A familiar pattern appears in small advertising accounts. A business launches paid search, spend leaves the card every day, the dashboard reports conversions, and the sales figures at the end of the quarter look the same as they did before the campaign started. Nobody in the business can point to a decision that went wrong, because no decision was ever taken. The account simply ran the way it arrived.
This guide explains why that outcome is so common. It works through the settings that are enabled when an account is created, shows how each one raises spend and reported performance at the same time, names the two controls that genuinely remain with the buyer, and gives the reporting cut that makes waste visible whoever manages the account. The named examples come from Google Ads because it is the largest search auction, and the mechanics apply to every auction platform in every market.
The defaults that spend the budget
The usual explanation for a failing paid search account is a list of advertiser errors: the wrong keywords, weak adverts, careless budgets. That list describes symptoms rather than the cause. Almost every item on it is the account behaving as it was configured on the day it was opened, by a configuration the advertiser did not write.
Those settings move two numbers in the same direction. They increase how much of the budget is spent, and they increase the clicks, impressions, and conversions the interface reports. The graphs rise, the conversion column fills, and the money goes to searches and placements no owner of the business would have chosen to buy.
An untouched paid search account looks successful, because the settings that waste the budget also inflate the report that is supposed to detect the waste.
None of this requires dishonesty from the platform. A search auction is a marketplace, and the party that writes the default settings is the party selling the inventory. Defaults written by a seller favour volume, reach, and spend, because those are the seller's revenue. Efficiency is the buyer's responsibility in every media market, as the guide to what advertising really costs sets out.
A search auction sells a chance to appear when a query is typed. Four things in that arrangement are set by the seller: how widely a keyword's meaning is interpreted, where else the advert appears, what counts as a result, and who splits the budget between formats.
Broad matching without a negative list
The most consistently reported source of wasted paid search budget is broad matching used without a maintained list of negative keywords. Broad match allows the platform to serve an advert against any query it judges related to the keyword. A business selling commercial flooring can pay for searches about flooring courses, flooring jobs, and repairs it does not carry out.
The behaviour is not restricted to broad match. Phrase and exact match both serve close variants, which cover plurals, misspellings, and terms the platform reads as the same meaning. An advertiser who believes exact match means exact is usually wrong about a measurable share of the spend.
Optmyzr's study of more than 1,400 accounts spending above ten thousand dollars a month found exact match ahead of broad match on click-through rate in roughly nine accounts in ten, and on cost per acquisition and return on ad spend in roughly three in four. Broad match remains usable. It requires an exclusion list, and that list is what is missing in most accounts.
A working negative keyword list is a maintained structure rather than a set of obvious exclusions added once and left alone. It usually contains five groups.
- Intent exclusions. Words that reveal a searcher who will never buy: free, cheap, DIY, template, salary, jobs, course, definition.
- Product exclusions. Adjacent products and services the business does not sell, including the categories the platform most often confuses with its own.
- Audience exclusions. Terms used by students, researchers, job seekers, suppliers, and competitors rather than customers.
- Geographic and language exclusions. Place names outside the areas the business serves, which matter even when location targeting is set, because a query can name a place the account does not cover.
- Brand exclusions. Competitor names, where bidding on them should be deliberate, and the advertiser's own name, which belongs in a separate campaign.
The list is built from evidence. Every week or two the search terms report is read in order of spend, and anything that should never have been bought is added as a negative at the correct level. Six months of that discipline produces a list specific to one business and one product set.
The upgrade that widens matching without a decision
Match settings are not fixed for the life of a campaign. During 2026 Google began moving existing search campaigns onto AI Max for Search, a package in which search term matching is enabled by default. Search term matching allows a campaign to serve against queries the platform selects rather than only against the keywords the advertiser listed.
