The question is not which model is better
It is which model matches the shape of your problem. A problem with a definable end belongs in a project. A problem that regenerates every week belongs in a retainer. A problem that consists of occasional decisions belongs in hourly counsel. Buy the wrong shape and the engagement fails in a way that looks like a people problem and is actually a contract problem.
That failure is predictable enough to describe in advance. Bounded work bought as a retainer becomes maintenance around month four, when the real work is finished and the invoice is not. Continuous work bought as a project produces a document and no change, because nobody owns what happens next. And diagnosis bought hourly puts you and the provider on opposite sides of every hour spent understanding your site.
For reference on what is actually offered: in the largest published survey of search providers, 78.2% offered a monthly retainer, 48.9% offered per-project work and 34.8% offered hourly. Respondents offered more than one, so the figures exceed a hundred. The retainer's dominance is a fact about how the industry prefers to sell, and not evidence that it is the right purchase for you.
What a retainer actually buys
Continuity, capacity and accumulated context. The provider holds your situation in their head, notices things between scheduled reviews, and can act on something in week two rather than after a scoping call. On a site that changes constantly — a catalog with daily inventory movement, a publishing operation, an advertising account with live budget — that context is most of the value, and rebuilding it every quarter would cost more than retaining it.
What it does not buy, though it is often sold this way, is a fixed quantity of output. A month in which the correct action was to change very little is not a month wasted, and a provider who fills it with activity to justify the invoice is doing you harm at your own expense.
The honest structure has three properties: a stated minimum useful term, because search work has a lag and judging it after six weeks measures nothing; review points where either side can stop without penalty; and a written record of what changed and when, which is the only thing that makes later attribution possible at all.
When a retainer is the right purchase, and when it quietly stops being one
It fits when the problem regenerates. New products, new pages, new campaigns, a competitive set that moves, an advertising account where yesterday's search terms create today's work. It also fits when the constraint is capacity rather than knowledge, and when the site's own release cycle means work ships continuously rather than in a burst.
The moment it stops fitting is identifiable, and almost nobody names it out loud. Once the diagnosed problems are fixed and the remaining work is routine, the engagement has become maintenance. Maintenance is a legitimate purchase, but it is a smaller one, and it should be repriced rather than continued at the diagnostic rate. If you cannot get a straight answer to the question of what is left to do that only this provider can do, that answer has arrived.
Build the check in at the start. A review point at a stated interval, at which the question is not whether the provider is working hard but whether the remaining work still needs this level of person, prevents the drift that turns a good engagement into a subscription.
When a project is the right purchase
When the problem has an end. Migrations end. Audits end. A penalty determination and remediation ends. An international annotation rebuild ends. Each has a definable scope, a definable artifact and a point at which it is either done or demonstrably not.
Projects also suit situations where trust has not been established. A fixed scope with a defined deliverable is a cheap way to find out how someone works before committing to a standing arrangement, and any provider who will not sell one is telling you something about their business model. The reverse question is worth asking too: a project that cannot be described in terms of what will exist at the end is not a project, it is a retainer with a deadline attached.
The risk to manage is the handover. A project that ends with a document and no owner for what happens next is the most common way search money gets wasted, and it is avoided by deciding in advance who implements, on what timetable, and who verifies that the change behaves as specified.
When hourly works, and when it works against you
Hourly suits counsel. A decision needs a specialist opinion, a proposal needs reading, a developer needs twenty minutes to unblock an implementation question. In those cases the unit matches the work and nobody is guessing at scope.
Hourly works against you in diagnosis, and the reason is structural rather than moral. Diagnosis requires an unpredictable amount of time spent understanding a site before anything useful can be said. Under an hourly arrangement, you are paying by the minute for the provider to learn your situation, and you will feel the meter running during exactly the phase where they should be following an unexpected thread. Both sides then optimize against the arrangement: the buyer rations the questions, the provider stops at the first plausible answer.
It also fails at implementation oversight, because the work that matters most — noticing that a shipped change does not behave as specified — happens intermittently and cannot be scheduled into billable blocks.
Performance-based arrangements, honestly
Paying only for results sounds like the arrangement that removes all risk from the buyer. It is worth understanding why it remains uncommon in serious work.
There is no credible published measurement of how prevalent performance-based search pricing actually is, so treat any claim that it is standard as an assertion. More importantly, three structural problems apply regardless of prevalence. Attribution is contested, so both sides will end up arguing about what caused a change, usually at the worst possible moment. The incentive tilts toward outcomes that are easy to measure rather than valuable, which is how you end up ranking for phrases nobody buys from. And the thing being priced cannot be committed to: Google states plainly that there is "no guarantee that changes you make to your website will result in noticeable impact in search results," which makes a fee contingent on that impact a bet rather than a service.
