Agency pricing models compared

Owners tend to compare quotes on the number. The number is the least useful part. Two quotes can carry the same monthly figure and mean entirely different things about what happens in month five when you need something unplanned.

I think about vendor pricing the same way I thought about food contracts. The unit price mattered less than what the contract did to behavior over a year. A cheap case price with a delivery minimum that forced me to over-order was not cheap.

What are the four structures?

Most real engagements are hybrids. A common and reasonable shape is a fixed-price first phase for the one-time work, then a smaller retainer for the recurring part. If a quote bundles both into one flat monthly number, ask which portion is setup, because you will keep paying for it long after it is done.

The structure also decides what happens when the work turns out to be bigger than anyone thought, which in this field is often. That is the test to apply to each row above. Not what it costs. What it does the first time somebody is wrong.

Which structure puts the risk on me?

Hourly, straightforwardly. You pay for the hours whether or not they produced anything, and estimates slip.

That is not a reason to avoid it. It is a reason to cap it. Ask for a not-to-exceed figure per month and a note when the work is approaching it. Any competent shop does this without complaint.

Retainers hide the risk better. If the scope is written as activities rather than outputs, you are paying for presence, and presence is exactly what erodes when the agency gets busy elsewhere. This is the single most common way a good engagement goes soft, and the signs that an agency has quietly stopped doing the work are visible in the reporting long before anyone says anything.

Fixed-price projects push the risk onto the agency, which is why the price includes a cushion. That cushion is worth paying. What you get in return is a number that does not move and a defined finish.

The cushion also has an edge case, and it is the one to watch. A fixed price on genuinely unknown work pushes the agency to define the finish line narrowly, because that is the only lever it has left. A migration quoted fixed will be scoped tightly around the pages that were listed. The pages nobody listed become a change order.

When is hourly actually the right answer?

More often than people think.

If you have someone capable in your own team and you need a specialist for specific problems, hourly is the efficient purchase. You are buying judgment in small amounts rather than a relationship.

It also fits diagnostic work, where nobody can scope the fix until they have looked. Reading server logs to see which AI crawlers are hitting your site is a good example, because the answer to what is actually fetching your pages lives in the logs and takes an hour to establish. Pricing that as a monthly program before anyone has looked is backwards.

And hourly is right for oversight. Paying a second set of eyes for two hours a month to check another vendor's work is one of the better small purchases available to an owner.

Where hourly breaks down is anything that needs to happen every week for a year. Nobody wants to file a timesheet against a fifteen minute task, so the small recurring work quietly stops being done. If your need is cadence rather than expertise, hourly is the wrong instrument no matter how fair the rate looks.

What about performance pricing and spend-based fees?

Both are legitimate. Both have a pull you should understand before agreeing.

Performance pricing, where you pay per lead or per booked job, sounds like the perfect alignment. The complications are practical. Someone has to define what counts as a lead, someone has to arbitrate disputed ones, and you need call tracking clean enough to settle an argument. Where that infrastructure exists, this model works well. Where it does not, you will spend your Fridays disputing line items.

A fee set as a share of ad spend pushes toward spending more. That is the obvious objection, and it is only half true, since managing a larger budget genuinely is more work. The real problem is that it stops telling you anything about the quality of management. A flat management fee, or a fee tied to a defined set of tasks, keeps the two conversations separate.

Performance pricing on search visibility has a harder problem underneath it, and it is measurement. If a proposal offers to charge you on AI search visibility, ask what the measurement is. Research from the University of St. Gallen, published in April 2026, ran identical prompts across four AI engines repeatedly and found that two consecutive days of the same prompt shared only 34% to 42% of their cited sources. Run the same prompt several times on the same day and source overlap fell to 32% to 43%. ChatGPT was the least stable engine measured. A fee that pays out on a number that moves that much on its own is not a performance fee, it is a coin toss with an invoice attached. The same paper's prescription is worth borrowing as a contract term: at least seven runs per prompt per day, reported on a two to four week rolling window, never week over week.

Is there a standard rate I should be comparing against?

No, and the confident answer you will find online is the problem.