The timing matters, because the change reaches accounts whose owners have taken no action. Campaigns using the campaign-level broad match setting and campaigns using automatically created assets are upgraded automatically from September 2026, and search term matching arrives switched on in both cases, and text customisation arrives switched on for automatically created assets. The equivalent upgrade for Dynamic Search Ads was deferred to February 2027.
The control still exists. Search term matching is a setting at ad group level, and a campaign can be returned to the keywords its owner chose. Three checks establish where an account stands.
- The search term matching setting on every search campaign. Record whether it is enabled by choice or by upgrade.
- The search terms report, read for query types the keywords do not name. New categories of query appearing without a keyword change are the signature of widened matching.
- Spend, orders, and cost per order for the ninety days either side of the upgrade, taken from business records rather than the platform's conversion column.
Wider matching with an unchanged exclusion list moves budget to queries nobody selected, so the negative list grows with it.
Partner and extended inventory included by default
A search campaign does not only appear on the search engine. Search partner networks place the same adverts on other search sites, directories, and apps, and display expansion can place them on ordinary web pages. Search partner placements are included by default on a new search campaign. Display expansion is a separate checkbox, most often found switched on in accounts built through a guided setup flow. Many advertisers have seen neither.
That inventory is cheaper per click, which is why blended cost figures often improve when it is included. It converts far less reliably, and in some accounts it produces a large share of clicks and almost no revenue. The improvement in the average is a mathematical artefact rather than a commercial gain.
Some accounts do earn from partner inventory, so no campaign should be assessed until its performance has been separated by network and by placement. A single blended number hides two segments performing very differently.
The same reasoning applies to search terms that never appear. Search terms reporting has excluded low-volume queries since 2020, so a share of clicks is grouped under an unnamed category. Reported estimates commonly put the visible share at around seventy to eighty per cent of query data. A buyer cannot exclude a term that is never shown, which makes account structure, covered below, the practical defence.
Recommendations that are applied automatically
Advertising platforms produce a stream of suggestions: raise a budget, add keywords, widen a match type, change a bidding strategy. Presented one at a time with an optimisation score attached, they read as maintenance. Many accounts have automatic application switched on for some categories, which means the suggestion is implemented without a person approving it.
The pattern of those suggestions is consistent. They broaden targeting, raise budgets, loosen match types, and add audiences. Very few narrow anything. A system that recommends the features it has will recommend expansion, because expansion is what the features do.
An advertiser can inspect a campaign, agree with its structure, return a month later, and find the targeting materially wider than the version approved, with no record of anyone approving the change. The account did not drift. It was changed.
Automatic application converts a suggestion into a change of strategy. The advertiser keeps the invoice and loses the decision.
The remedy is administrative rather than technical. Automatic application is switched off for anything that widens targeting or raises spend, and each recommendation is accepted only when it survives the same scrutiny as any other budget decision.
Conversions counted more generously than they should be
The conversion column is the number every judgement about the account depends on, and the advertiser defines it. Two habits inflate it, and both are common in accounts set up quickly.
The first is counting actions with no commercial value. Page views, scroll depth, video plays, and newsletter sign-ups from people who will never buy can each be registered as a conversion, and once registered each is fed to the bidding system as a target. The platform then buys more of the traffic that produces them, which is the traffic least likely to produce revenue.
The second is generous counting rules. Conversion actions can be set to count every occurrence rather than one per customer, which turns a refreshed confirmation page into several sales. View-based counting credits adverts that were seen rather than acted upon.
The consequence is a report that no longer measures what the business cares about. Where several platforms claim the same order, the reconciliation is covered in the guide to which marketing channel is working. Inside one account the repair is immediate: one primary conversion, defined as an event that produces money or a qualified enquiry, counted once per customer, with everything else demoted to a secondary metric that informs nothing about bidding.
Automated bidding and bundled campaign types
Automated bidding sets the price of each auction entry on the advertiser's behalf, using signals no manual bidder can see. Bundled campaign types distribute one budget across search, shopping, video, and display according to the platform's own judgement. Both are genuinely capable, and both transfer budget allocation from the buyer to the seller.
