Overview
Billd's paid media program had channel coverage but not channel clarity. Conversion counts looked reasonable on the surface; beneath the surface, low-quality traffic inflated lead volume, campaigns competed for the same audiences, and budget flowed toward angles that cost several times more per lead than the best performers. Over six months, Omni Lab addressed each of these issues systematically. The result was a cost per SQL that fell by more than 75% and a program structured to scale on quality rather than volume.
The Challenge
Billd serves multiple buyer personas within the construction industry, each with different needs, search behaviors, and conversion patterns. A paid media program serving all of them without segmentation produces a noisy signal: lead volume that looks promising in-platform but doesn't hold up when measured against the downstream pipeline.
That was the starting position. Conversion tracking captured top-of-funnel activity without qualification signals further down the funnel. On paid social, campaigns were running with overlapping audiences, driving up costs by bidding against themselves. Certain traffic segments were generating clicks and conversions at scale without advancing through the sales funnel at any meaningful rate. And within Meta, the cost-per-lead variance between the best and worst-performing audience angles was more than 5x, with budget spread across both.
The Approach
Prioritizing lead quality over lead volume. The most consequential decision of the engagement was removing low-quality traffic segments from paid search entirely. An initial bid reduction had already been applied and was still producing too much traffic that wasn't converting downstream. A full exclusion followed across both Google and Bing. Conversions dipped initially, which was expected and accepted. What improved was what mattered: lead quality increased, downstream qualification rates followed, and the cost per SQL began its sustained decline. The same logic was applied to keyword coverage, tightening the traffic signal across both platforms.
Restructuring paid social around what was actually working. An audit identified that multiple Meta campaigns were running overlapping audiences and competing against each other in the same auction. Consolidating to a smaller set of core campaigns eliminated self-competition and clarified the performance picture. More importantly, it surfaced a dramatic cost-per-lead variance between audience angles, with the strongest performer producing leads at roughly a third of the cost of the weakest. Budget was concentrated on the winner. Cost per click on conversion campaigns improved by more than 35% following the restructure.
Building a full-funnel channel architecture. Rather than asking every channel to produce direct conversions, the program was structured around channel roles. Paid search captured high-intent demand from buyers actively evaluating solutions. Meta drove website conversion volume from warm audiences. LinkedIn played a role in demand generation and brand awareness, supporting the retargeting layer rather than being held to direct lead-gen standards. Retargeting audiences across both LinkedIn and Meta were launched once pools reached minimum delivery thresholds, adding a re-engagement layer that hadn't existed at the start of the engagement.
Managing the budget with discipline across a volatile period. More than 20 budget and bid adjustments were made over the 90-day window. Underpacing on secondary search channels required multiple budget increases and targeting adjustments before reaching efficient delivery. Bidding strategies were updated following major structural changes to restore spend velocity without sacrificing quality gains. On paid social, overpacing in certain campaigns required reductions to maintain monthly targets. Each adjustment was tracked and treated as part of an ongoing optimization process rather than a reactive fix.
Testing methodically before scaling. New audience segments and targeting approaches were introduced, with clear validation criteria in place before any scaling decision was made. A smaller, more tightly filtered audience was tested on LinkedIn against a much broader version to determine whether ICP precision produced better engagement at the cost of reach. New match-type coverage on Bing was launched with negatives in place to capture additional search volume without self-competition. Both tests were structured to produce a clear signal before budgets moved.
The Results
The downstream funnel tells the clearest story. Cost per SQL fell by more than 75% over the six-month period, and cost per SAL declined by more than 80%. SQL volume grew more than 4x from the start of the engagement to its peak month. MQL cost per lead improved by more than 40% from January through its lowest point.
Meta drove the highest raw conversion volume among channels. Paid search drove the strongest downstream qualification rates. LinkedIn reinforced both through brand presence and retargeting. Opportunity creation, the furthest-downstream signal tracked, showed consistent growth beginning in the middle of the period and continued to build through the final months of the engagement.
The cost improvements were not the result of spending less. Total investment remained consistent or grew across the period. The efficiency gains came entirely from structural changes: removing traffic that wasn't converting, consolidating campaigns that were competing with themselves, and concentrating budget on the angles and segments that had proven they could produce pipeline.
Strategic Takeaway
Billd's results reflect what disciplined paid media management looks like when quality is held as a non-negotiable constraint. The willingness to accept a short-term dip in conversion volume in order to improve downstream lead quality was the decision that unlocked everything else. That foundation, combined with structural consolidation on paid social and channel role clarity across search and social, produced a program that now operates at a fraction of its original cost per opportunity. The architecture is in place to scale efficiently from here.
