One difference separates a B2B Google Ads account from a consumer one, and almost everything follows from it: search volume is the wrong selection criterion.
On a market of a few hundred to a few thousand addressable companies, a query typed fifty times a month by people looking for exactly what you sell is worth more than five thousand generic searches. Volume measures an audience. In B2B, intent decides.
This page sets out the structure of an account built for that reality, from what Google’s documentation actually establishes, including three places where it contradicts common practice.
Which keyword match type should you choose?
Match type decides what you pay for, and it is usually left at the default.
What Google documents. Broad match, the default, serves searches related to your keyword including queries that do not contain the direct meaning of your keywords. Phrase match targets searches with the same meaning. Exact match targets searches with the same meaning or intent.
The sentence that matters most to a B2B advertiser. Google states the relationship between the three: broader match types capture all the queries of narrower ones, and more.
Read literally, that changes the diagnosis. Drift toward neighbouring queries is not a malfunction. It is the documented behaviour of the default setting. On a broad market that latitude earns its keep by surfacing phrasings you had not anticipated. On a narrow market it spends budget on searches that resemble yours without being yours.
The practical rule. Exact and phrase for the majority of the account. Broad on a separate, watched budget only, and treated as a query discovery tool rather than a delivery mode.
What that costs you if you ignore it. Not a percentage anyone can quote for your account, but a category of waste you can see directly: open the search terms report and read what broad match actually bought last month. It is the fastest way to settle the argument internally.
Quality Score, and what Google actually says about it
This is the least understood point in the subject, and it should get several dashboards rewritten.
How Google frames it. The Quality Score documentation presents it as a diagnostic tool built on three components: expected clickthrough rate, ad relevance and landing page experience. Each is rated above average, average or below average against other advertisers over the preceding 90 days.
The sentence to read literally. “Quality Score is not an input in the ad auction. It’s a diagnostic tool to identify how ads that show for certain keywords affect the user experience.”
What that does not mean. It does not mean ad and page quality are irrelevant. The auction accounts for them through other paths, which is why a better ad can win a better position at a lower cost. It means the number on your screen is an instrument, not an objective.
What it does mean for anyone managing an account. An advertiser who optimises toward Quality Score is chasing a figure the platform itself does not treat as a performance metric. Worse, they will attribute cost changes to a number that, by Google’s own statement, did not cause them.
The correct and more modest use. A low score signals an inconsistency between what the person typed, what your ad promises and what your page delivers. It is a break detector in that chain, and it usually points at an ad group that has grown too heterogeneous to be answered by one ad.
A second regulator says the same thing, independently. Microsoft Advertising is blunter still: “It is not used at auction time to determine ad rank, and does not affect cost or spend.” Two competing platforms, documenting the same thing without coordinating, is about as solid as this kind of claim gets.
What you actually pay, and why it is not your bid
A documented mechanism escapes many B2B advertisers, and it changes how you set a bid.
You almost never pay what you entered. Google’s documentation is explicit: “You’re often charged less, sometimes much less, than your maximum cost-per-click (max. CPC) bid, which is the most you’ll typically be charged for a click.”
The stated principle. “With the Google Ads auction, you only pay what’s minimally required to clear the Ad Rank thresholds and beat the Ad Rank of the competitor immediately below you.” With no qualified competitor below you, you pay the reserve price. The minimum billable increment is one cent.
The first consequence, and it is the expensive one. Setting a cautious low bid to avoid overpaying does not make you pay less. It excludes you from auctions you would have won below that ceiling. You do not get a discount, you get no impression.
A higher bid does not buy a better position either. Google writes that this is not a conventional auction in which the highest bid always wins, and that you can hold a better position at a lower cost than a competitor bidding more. Ad Rank combines bid amount, ad and landing page quality, the Ad Rank thresholds, auction competitiveness and search context, and Google notes it is computed twice, once for eligibility and once for ranking.
The exception that turns the ceiling into a suggestion. The actual cost can exceed your maximum in two documented cases: if you have enabled enhanced CPC, or if you have set a bid adjustment. The ceiling stops being a ceiling the moment you delegate part of the setting.
