Google Ads Agency or Do It Yourself — The Honest Maths

Every page that ranks for "google ads agency" is either an agency selling itself or a directory selling leads to agencies. They all say the same three things: agencies save you time, agencies have experience, agencies get better ROAS. None of that is false. It is just incomplete, because none of them tell you what an agency does on a Tuesday, what it does not do at all, and at what account size the maths stops working.

Here is the version with numbers in it.

What you actually pay a Google Ads agency for

You pay for the decisions that happen a few times a year and move a lot of money: account structure, conversion tracking, bidding targets, creative, and feed quality. Not for daily account hygiene.

Run through what a competent agency earns its fee on:

• Conversion tracking that is actually right. Deduplicated, server-side where it matters, offline conversions imported from the CRM so Smart Bidding optimises to booked revenue instead of form fills. This is the single highest-leverage thing most accounts are missing, and it is genuinely hard.
• Account structure. How many campaigns. What deserves its own budget. When to pull a product line out of Performance Max and into Standard Shopping so you can see the search terms and use shared negative lists. When to consolidate instead of split.
• Bidding. Choosing between tCPA and tROAS, setting the target, knowing when a target is starving a campaign rather than disciplining it, resisting the urge to change it every four days.
• Creative and assets. RSA headlines that match the landing page, asset groups with images that do not look like stock photos, video that exists at all so PMax does not auto-generate something embarrassing.
• Shopping feed work. Titles, product types, custom labels for margin tiers. A feed fix often beats a month of bid tinkering.
• Accountability. Someone whose job it is to notice that last month was bad and explain why.

All of that is real work. Most of it is monthly or quarterly. None of it happens by reading a list of search terms.

What Google Ads agencies charge, and how the pricing models differ

Three pricing models dominate: a flat monthly retainer, a percentage of media spend, or a hybrid of both plus a setup fee. From what clients and prospects quote to us, flat retainers for small accounts land somewhere around $500–$2,500 a month, mid-market engagements around $2,500–$10,000, and percentage-of-spend deals typically 10–20% with a floor so the agency does not work for $90.

A few things that pricing page comparisons rarely mention:

Minimum spend requirements are the real filter. Plenty of agencies will not take an account under $5,000–$10,000 a month in media. Not snobbery — arithmetic. Which brings us to the second thing.

Agencies pay a fortune to find you. Keyword tools put the average cost per click on "google ads agency" near $100 in the United States. Ten clicks from a searcher who never replies is a thousand dollars. A handful of clicks costs more than a month of a small retainer. That is why the good ones want twelve-month contracts, minimum spends and onboarding fees, and why the cheap ones are cheap because they are running a template across forty accounts.

Percentage of spend has an incentive problem, and it is worth naming. If the fee is 15% of spend, the fee goes up when spend goes up — including when the extra spend is wasted. That is not fraud and most agencies behave well anyway. But it means the person paid to scale the account is not the person who feels the pain of $1,400 a month going to plumber salary, how to fix a leaking tap youtube and plumbing courses near me. Nobody's incentive is pointed at that money.

Freelancers are the underrated middle. A good freelancer at $500–$1,500 a month, or $60–$150 an hour, often does more per dollar than a junior at an agency, because you get the senior person directly. The risk is obvious: holidays, illness, and one client who suddenly needs everything.

The four options side by side

Four ways to run the account. Costs are monthly, time is weekly, and the last column is the one people forget to ask about.

Read the bottom two rows together. They are not competitors. An agency that does strategy well and touches search terms monthly, plus something that handles the daily sweep, is a sane combination. So is a skilled in-house person plus an agency for a quarterly structural review. The mistake is assuming any single row covers the whole table.

Doing it by hand: what the weekly pass really costs you

A proper search terms pass on one account takes 30 to 90 minutes a week, depending on spend and how many campaigns you run. It is free in cash and expensive in attention, and it is the first task that gets dropped when a client emails.

The workflow, concretely: open the search terms report, set the window to the last 7 or 14 days, filter to conversions = 0, sort by cost descending. Read down the list. For each term you decide three things — block or keep, which match type, and at what level.