The relevant question is what the optimisation is aimed at. The platform optimises towards the conversion signal it has been given, within its own inventory. It does not know the gross margin on any product, or the fact that one order type is profitable and another loss-making at the same revenue. It cannot allocate towards a margin it has never seen.
Automated systems find volume against a badly chosen target with the same efficiency they apply to a good one. Automation amplifies whatever the conversion definition contains. A good definition produces more of the right traffic, and a poor one produces more of the wrong traffic.
Most advertisers should neither refuse automation nor accept all of it. Automation is used where the conversion signal is trustworthy and the volume is sufficient to learn from, and the buyer keeps the decisions automation cannot take: which intents get their own budget, and what counts as a result. Those two are the substance of professional paid media management.
The two constraints that remain with the buyer
As platforms automate more of the buying decision, the list of controls that matter gets shorter. Two cannot be overridden by the platform's own optimisation, and an advertiser holding both can run a profitable account with a great deal of automation switched on.
Account structure that separates intents
Structure controls where money goes when the platform decides which query to answer. Searches that mean different things belong in different campaigns, because a campaign is the unit that holds a budget. Mixing them means the platform allocates between them, towards whichever produces the cheapest recorded conversion rather than the most valuable one.
- Brand searches separately from everything else. People searching for the business by name are cheap to reach and would often have arrived anyway. Left in a general campaign they improve the account's average without improving the result.
- Buying intent separately from research intent. A query naming a product or a price behaves differently from a query asking how something works, and the two should not compete for the same budget.
- Profitable product lines separately from the rest. If margin differs materially across what the business sells, the campaigns should differ too, because that is the only way a margin difference reaches the bidding.
- Markets and languages separately. Costs and conversion rates differ by market, and a blended figure will misprice all of them.
A conversion definition tied to money
The second constraint is the conversion signal. It should represent an event a finance director would recognise: a sale, a booking, a qualified enquiry that a person has assessed. Where values differ between events, they should be passed into the platform so that automation optimises towards revenue rather than a count. Where a lead requires qualification, the qualified stage is what should be counted, even though that reduces the number in the report.
The reporting cut that exposes waste
Three reports will show whether the budget is buying anything worth having, and they apply equally to an account managed in-house and to one managed externally. They are deliberately simple, and an owner should ask for them by name.
- Search terms sorted by spend, highest first. Spend is the sort order, in place of clicks or conversions. Read the top fifty and mark every term the business would never have chosen to buy.
- Spend against revenue by campaign, over ninety days. Cost on one side, real orders or qualified enquiries on the other, taken from business records rather than the advertising interface. A campaign that cannot show the second column is assessed only on figures the platform reports about itself.
- Performance separated by network, placement, and device. Each reported apart from the others, so that a blended average stops hiding a segment that consumes budget and returns nothing.
From the first report comes the number that settles most arguments: the share of total spend that went to search terms nobody in the business would have selected. In a maintained account that share is commonly in the range of five to fifteen per cent. Above roughly a quarter, the account is being run by its defaults, whatever explanation is offered. The exact threshold depends on margin, so the figure worth tracking is the direction it moves over three months.
These reports assess the buying of the click. An account can be bought perfectly and still produce nothing if the destination fails, which is why media spend and landing page optimisation belong to the same review, and why why a website is not converting is a separate diagnosis.
The same audit on any platform
The mechanism has nothing to do with any one company. It follows from the structure of an auction marketplace, so it appears wherever that structure appears. Only the vocabulary changes.
- Social advertising platforms. Automatic placements extend delivery beyond the surfaces that were chosen, and audience expansion adds people outside the audience that was defined.
- Retail media inside marketplaces. Automatic targeting adds search terms and product pages the advertiser never selected, billed alongside the ones it did.