One term to stop using. You will read everywhere that Google Search runs a second-price auction. Google never writes it. The documentation describes the principle without naming the model, and when Google moved Ad Manager to first-price bidding in 2019, the announcement stated that the change would have no impact on auctions for ads on Google Search, without saying what those auctions were instead.
Structure by intent, not by catalogue
The structure inherited from consumer retail splits the account along the product catalogue, with very granular ad groups. It justifies itself poorly in B2B, for a reason of volume: an ad group receiving three clicks a week never accumulates enough signal for anything to be learned from it.
One campaign per major buying intent, not per product. Someone trying to solve a problem, someone comparing vendors and someone searching a specific brand carry different value and need different messages. Separating those intents is what lets you allocate budget differently, which is the primary job of a campaign.
Tight ad groups built around keywords that share a meaning, not a spelling. The test is not whether the words look alike. It is whether one ad and one landing page can honestly answer every query in the group.
A landing page that keeps the ad’s promise. This is the most frequent and most expensive break in the chain: a precise ad pointing at a homepage forces the visitor to find on their own what they were looking for. We treat page structure in what makes a B2B landing page convert.
Why granularity backfires here specifically. On a consumer account, splitting into forty ad groups gives each of them enough traffic to be judged. On a B2B account with two hundred clicks a month, the same split gives you forty samples of five clicks, none of which supports a decision. Granularity is only free when volume is.
Negative keywords are the cheapest line item you are not using
Negative keywords are the least visible and most profitable work on a search account, and they are missing from most of them.
On a professional market, a meaningful share of searches do not come from buyers. Six families come up systematically.
Job seekers and students looking for information about your trade. Terms like jobs, salary, career, training, internship, definition and course remove most of it.
Free-tool hunters, with free, open source, download and template.
Competitors doing research, harder to exclude but visible through queries carrying sector brand names.
Consumers, whenever your product category also exists in a retail form.
People applying to work for you, if your brand name is being searched.
Out-of-area queries, if your delivery is geographically constrained.
Where these lists come from, and where they do not. They are built from the search terms report, which shows what people actually typed, as opposed to the keywords you bought. Confusing the two is the most common analytical error on a search account. A monthly review of that report is the highest-return maintenance task on a B2B account, and it takes about ten minutes.
What you count as a conversion decides everything else
An account is managed on what it measures, so a wrong measurement produces wrong decisions with perfect regularity.
Counting actions with no commercial value. Document downloads, contact page views, time on site. These fire often and produce an account that looks like it is performing while the pipeline stays empty. The algorithm, meanwhile, conscientiously optimises toward whatever you pointed it at.
Counting the same thing twice. A form firing two events, or a conversion recorded on every load of the thank-you page, inflates results artificially.
Counting only what happens online. In B2B a share of enquiries arrive by phone or direct email after a visit. An account that only sees forms understates its own effect, which leads people to cut campaigns that were working.
The fix is methodological before it is technical. Decide what counts as a conversion, write it down, and verify the tool counts that and nothing else. If your CRM lets you import closed deals back, the nature of the exercise changes: you optimise toward revenue rather than toward forms. Where your counters disagree is treated in why GA4 and Meta conversions never match, and the plumbing in server-side tracking.
Smart Bidding, and the threshold most B2B accounts never reach
This is where the gap between platform messaging and the reality of a small B2B account is widest, and it can be quantified.
What the strategies have in common. Maximise conversions, target CPA, maximise conversion value and target ROAS all rest on the same principle: the algorithm adjusts the bid at every auction using signals you cannot see.
The threshold Google itself publishes, and nobody quotes. Its guidance points at a minimum of thirty conversions, and fifty for target return on ad spend, for performance to be measured properly. It also asks that results be judged over periods of a month or more.
Now hold that against your reality. A B2B company producing ten to fifteen enquiries a month, which is already respectable on a market of a few hundred accounts, never reaches thirty conversions inside a window where the signal is still fresh.
Three observable effects follow. Results swing week to week with nothing having changed, because each additional conversion moves the model a lot when conversions are few. Unable to tell good conversions from bad ones, the algorithm optimises toward whatever happens most often, which means it will go and find your document downloads if downloads are in the conversion set. And no test comparing two strategies on samples that small can be concluded, whatever the apparent gap.