That third decision is where hand-work gets slow. A phrase negative for salary kills plumber salary and plumbing salary uk in one line. But applied at account level it also kills salary sacrifice scheme plumber if that happens to be something you sell. Broad negatives behave unpredictably across match types, which is why negative keyword match types don't work like positives and why a rushed pass creates problems you find three months later.

Then there is the part nobody enjoys: Performance Max takes up to 10,000 campaign-level negative keywords and accepts no shared lists. So every negative you decide on has to be written into each PMax campaign separately. Five PMax campaigns means five copies of the same list, maintained by hand, against a hard ceiling — the detail behind negative keywords in Performance Max and the 10,000 cap.

Two honest limitations of the manual route:

The report does not show you everything. Google withholds low-volume and privacy-filtered terms, so the list you are reading is a subset of what you paid for, and the withheld portion is disproportionately made of one-click oddities. That is covered in the search terms report and what it hides.

And weekly is a compromise, not a standard. A term that starts spending on Monday spends all week before you see it. Doing it daily costs you five slots instead of one. Almost nobody sustains that across more than two or three accounts.

The script option: free, fast, and blind

A Google Ads script that adds any term over a cost threshold with zero conversions as an exact negative costs nothing and runs every night. It also cannot tell the difference between a bad query and a slow one.

The typical rule: cost over $40 in the last 60 days, zero conversions, add as exact negative. It catches the obvious offenders and it is better than nothing. Where it breaks:

Exact negatives only kill that exact string. Block plumber salary uk and you have done nothing about plumbers salary uk, plumber salaries, average plumber salary. Each variant has to independently cross the threshold before it gets blocked. You are always one step behind the long tail, which is exactly where wasted spend hides.

A rule cannot read intent. cheap crm and crm for 3 person team look identical to a threshold script. One is a tyre-kicker, one is your ideal customer. Only a human, or something trained to judge, separates them.

Thresholds cut both ways. Set it at $40 and a long sales cycle B2B term that would have converted in week nine gets killed in week three. Set it at $200 and you burn $200 per bad term before anything happens.

Undo is manual. Scripts write. They do not remember why, and they do not roll back cleanly when you realise the rule was wrong.

Scripts are a good fit for one specific reader: someone technical, running one or two accounts, who wants the worst 20% of waste gone and is happy to review the log monthly. They are a bad fit for an agency managing thirty accounts with different verticals, because the same threshold cannot be right for a $9 impulse product and a $40,000 service.

Why almost no agency reads the search terms report daily

It is not incompetence. Daily search term cleanup is unbillable, invisible, never finished, and gets an account manager no credit. So it becomes a monthly line on a checklist, or it quietly stops.

Think about how an agency day is structured. A manager holds eight to twenty accounts. The calendar is full of client calls, reporting, creative approvals and the campaign launch that was promised for Friday. Opening one account's search terms report and reading 400 lines takes an hour and produces a slide that says "added 37 negative keywords." The client does not applaud. Next month there are 400 more lines, because the report refills itself continuously — a term you have never seen can appear tomorrow and spend $80 by Thursday. Google shows irrelevant search terms as a structural consequence of how close-variant matching works, not as a bug that gets fixed once.

Meanwhile the same hour spent fixing conversion tracking or rewriting feed titles produces a visible result. The agency is rationally allocating attention. The account still leaks.

The categories that leak most are predictable. Job hunters: plumber jobs, apprenticeship, vacancies, salary — covered in job seekers clicking Google Ads. DIY and research intent: how to, tutorial, free template, pdf. Competitor names, which are a real decision rather than an obvious block — see competitor searches, block or keep. And in Shopping and PMax, whole product categories you do not stock, because Shopping negatives work differently and the query is matched against your feed rather than your keywords.

If you have an agency, there is a simple test. Ask for the change history filtered to negative keywords, last 90 days. You will see either a steady trickle of additions across all campaigns, or a burst of forty on the day before a quarterly review, or nothing. All three answers are informative.

What a negatives tool does and does not replace

A tool like PPCleaner reads the search terms your account paid for, judges which ones were never going to convert, and writes them in as negatives every night across Search, Shopping and Performance Max. That is the entire scope. It is a replacement for the weekly spreadsheet pass, not for a strategist.

