Most B2B paid media accounts optimize for the wrong outcome. They chase form fills, celebrate MQL volume, and wonder why pipeline quality keeps declining. Mike Matta, now Account Director of Paid Media at Directive Consulting, has been writing about this problem for years on PPC Hero, and his conversion weighting framework deserves a closer look from anyone trying to connect ad spend to actual revenue.
The core insight is deceptively simple: not all B2B conversions are equal, so stop treating them that way. A PDF download is not worth the same as a demo request. An MQL is not worth the same as an SQL. Yet most Google Ads accounts still optimize toward a single conversion action, usually the one that happens most frequently, which is almost never the one that matters most to the CFO.
The Funnel Math That Changes Everything
Matta's approach starts with working backward from closed revenue. In his PPC Hero breakdown, he walks through a simple example: if your average deal value is $10,000 and it takes 100 MQLs to generate one closed deal, each MQL is worth $100. If 25% of MQLs become SQLs, each SQL is worth $400. If 50% of SQLs become opportunities, each opportunity is worth $2,000.
This arithmetic forces a conversation most marketing teams avoid. When you assign explicit values to each funnel stage, you expose the true cost of optimizing for volume over quality. A campaign generating 200 MQLs at $50 each looks efficient until you realize only 2% convert to SQL, making your effective cost per SQL $2,500. Meanwhile, a campaign generating 50 MQLs at $150 each with a 20% SQL conversion rate delivers SQLs at $750.
The numbers don't lie, but they do require you to track them.
Why Google's Algorithm Needs Your Revenue Data
Directive's 2026 Google Ads playbook makes the case bluntly: Google is only as smart as the data you feed it. Unlike paid social platforms with firmographic targeting, Google learns who your best prospects are through the conversions you send back. If you only send form fills, Google optimizes for form fillers. If you send revenue signals, Google optimizes for revenue.
This is where Matta's weighting framework connects to platform mechanics. By using Max Conversion Value bidding with differentiated values for each funnel stage, you're teaching the algorithm what actually matters. An MQL might be worth $100, an SQL $400, and a closed deal $10,000. Google's Smart Bidding then allocates budget toward the clicks most likely to generate higher-value outcomes.
Optmyzr's value-based bidding guide explains the underlying logic: you're telling Google things that Smart Bidding can't measure on its own, like which conversions turned into actual revenue and which ones didn't. The platform's machine learning then incorporates these signals into auction-time decisions.
The Implementation Gap
The theory is sound. The execution is where most teams stumble.
Recent B2B SaaS benchmarks show that companies importing offline conversions from their CRM and using value-based bidding generate 3x more pipeline at 31% lower cost per lead. Yet most accounts still optimize for first-touch form fills because the CRM integration work feels daunting.
Matta acknowledges this in his original framework. If you don't have clear visibility into deal size or funnel progression, you can still assign relative weights based on importance. An SQL might be worth 5x an MQL. A demo request might be worth 3x a content download. These aren't perfect numbers, but they're better than treating every conversion identically.
The key is starting somewhere and iterating. Run the weighted approach for 30 days, compare SQL volume and quality against your baseline, and adjust the ratios. The algorithm learns from the signals you provide, so even imperfect signals beat no signals.
Attribution Matters More Than You Think
Matta's framework includes a critical footnote that many practitioners skip: attribution model selection changes everything. Position-based or data-driven attribution reveals which keywords and campaigns actually contribute to high-value conversions, not just which ones happen to be last-click.
Current MQL-to-SQL benchmarks show B2B SaaS companies averaging 18-22% conversion rates, with top performers hitting 25-35%. The gap between average and excellent often comes down to whether teams are measuring the right things. Loose MQL definitions tank conversion rates because marketing counts every content download as qualified, then wonders why sales rejects most of them.
The fix isn't just tighter definitions. It's feeding SQL data back to the ad platforms so targeting shifts toward higher-quality leads within weeks. This creates a virtuous cycle: better signals produce better targeting, which produces better leads, which produces better signals.
What This Means for Your Next Pipeline Review
If you're presenting paid media results to your CFO and the conversation centers on cost per lead without any mention of lead-to-revenue conversion rates, you're having the wrong conversation. The Matta framework reorients the discussion around what actually matters: how much pipeline did we generate, at what cost, and how does that compare to our CAC payback targets?
Pipeline conversion benchmarks show the biggest drop-off typically happens at the MQL-to-SQL stage, where many marketing teams hand over leads that aren't truly sales-ready. By weighting conversions appropriately and feeding those signals back to your ad platforms, you're addressing the root cause rather than celebrating vanity metrics.
The math here isn't complicated. It just requires you to do it.
Start with your average deal value. Work backward through your funnel stages. Assign values that reflect actual progression rates. Implement those values in your bidding strategy. Measure results over 60-90 days, not 7. Adjust based on what you learn.
Matta's been writing about this approach since his Hanapin Marketing days, and the core logic hasn't changed because it didn't need to. What's changed is the platform sophistication available to act on it. Google's Smart Bidding, enhanced conversions for leads, and offline conversion imports all exist to solve this exact problem. The question is whether you're using them.