Where something like it can work is narrow and worth naming: a bounded outcome that is unambiguous and verifiable by both parties, such as a specific technical state being reached, or a manual action being lifted. Those are milestones, not performance pricing, and they are better written as project stages.
Mixing models, which is what usually actually happens
Most sensible arrangements are not pure. A diagnostic project first, because nobody should retain a provider to solve a problem neither party has defined. Then either a scoped implementation project, or a retainer if the work turns out to be continuous. Then, once the site is stable, a smaller standing arrangement for counsel — which is a different and cheaper thing than delivery.
Sequencing that way has a commercial advantage beyond the fit. Each stage is a real test of the provider before more is committed, and each stage produces something you keep regardless of whether the relationship continues. A provider who insists on a long retainer before any diagnosis exists is asking you to take the entire risk at the point where you know least.
How to choose in one pass
Four questions settle it, in this order.
- Does the problem have an end? If yes, project. If it regenerates, retainer. If you are not sure, buy a diagnostic project and find out, because that uncertainty is itself the thing to resolve first.
- Who implements? If your team does, you are buying diagnosis and specification, which is smaller and more project-shaped. If the provider does, it is larger and more retainer-shaped.
- How often will decisions arise? Frequently and unpredictably means a standing arrangement. Occasionally means hourly counsel, and buying a retainer for that is buying availability you will not use.
- What will you look at to judge it? Agree that before signing, whichever model you pick. The model determines what is reasonable to expect: a project is judged on whether the artifact was correct and the change shipped, a retainer on whether the site's trajectory moved over a period long enough to mean anything, and hourly counsel on whether the decisions it informed turned out well.
If a provider cannot answer the fourth question in your terms rather than theirs, the model is not the problem.
Frequently Asked Questions
Should I hire an SEO on retainer or per project?
Match the model to the shape of the problem rather than to a preference. If the work has a definable end — a migration, an audit, a penalty determination, an international rebuild — buy a project, because it finishes and produces an artifact you keep. If the work regenerates weekly, as it does on a large catalog, a publishing operation or a live advertising account, a retainer fits because most of the value is accumulated context. If you genuinely cannot tell, buy a diagnostic project first; resolving that uncertainty is what it is for.How long should an SEO retainer run?
Long enough that the work has had time to be read by a search engine, and no longer than that by default. Google states that some changes take effect in days while others can take several months, and offers no guarantee of an outcome, so judging a retainer after six weeks measures nothing. The better structure is a stated minimum useful term, then review points where either side can stop, rather than a fixed twelve-month commitment. Long lock-ins mainly protect the provider's forecasting.Is hourly SEO consulting a good idea?
For counsel, yes: a decision to be reviewed, a proposal to be read, a developer to be unblocked. For diagnosis, it works against both sides. Understanding a site takes an unpredictable amount of time before anything useful can be said, and an hourly meter makes the buyer ration questions and the provider stop at the first plausible answer. It is also poor for implementation oversight, where the valuable work is noticing intermittently that a shipped change does not behave as specified.What should a retainer include so it does not become maintenance?
A written record of what changed and when, review points at stated intervals, and an agreed measure set before the work starts. The specific check that matters is asking, at each review, what remains that genuinely needs this level of person. Once the diagnosed problems are fixed and the remaining work is routine, the engagement has become maintenance — which is a legitimate purchase at a smaller size, and should be repriced rather than continued unchanged.Can a retainer be stopped, and what happens to the work?
It should be stoppable at review points without penalty, and what you keep should be settled before you start rather than negotiated on the way out. At minimum that means the documentation, specifications and change records produced during the engagement, plus administrative ownership of your own accounts and properties. A provider holding your advertising account, analytics property or Search Console access as their own asset is a problem to solve at the beginning, when it is a paragraph, not at the end, when it has become a bargaining chip.Is performance-based SEO pricing legitimate?
It is legitimate in principle and awkward in practice, and there is no credible published measurement of how common it actually is. Three problems recur: attribution is contested, so both parties will eventually disagree about cause; the incentive favors outcomes that are easy to measure over outcomes that are valuable; and Google states there is no guarantee that changes to a site produce noticeable impact, which is precisely what the fee is being staked on. Milestone-based project stages achieve most of the intent without those defects.Can a project turn into a retainer later?
That is the usual and sensible sequence: a diagnostic project first, then implementation, then a standing arrangement only if the work turns out to be continuous. Each stage tests the provider before more is committed and leaves you with something you keep either way. Be alert to the reverse pressure — a provider whose diagnostic project concludes, without much argument, that a long retainer is the only possible next step. The finding may be right, but the incentive deserves noting.Published