funnel-diagnostic events as you need without polluting the training data: pricing page views, cart adds, checkout starts, shipping page views, account creation. These map where users fall off without instructing the algorithm to chase low-intent traffic.

One caveat worth flagging: while Google states secondary actions are excluded from bidding, many practitioners believe the system still uses them as predictive signals of intent, even though Google hasn’t confirmed this publicly. That means secondary slots shouldn’t be filled with vanity events either — track legitimate funnel steps, not noise.

The custom-goal override most strategists miss

Custom goals bypass the primary/secondary tagging entirely. If you build a custom goal and include a secondary action inside it, that action gets used for bidding in any campaign assigned that goal — regardless of how it’s tagged at the account level. This is a frequent landmine: strategists assume “secondary means observation-only” and never notice a custom goal has quietly re-promoted a low-intent action back into the training signal. Audit every custom goal in the account before assuming the framework is intact.

There’s a similar default-state trap with imported data. When you pull conversions from Google Analytics into Google Ads, they arrive tagged secondary by default. If your real macro-goal lives in GA4, assuming the import handled optimization status will silently disconnect your bid strategy from actual revenue. Verify status manually after every import.

Edge cases that require judgment, not defaults

Phone calls are the most context-dependent case. For some businesses, calls are informational (“what time do you close?”) and belong in secondary. For others, calls are the macro-goal because they end in booked demos or sales conversations, and belong in primary. The right classification depends on listening to a sample of actual calls, not the action label itself.

Low-volume accounts that haven’t hit Smart Bidding’s data threshold — typically 30 to 50 conversions in a rolling 30 days, though this varies by strategy — still need the same discipline, but secondary-conversion quality matters even more here. With few primary events to learn from, meaningful secondary signals (a pricing page view, a demo video watched, a configurator used) can give the algorithm enough texture to start pattern-matching before volume arrives. Garbage secondary events just create false positives in that early model, so apply the same discernment you’d use for primary conversions.

When to promote a secondary action to primary

Almost never. Cart adds, checkout starts, page views, and chat opens shouldn’t graduate to primary no matter how much volume they generate — volume isn’t intent. The legitimate reasons to change status are narrow:

• Call quality data shows calls are genuinely higher (or lower) intent than assumed.
• A GA4-imported macro-goal was left in secondary by default and needs correcting.
• The business redefines what counts as a qualified lead, and the qualification step becomes the new macro-goal.

Beyond that short list, regularly promoting micro-actions to primary isn’t using the framework — it’s eroding it back into the same noisy setup that caused the problem.

What happens after you fix it

Correcting a polluted conversion setup isn’t instant relief. Moving micro-conversions back to secondary forces a relearn phase, because the model was shaped almost entirely around false signals. Expect performance to look volatile, and often depressed, for several weeks while Smart Bidding rebuilds its understanding from cleaner data — typically a 7-to-14-day learning window for a normal change, extending toward 30 days for a full cleanup of a badly polluted account. Document the change so stakeholders understand a temporary dip is expected, not a failure of the new setup. A clean architecture from the start compresses this window, since fewer, higher-quality signals give the algorithm clearer contrast between buyer and non-buyer behavior.

Anyone running Google Ads bidding strategy work in 2026 should build this audit into every account before testing a new bid strategy — and it pairs naturally with reviewing the broader bidding changes PPC managers need to track this year.

Audit checklist before your next Smart Bidding change

• One macro-goal per campaign objective, mapped directly to revenue.
• All micro-actions — cart adds, clicks, checkout starts, page views, chats — tagged secondary.
• Every GA4-imported macro-goal manually verified and promoted to primary.
• Phone calls evaluated by sampled call quality, not assumed value.
• Every custom goal audited for force-promoted secondary actions.
• Full funnel coverage represented in secondary, for diagnostics only.
• No vanity events sitting in secondary slots.
• Seven to 14 days budgeted for a standard learning phase after any change.
• Up to 30 days budgeted for a relearn period after a full architecture cleanup.

Where human judgment still sits in 2026

Smart Bidding will keep absorbing tasks that used to require manual bid management, and the visible surface of paid search keeps shrinking as a result. What doesn’t get absorbed is signal architecture — deciding what data the algorithm is allowed to learn from in the first place. That’s a business decision built on pipeline math, sales cycles, and revenue attribution, not something the platform can infer on its own.

Configured well, the primary vs secondary framework lets the algorithm scale toward the outcomes that actually matter. Configured poorly, it scales the wrong outcomes faster and finds more ways to be wrong. The framework is the strategy; the bid itself is just the output of how well that architecture has been set up.

Frequently asked questions

What is the difference between primary and secondary conversions in Google Ads?

Primary conversions populate the “Conversions” column and are actively used by Smart Bidding to train and optimize. Secondary conversions populate “All conversions” for reporting purposes but are excluded from the bidding algorithm’s optimization target.

Why would fixing conversion tracking hurt performance short-term?

Removing polluted signals forces Smart Bidding into a relearn phase. Because the model was previously shaped by low-intent signals, performance often dips for one to several weeks while it rebuilds around cleaner, revenue-aligned data.

Should phone calls be primary or secondary conversions?

It depends on call quality, not the action label. Sample recent calls: if most are informational, keep calls secondary; if most result in bookings or sales conversations, calls should be primary.

Do GA4 imported conversions automatically optimize Google Ads bidding?

No. Conversions imported from Google Analytics default to secondary status in Google Ads. If that imported action is your real macro-goal, it must be manually promoted to primary.