- Video and app inventory. Content suitability settings arrive at their broadest usable level, so adverts appear beside material the business would not have chosen.
Five questions do most of the work on any of them: what is being interpreted more widely than intended, where else the advert is appearing, what is counted as a result, who decides how the budget is split, and what has been upgraded since the campaign was last approved. An advertiser who asks all five on the day a new account is opened will avoid most of what this guide describes.
Broad awareness formats, cheap inventory, and untended settings absorb budget in every channel, as the survey of advertising services sets out.
Doing it yourself versus hiring help
A small paid search account with a narrow product set is manageable in-house, and many businesses should manage it themselves. The requirements are modest: a few campaigns separated by intent, tight match types, a negative list read from the search terms report every week or two, one conversion that means money, and an honest monthly comparison of spend against real orders. The owner usually knows better than anyone which searches are irrelevant.
The economics change with three factors. The first is structural complexity: many products, several markets, different margins, or a long sales cycle where the recorded conversion sits far from the revenue. The second is testing volume, because testing adverts and landing pages requires enough traffic to reach a conclusion. The third is the cost of the mistake, since a large monthly budget makes a month of misdirected spend more expensive than a year of professional management.
One filter is worth applying to any prospective manager. Ask which of the three reports above they propose to send, and how often. An answer built on impressions, clicks, and platform-recorded conversions describes the interface. An answer built on spend against revenue and search term quality is rarer, and it is the one that still holds when the figures are compared against the accounts. The same test applies to fees, covered in the guide to what a marketing agency costs.
Key takeaways
- Waste is the starting condition of a paid search account, because the defaults raise spend and reported performance together.
- Broad and close-variant matching, partner inventory, and automatically applied recommendations all widen delivery without a decision being taken.
- The 2026 matching upgrades arrive with search term matching enabled, so the setting is worth checking on every campaign.
- A conversion definition that includes actions with no commercial value teaches automation to buy worthless traffic.
- Account structure and a money-based conversion definition are the two controls automation cannot overrule.
- Search terms by spend, spend against revenue, and performance split by network expose the waste in any account.
The question of why paid search is not working usually resolves into a simpler one about who has been making the decisions. In an account left as it arrived, the answer is the platform, and the platform is optimising a marketplace it owns rather than a margin it has never seen. Reclaiming a small number of those decisions changes the outcome more than any change of copy.
For an independent read of an existing account, book a strategy call with Reachford. The response addresses account structure, the conversion definition, and analytics and performance reporting, and states plainly where the current arrangement is working.
Frequently asked questions
Why are my Google Ads getting clicks but no sales?
Usually because the clicks are being bought against searches with no buying intent. Broad and close-variant matching serves adverts on loosely related queries, and without a maintained negative keyword list a large share of spend goes to research, job, and free-alternative searches. Read the search terms report sorted by spend before changing anything else. If the terms look right, the problem has moved to the page the click lands on.
Should I turn off search partners and display expansion?
Separate the reporting first, then decide. Search partners is on by default, and display expansion is often on in accounts built through a guided setup. Both are cheaper per click, and in many accounts they convert far less reliably than the search results themselves. Report their spend and revenue apart from the main network for at least a month. Where they earn, keep them. Where they consume budget without producing orders, exclude them.
Is automated bidding a bad idea for a small budget?
It depends on the conversion signal rather than on the budget size. Automated bidding optimises towards whatever it has been told counts as a result, so a poor conversion definition makes it efficient at buying the wrong outcomes. With too little data it also has nothing to learn from. A small account should fix the conversion definition and campaign structure first, then consider automation.
How often should a paid search account be reviewed?
Search terms deserve a reading every week or two, because that is where waste appears first and where negative keywords come from. Structure, budgets, and conversion definitions justify a fuller review each month, compared against real orders rather than platform-recorded conversions. Judge the channel itself over a quarter, since shorter windows rarely contain enough data to separate a trend from noise.