Three practical answers, none of which is refusing automation on principle. Broaden what counts as a conversion without falsifying it, counting the steps that carry real commercial value such as a demo request or a quote request, weighted if useful. Start on manual or maximise clicks long enough to build history, then switch, because starting a conversion strategy with zero data is asking for a judgement from someone who has seen nothing. And judge on the month rather than the week, which is the platform’s own recommendation and the first one internal pressure discards.
Which default settings cost you money?
An account created quickly inherits several settings that suit consumer retail and suit B2B considerably less. Four deserve an immediate check.
Search partners. On by default, they serve your ads on third-party sites using Google search. The traffic is less qualified and reporting does not always separate it clearly. Given B2B click costs, the question is worth asking rather than inheriting.
The attached Display network. Some campaign types offer to extend delivery to banners. Mixing search and display inside one campaign makes analysis impossible, because two entirely different mechanics share a budget and a report. If you want display, run it separately.
Geographic targeting. The default targets people in or interested in your area. For a locally delivered service, that pays for clicks from people interested in the region without being in it. Presence-only is almost always the right setting in local B2B.
Accelerated budget delivery. A daily budget consumed in three hours puts your ads in front of a fraction of the day. In B2B, where most professional searching happens on weekdays during business hours, distribution matters more than it does in consumer retail.
One decision rather than a setting: responsive search ads. They generate automatic combinations from the headlines and descriptions you supply. That works, provided you do not mix incompatible promises in one asset group, because the algorithm can then pair a headline and a description that together say something you never intended.
At what pace should a B2B account be optimised?
A B2B account is not managed at the tempo of a consumer one, for an arithmetic reason: with few conversions, every decision taken early rests on noise.
The week says nothing. On ten monthly conversions, a week at three and a week at one indicate no trend. Looking at numbers daily mostly produces anxiety and intervention, and every intervention restarts the algorithm’s learning phase.
The month is the useful unit, and the quarter the one for structural decisions. That is consistent with the platform’s own guidance to judge over a month or more.
The sales cycle shifts everything. If your deals close in six months, a campaign launched in January produces revenue in June. Judging its profitability in March is measuring a harvest during germination.
Seasonality is real but market-specific. Professional search volume drops in the weeks most of your buyers are away, and rises at the two points in the year when budgets unlock. Which weeks those are depends on your market, not on a general rule, and the answer sits in your own last two years of data rather than in anyone’s article.
What to do about it, and what not to. Cutting a campaign entirely during a slow period costs you history and algorithmic stability, which you pay for on restart. Reducing the envelope without stopping is usually preferable.
Where to go next
You are choosing between platforms. How to split budget between Google and Meta in B2B.
You run Meta as well. Meta Ads account structure, and Meta Advantage+.
You are setting a budget. Small business advertising budget, and marketing budget as a percentage of revenue.
Your counters disagree. Why GA4 and Meta conversions don’t match, and server-side tracking.
You want the metric to run on. ROAS vs MER vs CAC vs LTV, and how to calculate CAC.
Your landing page is the weak link. What makes a B2B landing page convert.
In short
- Volume is the wrong criterion in B2B. Fifty high-intent searches beat five thousand generic ones, because volume measures an audience and not an intention.
- Broad match drifts by design. Google writes that it serves queries which do not contain the direct meaning of your keyword. On a narrow market it is a discovery tool, not a delivery mode.
- Quality Score is not an input in the ad auction, in Google’s own words, and Microsoft states it does not affect cost or spend. It is a coherence diagnostic between query, ad and page.
- Your maximum bid is a ceiling, not a price. Bidding low to be careful does not earn a discount, it loses auctions you would have won cheaply. Enhanced CPC and bid adjustments break the ceiling.
- Google never calls Search a second-price auction. It describes the principle without naming the model, and its 2019 move to first-price bidding in Ad Manager explicitly excluded Search.
- Structure by intent, not by catalogue. An ad group getting three clicks a week learns nothing, and granularity is only free when volume is.
- Negative keywords are the cheapest line item, and they are built from the search terms report rather than from your keyword list.
- Smart Bidding wants thirty conversions, fifty for target ROAS. Below that it is unstable rather than wrong, and it drifts toward whatever event fires most often.
If you have a live account and a doubt about what it is actually buying, the search terms report answers in ten minutes. Book a diagnostic, or see how we approach B2B paid acquisition.