Search for what an agency should charge and you will find the same figures everywhere: a percentage of ad spend described as the industry standard, a monthly floor for small accounts, a build range for a website. They read like benchmarks. They are not. Every carrier of those Google Ads management fee ranges is an agency blog citing another agency blog, and no survey with a disclosed sample sits behind any of them. The same is true of small business website build and maintenance ranges, which trace to agency pricing pages rather than to research.

That matters more than it sounds. Clutch surveyed 406 US small business owners in August 2025 and found 83% now have a website, with 41% using a builder such as Wix or Squarespace. Real survey work is being done on what small businesses own. None is being done on what they pay agencies for it.

What does have a published benchmark is the media, not the management. LocaliQ's 2026 search advertising benchmarks, drawn from thousands of US search campaigns, put the median cost per click across all industries at $5.42, the median conversion rate at 8.18%, and the median cost per lead at $66.69. By category the spread is wide: Restaurants and Food at $2.05 per click and $30.57 per lead, Dentists at $8.00 and $72.97, Home and Home Improvement at $8.33 and $90.92. LocaliQ sells advertising services, and the 2026 edition does not publish its sample size, so treat the figures as directional. The 2025 edition of the same series disclosed 16,446 US campaigns and noted that its averages are technically medians.

So you can benchmark what a lead costs on the platform. You cannot benchmark what an agency charges to get it. When someone tells you their fee is standard, the honest reply is to ask which survey.

How do I compare two quotes with different structures?

Convert both to the same three answers.

What gets produced in a month, counted in units. Not activities. Pages, listings, campaigns, reports, calls. If a quote resists being expressed this way, that is information, and the checkable contents of a monthly retainer are a short and specific list.

What it costs to stop, including notice period, and what you keep.

What you have to supply. Hours of your time, approvals, photos, access. A cheaper quote that needs six hours of your week is not cheaper.

Then add the one-time work you know is coming. If you are opening a second location or relocating, that is project work with its own scope, since a move creates ranking problems that do not resolve on their own and no retainer covers it by default.

The arithmetic that makes any fee comparable

Here is the calculation to run before you compare anything, and it works on all four structures because it ignores the structure entirely.

The National Restaurant Association's July 2026 analysis describes the cost shape of a typical independent restaurant before the pandemic: about 33 cents of every sales dollar to food, about 33 cents to labor, about 29% to everything else, leaving a pre-tax margin near 5%.

At a 5% margin, one dollar of profit requires about twenty dollars of sales.

So take whatever monthly figure you are being quoted and multiply it by twenty. That is the incremental sales the engagement has to produce, every month, before it has broken even at restaurant economics. Divide that by your average check to get covers. Divide by the days you open to get covers per day.

Most owners run that and find a smaller number than they feared. Some find it is four tables a night in a room already full on the nights that matter, which is the more useful outcome, because it says the constraint is capacity rather than demand.

Two honest limits. Your margin is probably not 5%. The Association's analysis says total expenses for an average restaurant rose 36% between 2019 and 2026 and that 42% of operators reported no profit in 2025. And a dental practice or an HVAC company carries a completely different margin, which changes the multiplier entirely. Use your own margin. The method transfers, the number does not. The Association also states plainly that its figures are management tools for gauging performance, not standards or goals.

And agree the communication rhythm at the same time as the fee, because price and contact frequency get negotiated together or not at all. How often you should expect to hear from an agency is part of the deal you are signing.

What to do this week

Take whatever quotes you have and rewrite each one on a single page in your own words. Three headings: what I get monthly, what it costs me to leave, what I have to do.

Anything you cannot fill in becomes an email. Send it. How fast and how plainly it comes back is part of the evaluation.

Run the multiply-by-twenty arithmetic with your own margin, not mine. Write the answer in covers, jobs, or appointments per day rather than dollars, because that is the unit you can judge.

Then decide which structure fits your year. Predictable, ongoing work suits a retainer. A defined problem suits a project. An unclear problem suits a few hours of diagnosis before you buy anything larger. Structure follows certainty, not preference.

Questions about a quote structure you were given?

Email me at eric@seod.com and tell me only the structure you have been offered, retainer or hourly or project or performance, plus what is in scope. No dollar figures needed and I would rather not see them. I will tell you which risks that structure puts on you and which two clauses would balance it. No pitch, and nothing sent afterward.

There is more on hiring and working with agencies if you are still comparing.