Smart Bidding Is Optimizing for the Wrong Conversions. Here's the Fix.

Smart Bidding learns from whatever you tell it counts. If your primary conversion is a raw form submission, the algorithm will spend your budget finding people who fill out forms, not people who buy. Value-based bidding advertisers see an average 14% lift in conversion value at a similar ROAS once they import qualified, revenue-tied signals, according to Optmyzr's value-based bidding benchmarks. The fix is not a new bid strategy. It's a conversion signal overhaul.

I keep seeing the same pattern in paid search accounts. The team is running tCPA or tROAS, hitting the target, and the CMO is asking why pipeline is flat. The algorithm is doing exactly what it was told. Nobody told it what a good lead looks like.

The Part Nobody Reads: Include in Conversions

Every conversion action in Google Ads has an "Include in Conversions" toggle. Toggle on means the action counts in the Conversions column and feeds Smart Bidding. Toggle off means it lives in All Conversions for reporting only. In the primary/secondary framing Google rolled out, primary actions are used for bidding, secondary actions are observed (unless the secondary is part of a custom goal).

Here's the audit that actually matters. Open Tools & Settings, go to Conversions, and look at every action sitting in primary status. Ask one question per row: does more of this, by itself, make the business more money? If the answer is "sometimes" or "it depends on what happens after," that action does not belong in primary.

The usual suspects that get left in primary by accident:

• Newsletter signups and content downloads on a lead-gen account
• "Add to cart" or "Begin checkout" on an ecommerce account where purchase tracking is already firing
• Any conversion action imported from GA4 that overlaps a natively tracked Google Ads conversion (double-counting, silent but expensive)
• Phone call conversions with a 15-second threshold (catches robocalls, wrong numbers, and curiosity clicks that are not buyers)

A 3-day conversion tracking outage can depress bid accuracy for two weeks if you don't log a data exclusion, because Smart Bidding keeps metabolizing the bad signal well after you notice the issue, as Cometly's tracking diagnostics guide lays out. Every day of corrupted data compounds. The longer the algorithm trusts the wrong thing, the longer the recovery.

The Lead-Gen Trap: Optimizing for Forms, Not Pipeline

Lead gen accounts are where this breaks most visibly. B2B teams wire up a form submission, flip it to primary, run tCPA, and the algorithm faithfully delivers more form submissions. A lot of them are students, competitors, job seekers, and bots. The sales team complains. Marketing blames intent. Nobody blames the signal.

There's a detour problem specific to search that Search Engine Land covered in 2026: buyers now route through Reddit, Perplexity, and AI Overviews before they ever hit your ad, which changes what a "short path to conversion" actually means. The shortest path is usually the least-informed buyer. Smart Bidding, optimizing for cheap and fast conversions, systematically pulls spend away from your most qualified traffic because that traffic takes longer to convert. It looks like underperformance in the UI. It's the opposite.

The counter-move is old news that most accounts still haven't executed: enhanced conversions for leads. You capture hashed email and phone from the form, wire your CRM to upload qualification status back to Google Ads as a conversion event, and Smart Bidding starts optimizing on SQLs or closed revenue instead of raw form fills. PPC Mastery's offline conversion tracking breakdown runs through the CRM-stage mapping in practical detail.

The benchmark most teams don't check: after you switch to CRM-stage conversions, expect raw lead volume to drop 30-50% in the learning period. That's not the strategy failing. That's the strategy working. The algorithm was overbidding on junk. Pipeline value per dollar is the number that actually moves. If your dashboard still grades the campaign by lead count, rebuild the dashboard before you flip the switch, or the first weekly review will get the change reversed.

Micro-Conversions: Sometimes Useful, Usually Dangerous

Micro-conversions have a legitimate job. In low-volume accounts where the algorithm can't learn on terminal events (fewer than roughly 30 conversions per 30 days, which is Google's threshold for tCPA learning exit), upper-funnel events feed the model enough data to bid intelligently. That's the real use case.

The danger comes when micro-conversions stack up in primary status alongside the real goal. Count a purchase as 1, count a PDF download as 1, count a pricing page view as 1, and Smart Bidding cannot tell which one it should actually chase. As ALM Corp's micro-conversion analysis puts it, too many events in primary dilutes intent and pushes automation toward cheaper, easier actions with no corresponding revenue lift.

