The Broad-Match Spend Spiral at an Industrial Distributor
A B2B distributor of industrial fasteners spends $18,000/month on a single Search campaign built from a 4,200-keyword broad-match list assembled by a contractor. Target cost per lead is $120; the account is running at $210. The search terms report shows 38% of spend going to queries like "free CAD file", "fastener torque chart PDF" and "warehouse jobs near me". Twelve keywords absorb 61% of spend with no recorded conversions. Conversion tracking counts a quote-request form plus phone calls longer than 90 seconds. Sales says roughly one in five form fills is a student or hobbyist. The CFO wants a 30% spend cut within two weeks; the marketing manager wants to add 1,500 more keywords to "cover more demand".
1. The CFO wants to cut 30% of spend and the marketing manager wants to add 1,500 keywords. Which direction do you take, and what is your first two weeks of work?
Read the worked answer
Neither proposal addresses the actual defect. The evidence points to intent mismatch, not to a budget level or a volume shortage.
- 38% of spend on research/job-seeker queries means the account is buying traffic that can never convert at $120, no matter the bid.
- 61% of spend on 12 zero-conversion keywords may be waste, or may be assisting conversions elsewhere. You cannot tell from a keyword-level conversion column alone.
Sequence: (1) mine the search terms report and ship negative keyword lists by theme (CAD/download, jobs, student, DIY repair); (2) pull the 12 keywords' search terms and check assisted conversions and conversion lag before pausing anything; (3) pause them one at a time over 2-3 weeks and watch total account conversions, not just that keyword's column. Adding 1,500 keywords before cleaning intent multiplies the defect. Report a reallocation to the CFO, not a flat cut.
2. Sort the query types visible in the account into those Search can profitably serve and those it cannot, and state the test you would use for a borderline case.
Read the worked answer
The dividing line is whether the searcher has a task your product completes and a plausible path to purchase.
- Serve: specification-led queries ("M8 stainless hex bolt bulk", "ISO 4032 supplier"), comparison and supplier queries, and part-number searches. These map to a quote request.
- Do not serve: "free CAD file", "torque chart PDF", "warehouse jobs near me". The first two want a document, the third wants employment. No landing page converts these at a defensible CPA.
- Borderline: "fastener torque chart" could be a maintenance engineer who also buys. Test it as a small, separately budgeted ad group with a landing page that offers the chart and a quote CTA, and judge it on its own conversion rate after enough volume to be meaningful - not on clicks or CTR.
Exclude borderline queries until they earn their place; do not carry them inside a broad-match keyword where you cannot see or control them.
3. How would you determine whether the remaining spend is producing genuinely new business rather than capturing inquiries that would have arrived anyway?
Read the worked answer
You need a comparison group, because the account's own conversion count cannot answer this.
- Split branded from non-branded. Branded search mostly harvests demand you already created. Judge it on incremental orders, not on its attractive CPA.
- Run a geo experiment or holdout where feasible: hold Search off in matched regions for a period and compare quote volume against regions left on, adjusting for baseline trend.
- Import CRM outcomes (quote accepted, order value) so you can compare lead quality across query classes, not just form fills. Sales' "one in five is a hobbyist" complaint is a tracking gap, not a fact you can act on yet.
- Look for cannibalisation signals: rising conversions on generic terms with flat total sales usually means you are paying for traffic that would have found you.
Until one of these exists, treat reported CPA as a directional number and say so.