A well-configured search campaign in a normal US vertical produces a readable result in four to six weeks. Revenue can arrive in week one. The gap between those two facts is where most disappointment lives, because a first sale is not evidence and most accounts treat it as though it were.
Search marketing gets sold on speed, and the speed is real in a narrow sense. You can launch Monday and take an order Tuesday. What you cannot do on Tuesday is know whether the campaign works, and acting on early numbers is the most reliable way to make a good account worse.
Week one
Clicks, impressions, sometimes sales. None of it is decision-grade.
The trap is arithmetic. In a week with 60 clicks and 4 conversions, the confidence interval around that conversion rate is wide enough to contain almost any conclusion you want. Pause a keyword on that evidence and you may have cut your best performer during a quiet week.
What week one is good for is catching what is broken. Are conversions firing. Does the page load on mobile. Are the search terms related to what you sell. Those are binary checks and need no sample size.
The volume a decision needs
Roughly 30 conversions per variant before comparing anything, and 50 or more before acting on a difference that is not dramatic. That converts directly into a budget question.
At LocaliQ and WordStream’s 2026 US average of $5.42 per click and 8.18% conversion rate, 30 conversions costs around $1,990 in media. In legal services, where they record a $9.87 click and a $131.63 cost per lead, the same 30 conversions runs closer to $3,950.
30 conversions at US averages: about $1,990. In legal services: closer to $3,950.
The same campaign structure returns a verdict in three weeks for one business and three months for another, because one vertical buys its evidence at a quarter of the price. Work out what yours costs before agreeing to a testing plan.
The automation floor
Automated bidding drives around 78% of Google Ads spend, per WordStream’s Q1 2026 data, and it learns from your conversion data. Below a certain volume it has nothing to learn from, so it approximates and results look erratic.
This gets misdiagnosed constantly. A small account switches to target-CPA, performance turns volatile, and the conclusion drawn is that the strategy does not work. What is happening is a model running on too little data, and the volatility is it guessing.
The sequence that avoids it: manual or maximize-clicks while volume builds, consolidated structure so volume pools rather than fragments, then automated bidding once there is real history. A tidy account with fourteen tightly themed campaigns has divided its learning data fourteen ways.
| Small accounts get made worse by good practices borrowed from large ones. Granular structure is correct at scale and actively harmful under a few hundred monthly conversions, because you have split one usable data set into a dozen unusable ones.— Vishal Singh, Performance Marketing Specialist, QlikMatrix |
What slows accounts down
Three things, in rough order of frequency in US mid-market accounts.
Conversion definitions counting the wrong action. A conversion firing on page load rather than form submission inflates volume and teaches the algorithm to buy traffic that does nothing. It produces the account that looks excellent in-platform and generates no pipeline, which clients notice last because the reports look fine.
Landing pages arriving after the campaign, so money goes into teaching the platform about a conversion path that is about to change. When the page ships, the learning resets.
Budget changed weekly. Every substantial change disrupts accumulated learning, so an account whose budget moves every Monday never builds enough stable history to be optimized. The constant adjustment feels like active management while functioning as sabotage.
Why costs rose and returns held
The market got more expensive and more efficient at once, which is unusual.
WordStream’s 2026 US average of $5.42 per click is more than double the $2.32 recorded when these benchmarks began in 2016. Over the same period cost per lead fell for the first time in five years to $66.69, with conversion rate improving in 87% of industries.
Clicks cost 2.3x what they did in 2016. Cost per lead just fell for the first time in five years.
The conclusion is specific: returns in search are now defended on the conversion side, not the cost side. You are not buying your way back to 2016 click prices. Conversion rate is what remains available, and it is owned by the landing page, the offer, and the tracking rather than by the bid.
Which reframes what a search marketing firm is paid for. Bidding is largely automated and click price is set by an auction you do not control. The work that remains sits on your side of the click, and any engagement scoped entirely inside the ad platform has excluded the part of the account where the leverage lives.
What clients ask about timelines
Q1. How long before I can judge the campaign?
Four to six weeks in a normal-cost vertical at adequate budget, longer if clicks are expensive. The better question is how many conversions you will have by then. Work backward from 30 per variant.
Q2. Why did performance get worse after we made changes?
Most likely the learning reset. Changes to budget, bidding, or conversion actions restart optimization, and recovery looks like decline. The instinct to change something else during the dip usually deepens it.
Written by Vishal Singh, Performance Marketing Specialist, QlikMatrix
Most of my week is inside search accounts where someone is trying to draw a conclusion from a fortnight of data. The conversion-volume math above is the calculation I run before agreeing to any test plan.