A cleaner structure, from what I've seen work:

1. One terminal primary (purchase, SQL, closed revenue). This is the thing you actually care about.
2. Micro-conversions in secondary status for observation. You see them in All Conversions, you can still diagnose funnel drop-off, but they don't pull the algorithm off-course.
3. If volume is genuinely too low for the terminal event to drive learning, promote ONE micro-conversion to primary and assign it a conversion value that reflects its probability of closing. Not 1. Use your actual close rate against contract value as the multiplier.

This is value-based bidding's real promise. You stop treating every conversion as equal. The 14% lift figure cited earlier isn't magic. It's what happens when the algorithm finally gets told that a $50,000 enterprise deal and a $500 SMB trial are different.

Conversion Value Rules: The Lever Most Accounts Never Touch

Conversion Value Rules let you adjust the value of a conversion based on device, geography, or audience, applied on top of whatever value your tracking sent. The adjustment range runs from 0.5x to 10x, or you can add a flat amount.

Practical examples where this pays for itself fast:

• Mobile converts fine but desktop closes at 2x the revenue. Multiply desktop conversion value by 1.8 and let Max Conversion Value shift spend accordingly.
• Your first-party data says customers in three specific metro areas have 3x LTV. Audience rule, multiplier of 3.
• Returning visitors close at higher rates than new users. Audience list rule, modest multiplier (1.3-1.5 is usually defensible before it starts cannibalizing prospecting).

The rule I wish more accounts used: a negative rule for low-intent geographies or devices where historical data shows the CPA is real but the lifetime value never materializes. Set the multiplier below 1 and the algorithm pulls back without you having to carve out separate campaigns.

One caveat I'd flag. These rules are only as good as your first-party data. If your CRM says returning users close 1.3x but your sample is 40 customers across 6 months, you're overfitting. Give it volume before you tune the multipliers.

The 2026 Shift: Journey Aware Bidding

Google announced Journey Aware Bidding in late 2025, with a 2026 rollout for Search campaigns using tCPA. Search Engine Roundtable's coverage captures the change. Today, Smart Bidding only uses primary conversions to optimize. Secondary conversions sit in reporting purgatory unless they're in a custom goal. Journey Aware Bidding flips that. You'll be able to let secondary actions influence the bidding algorithm without counting them in your Conversions column or hitting your CPA math.

The practical implication: the whole "promote a micro-conversion to primary to feed the algorithm" workaround gets smaller. You can leave your terminal event as the only thing in primary, keep reporting clean, and still hand Google the intermediate signals (content downloads, MQLs, webinar registrations) it needs to infer high-probability users earlier in the journey. That's a real upgrade for lead gen accounts where terminal-event volume is too thin to drive learning on its own.

I think most teams will get this wrong the first time. The instinct will be to fire everything in the funnel at the algorithm and hope it sorts itself out. From what Google has previewed with Search Engine Land's preview coverage, the early data looks best when the secondary signals are tightly mapped to funnel stages you've actually validated against closed revenue. Random middle-funnel events don't help. They probably hurt.

A 30-Day Signal Audit You Can Run This Week

None of this requires new tooling. It requires an afternoon with the conversions table open.

1. Day 1. Export every conversion action, its status (primary/secondary), its 30-day volume, its value rule (if any), and its Include in Conversions setting. This is your baseline.
2. Day 2-3. For every primary action, map it to a dollar value that reflects your actual close rate or ROAS. Unassigned values or "1 per conversion" on a lead-gen account is the root cause of half these problems.
3. Day 4-7. Demote any primary action that doesn't pass the "more of this, by itself, makes money" test. Move it to secondary. Yes, reported conversion volume will drop. Your CPA will look worse. The algorithm will start optimizing on the right thing.
4. Week 2. Stand up enhanced conversions for leads if you don't have it. Even without a full CRM integration, the user-provided data tag alone improves match rates enough to be worth the hour of setup.
5. Week 3. Add conversion value rules where your first-party data has volume. Start conservative, one rule at a time, monitor for 14 days before layering another.
6. Week 4. Review the results against pipeline, not against the Google Ads UI. If pipeline is up and reported conversions are down, the fix worked. If both are down, roll back one change at a time.

The pillar on this ran through the broader argument for running Google Ads without letting Google run you. This piece is the zoom-in on the single lever that decides whether Smart Bidding is your employee or your opposition. Signals, not strategies, are what the algorithm actually sees. And in the same way that Meta's attribution rewrite reshaped what counts as a click-through conversion, Google's moves around Journey Aware Bidding are a quiet invitation to rebuild the signal layer before the bid strategy does it for you.

If the algorithm is optimizing for the wrong thing, the answer isn't a different bid strategy. It's a different definition of winning.

Notice Me Senpai Editorial