Marketing managers in B2B businesses, especially engineering and industrial ones with long sales cycles, tend to arrive at paid search from one of a few starting points. Some have never run it, often because, as in many regional teams at multinationals, the work has always sat with a global team at headquarters, leaving no local account and no history to learn from. Others ran a campaign once, saw a handful of enquiries at an uncomfortable cost per lead, and quietly switched it off. Whatever the starting point, they now have a budget, a quarter or two to show something for it, and a choice between appointing an agency, rebuilding their own skills, or running the account with help from an AI assistant.

Whichever route they take, the first campaign will be judged on the wrong output if it is judged on leads. In a long-cycle B2B market, the early months of Google Search advertising produce something more useful than a lead count and considerably harder to get elsewhere: an accurate picture of who is competing for the business's buyers, what those buyers type when they are ready to look, and whether search can carry a meaningful share of the pipeline at all. That information is what the first budget is buying, and the leads, when they arrive, are a by-product of collecting it.

Why the early lead numbers look bad

Bottom-of-funnel demand in industrial markets is scarce by nature. A company buys a building management system, a fleet of forklifts or a fire suppression contract when an asset reaches the end of its life, when it opens a new facility, or when an acquisition forces it to standardise suppliers, and between those moments it has no reason to search for any of them. Across a whole market these windows open at different times, so the number of people searching with intent to buy in a given month is small, uneven, and only loosely connected to how much an advertiser is willing to spend.

Google's automated bidding strategies are built for the opposite situation. They learn by observing conversions and shifting bids towards the auctions that produced them, which works well when an account records conversions daily and poorly when it records a few a month with weeks of silence in between. A new account starts with no conversion history at all, so the learning period after launch stretches further, and any strategy that depends on conversion data is effectively guessing until enough of it exists. In the accounts we manage, those recording fewer than about thirty conversions a month do better under manual or click-based bidding, where the advertiser decides what a click is worth, and most first-time B2B accounts sit well below that line. A common workaround is to count a softer signal of intent, such as a visit to the contact or locations page, as a temporary conversion so the account has something to learn from, which helps provided everyone remembers that the algorithm is now being trained to find people who visit contact pages, and that the signal is retired once real enquiries arrive in enough volume to replace it.

Put scarce buying demand and a bidding system short of conversion data together and the cost per lead in the first months will be high and volatile, while saying very little about what the channel can eventually do. An enquiry in month five can belong to a contract worth many multiples of everything spent on the account, and an account with a flattering cost per lead can be filling the CRM with students, suppliers and job seekers. Negative keywords will clear most of those out within the first few weeks, although no exclusion list can catch the enquiry from someone who typed exactly what a real buyer types and has no budget, no authority and no live project, and that enquiry counts towards the cost per lead in the same way a qualified one does. None of this is visible in the CPL figure on its own.

Setting the budget before the first click

None of this means starting without a number in mind. Most companies already know, or can work out, what a qualified lead has cost them through other channels: the full cost of a trade show divided by the enquiries that came out of it, an outreach agency's fee divided by the meetings it booked, or an earlier campaign's spend divided by its results. The count that matters in each case is marketing qualified leads, meaning contacts who match the profile of a real buyer and have shown enough intent to justify sales attention. A facilities director at a target account who downloads a specification sheet and asks for a site survey qualifies, whereas a business card dropped in a bowl at a stand usually does not.

That historical cost per MQL is a reasonable anchor for the first search budget and a useful check on what the business is prepared to pay for a qualified conversation. As a threshold for switching the campaign off after six weeks it is a poor one, because it compares a mature channel against a channel that has had no time to learn. The budget itself should be large enough to buy a meaningful number of clicks each month on the terms that matter, and committed for long enough to cover at least one full swing of the market's demand, which in most industrial categories we put at three to six months.

The Insight in auction insights

The auction insights report is the part of the account most first-time advertisers never open, and it is the closest thing paid search offers to market intelligence that could not be bought any other way. For the searches the account entered, it shows which other advertisers appeared, how often they appeared relative to the business, how often their ad sat above the business's ad, and how often they held the top positions on the page. Third-party tools can show which keywords a competitor appears to bid on and what their ads say, whereas auction insights measure how the business actually stacks up against those competitors in the auctions it is paying to enter, with the figures coming from Google itself.

Read over several months, the report answers questions that would otherwise rest on guesswork. A market where two distributors share most of the impressions and nobody else shows up is a very different proposition from one where a dozen advertisers, including a global manufacturer bidding on its own product category, appear on every query. Impression share lost to budget tells the business how much more presence money could buy, and impression share lost to rank tells it whether more money would help at all or whether the ads and landing pages need work first. Those figures are the most honest basis available for setting next year's budget, since they describe the supply of attention in the market and how many others are competing for it.

Search terms: The buyer's own words

The search terms report lists the queries that actually triggered the business's ads, written exactly as buyers typed them. Sales teams hear buyers' language every day, though only from those who have already made contact and have usually framed the problem around a shortlist, whereas this report shows how buyers describe a problem before they know which product solves it, including the many who never get in touch. The gap between the company's vocabulary and the market's can be wide. A manufacturer advertising on its product category may find buyers searching for the regulation they need to comply with, the failure they are trying to prevent, or a competitor's model number, and each of those is a campaign, a landing page or a piece of content the business did not know it needed.

The same report surfaces demand nobody is serving, in the form of queries with clear commercial intent where few advertisers appear or where the organic results fail to answer what was asked. It also shows what to exclude, which in a new account tends to be a long list of job seekers, researchers, students and suppliers hunting for buyers of their own. Google withholds low-volume terms from the report on privacy grounds, so the view is partial, although it is a partial view of real searches made in the period, which is more than any keyword research tool can claim.

Running it as an experiment

Treating the first campaign as an experiment starts with writing down, before launch, the questions it is meant to answer: who competes in this auction and how hard, what a click on the core commercial terms costs, which terms buyers use that the business has not been targeting, and whether the volume of intent-bearing searches is large enough to justify a permanent channel. Those questions give the account a definition of success that three to six months can realistically meet, and they give the marketing manager something defensible to report upward while the lead count is still small.

Three to six months is the range we work to, although the experiment is better ended by its questions than by the calendar: it has run long enough when the auction insights show the same competitors at roughly the same strength month after month, the search terms report has stopped surfacing themes nobody anticipated, and the cost of a click on the core terms has settled. If the answers favour search, the next phase is built directly on them. Campaigns are restructured around the terms buyers actually used, pages and content are written for the problems the search terms exposed, the budget is set against the impression share the account was losing, and the temporary proxy conversion is replaced by real enquiries, with qualified outcomes from the CRM fed back into the account once there are enough of them to change how it bids. From that point the account is a channel rather than an experiment, and cost per lead becomes a reasonable figure to track, provided it sits alongside the pipeline value those leads go on to produce.

Sometimes the answer is that search will carry very little of the pipeline. If the main commercial terms return a few dozen searches a month across the whole region and three established competitors already hold most of the impressions, the business has learned that its budget belongs elsewhere, and it has learned this for the price of a quarter or two of clicks instead of a year-long retainer. That is a legitimate result, and a cheaper one than most other ways of finding out.

The outcome that actually wastes money is the campaign switched off after six weeks because the cost per lead looked wrong, with the auction insights unread and the search terms never exported. The business then chooses its next channel with exactly the information it had before it spent anything, and the buyer whose replacement cycle opens in month seven searches anyway and finds whoever stayed in the auction.