A mid-market SaaS client selling workflow automation to logistics companies came to us in Q1 2026 with a familiar problem: their Google Ads account was generating form fills, but fewer than 12% of those leads ever reached a sales-qualified stage. Their cost-per-SQL had ballooned to $1,840, and the sales team was spending two hours per week disqualifying inbound contacts from industries they never targeted. Over 90 days, we rebuilt their paid search approach around account-based targeting principles, and the cost-per-SQL dropped to $718, a 61% reduction, without cutting the total ad budget.
The Root Problem: Broad Intent Without Audience Filters
The original account was structured around high-volume intent keywords like 'workflow automation software' and 'logistics management platform'. These keywords pulled in real searchers, but with no audience layering, the ads reached everyone from solo operators to enterprise procurement teams and final-year students researching for university dissertations. The click-to-lead rate looked healthy at around 4.2%, which masked the quality problem completely.
When we audited the lead data against the CRM, 54% of form fills came from companies with fewer than 20 employees, a segment the client had explicitly excluded from their ICP. There was no suppression list, no firmographic bidding logic, and no Customer Match audience applied anywhere in the account. The budget was effectively subsidising a free discovery tool for people who would never buy.
This is a pattern we see constantly, and it is well documented in our breakdown of why Google Ads fails to generate quality B2B leads. Volume and quality are genuinely different metrics, and optimising for one without measuring the other will always produce this outcome.
The Account-Based Layer We Added
Account-based paid search does not require a separate ABM platform to work at a basic level. The core mechanic is applying audience bid modifiers and observation segments that approximate firmographic targeting inside Google Ads. We built three audience layers: a Customer Match list of 1,400 known target accounts uploaded from the client's CRM, a remarketing list of users who visited the pricing or features pages but did not convert, and a custom segment built around competitor brand searches and industry publication URLs.
Each of these ran in 'observation' mode first for two weeks so we could measure the conversion rate differential before committing to aggressive bid adjustments. The Customer Match segment converted at 3.1x the baseline rate. The pricing-page remarketing segment converted at 2.4x. The custom intent segment sat at 1.6x. These multipliers became the basis for our bid modifier stack, and we adjusted them weekly as data accumulated.
We also restructured the campaign architecture to separate branded, competitor, and generic intent into distinct campaigns with individual budgets, following the same logic outlined in this guide to structuring Google Ads for B2B. This prevented high-spend generic campaigns from cannibalising budget that should have gone to the higher-converting branded and competitor segments.
Suppression and Negative Keyword Work
Alongside the audience layering, we ran a full negative keyword audit. The account had 47 negative keywords across all campaigns. By the end of week three, that number was 310. The categories we prioritised for suppression were SMB-signalling queries (searches containing 'free', 'cheap', 'for small business', 'starter plan'), job-seeker queries ('workflow automation jobs', 'logistics software careers'), and research queries ('what is workflow automation', 'workflow automation examples thesis').
The immediate effect was a 22% drop in impression volume, which the client's previous agency would have flagged as a problem. In practice it was the intended outcome: the budget was now concentrating on a smaller, better-qualified pool of searchers. CPCs rose by about 18% as a result, since the cheaper, lower-intent traffic was gone, but conversion rates to SQL improved fast enough that the net cost-per-SQL still fell sharply within the first 30 days.
Landing Page Alignment With ICP Segments
One issue that became visible once the targeting improved was that the landing pages were built for generic visitors, not for the specific ICP. The headline read 'Automate Your Workflows Faster', which says nothing to a logistics operations manager evaluating software for a 200-person dispatch team. We rebuilt two landing pages, one targeting logistics and freight companies explicitly, and one targeting third-party logistics providers, with industry-specific copy, a relevant case study reference, and a lead form that asked for company size and industry upfront to pre-qualify submissions on the form itself.
The form qualification step reduced raw submission volume by 31%, which the client initially pushed back on. But the SQL rate from those submissions jumped from 12% to 49% within six weeks, because anyone who completed the longer form was a real buyer doing real evaluation. This dynamic is worth understanding in depth, and we cover it in detail in our analysis of why B2B landing pages fail to convert.
According to LinkedIn's B2B Buyer Journey research, B2B buyers now complete roughly 70% of their research before contacting a vendor, which means the landing page is often the first live interaction with a human-facing message. Generic pages waste that moment entirely.
Results After 90 Days
By day 90, the numbers were as follows: cost-per-SQL dropped from $1,840 to $718, a 61% reduction. SQL volume held at roughly the same level as before, meaning the improvement came from quality, not a trade-off against quantity. The sales team's disqualification time fell from two hours per week to under 20 minutes. Monthly ad spend stayed flat at $18,500.
The changes that drove the largest share of improvement were, in order: Customer Match bid modifiers (+3.1x conversion rate uplift), landing page ICP segmentation (+37 percentage points on SQL rate from form), and negative keyword expansion (reduced wasted impressions by 22%). The account-based audience layer was the highest-leverage single change, but it only produced these results because the structural and landing page work happened at the same time.
This is not a tactic that works in isolation. Account-based paid search requires clean CRM data to build the Customer Match lists, a well-segmented campaign structure to apply bid logic at the right level, and landing pages that speak to a specific buyer rather than a generic persona. When all three are in place, the economics of B2B paid search change substantially.
What to Replicate and What to Watch
The approach above is replicable for most B2B SaaS companies with a defined ICP and a CRM list of at least 500 target accounts. The minimum viable Customer Match list for meaningful signal is around 1,000 matched users in Google's system, which typically requires uploading 1,500 to 2,000 contacts to account for match rate attrition. Below that threshold, the audience is too small for Google to apply bid modifiers reliably, and you will see erratic performance data.
One risk to manage: Customer Match lists go stale quickly in B2B, where job changes are frequent. A contact uploaded 18 months ago may now be at a different company and a different email domain. We recommend refreshing the list from CRM every 60 days and suppressing churned customers separately so that win-back messaging does not mix with net-new acquisition campaigns. These two populations have very different intent signals and should never share a budget pool.
- Upload Customer Match lists from CRM, minimum 1,500 contacts, refresh every 60 days
- Run audiences in observation mode for two weeks before applying bid modifiers
- Separate branded, competitor, and generic campaigns before layering audience logic
- Expand negatives aggressively, targeting SMB-signal, research, and job-seeker queries
- Build ICP-specific landing pages with qualification questions on the form itself
- Measure success at SQL stage, not at form-fill stage