What it does not do, plainly: it does not touch budgets, bids, ads, keywords, targeting or your feed. It will not tell you your tROAS is set too high, that your landing page loads in nine seconds, or that half your conversions are duplicated. If your account is losing money because the offer is wrong, no negative keyword saves it — that is why there is an order in which to check things when there are no conversions, and negatives are not step one.

It also cannot see what Google does not report. Search terms exist in campaign_search_term_view for Search, Shopping and Performance Max. Display, Video and App report none at all. If your waste is in a Display campaign, a search terms tool is looking at an empty room.

What it does handle that hand-work struggles with: the volume. Writing the same negative into six Performance Max campaigns because PMax refuses shared lists. Choosing phrase over exact so one line covers a family of variants without over-blocking. Keeping the list under control, since there is a point at which too many negatives starts costing you. And two safety rails that matter when software is writing to a live account: nothing that converted in the last 90 days is ever blocked, and every block is reversible in one click for 90 days.

The review burden is the honest cost. Five to fifteen minutes a week reading what was blocked, and clicking undo on the two or three you disagree with. If you will not spend that, you should not hand write access to any automation, ours included.

Where each option becomes the right one

Thresholds, stated as plainly as we can. These are judgement calls from the accounts we audit, not laws.

Under roughly $1,500 a month in media spend. Do it yourself, monthly. An agency retainer would exceed your ad budget, and a percentage-of-spend deal would hit the floor fee anyway. Spend one Saturday learning conversion tracking properly. One hour a month on search terms is proportionate at this size, because there simply are not that many of them.

$1,500 to $10,000 a month, one person in-house who enjoys this. DIY on structure, and automate the hygiene — script if you are technical and patient, tool if you are not. A freelancer for a one-off audit every six months is usually better value than a retainer. This is the band where Performance Max waste starts to be real money and where nobody is watching it yet.

$10,000 to $50,000 a month, or you are selling something with a complicated sales cycle. Get a freelancer or a small agency. The structural decisions now move more money than your time is worth, and getting tROAS or offline conversion imports wrong costs more per month than the retainer. Separately, decide in writing who owns the search terms report. Not "we'll handle optimisation." Who, how often, and how you will verify it.

$50,000 a month and up, or multiple countries, multiple accounts, an MCC. An agency is the right answer and the maths is not close. At that spend, a 5% improvement in efficiency pays four retainers. Hire for the strategist, not the dashboard. Then assume the daily hygiene still needs its own solution, because at that account count no human is reading every search terms report every day.

You already have an agency and you are reading this because something feels off. The question is not whether to fire them. It is whether the search terms pass is in anyone's calendar. Pull the change history. If negatives are being added weekly across all campaign types, your agency is in the minority that does this properly and you should keep them. If not, that is one conversation, not a breakup.

If you cannot answer these questions, hire someone. Should this be PMax or Standard Shopping? Is my tROAS target achievable at current conversion rates? Are my Shopping feed titles matching how people actually search? Why did conversions drop 30% on the 14th? A tool answers none of those. A good agency answers all four, and that is worth more than any amount of keyword blocking.

Five questions to ask an agency before you sign

Ask these instead of asking for case studies. Case studies are marketing; these are diagnostics.

1. "Show me the negative keyword change history on an account you manage, last 90 days." Redact the client name. You are looking for a steady rhythm, not a pre-review burst. If the answer is a description of their process rather than a screenshot, that is your answer.
2. "Who specifically works on my account, and how many other accounts do they hold?" The person in the pitch meeting is rarely the person in the account. Twenty accounts per manager is normal. Forty is a template shop.
3. "How do you handle Performance Max search terms?" A good answer mentions campaign-level negatives, the absence of shared lists, and the fact that PMax search term data is thinner than Search. A vague answer about "optimising asset groups" means they are not looking.
4. "What happens to the account if I leave in month four?" You want the account under your own Google Ads ID, with them as a linked manager. Not their account with your money in it. This single point causes more pain than pricing ever does.
5. "What would you refuse to do?" Anyone who says yes to everything will also say yes to scaling spend past the point where it works.

If an agency answers all five well, the retainer is probably the best money in your marketing budget. If they answer three well and shrug at the search terms question, that is a normal, competent agency — hire them for what they are good at, and put the nightly cleanup somewhere else.