How to Run a 90 Day SEO Agency Pilot Before Signing a Retainer
The safest way to hire an SEO agency is to buy a 90 day paid pilot before you commit to a 12 month retainer. A pilot lets you test the agency's process, communication, and technical judgment on a small scope while the financial exposure stays low. OnyxRank offers this kind of structured start for ecommerce and SaaS teams, but the framework below works with any agency, and you can use it to hold us to the same standard.
Most founders do the opposite. They sit through a polished sales call, compare three proposals that all promise "growth," and sign a year of fees on trust. Then month four arrives, traffic is flat, and the agency explains that SEO takes time. Sometimes that is true. Often it is a sign nobody agreed on what would be visible by month four.
Why a pilot beats a proposal comparison
A proposal tells you what an agency says it will do. A pilot shows you what it actually does with your site, your developers, and your approval process.
Google's own guidance on [hiring an SEO](https://developers.google.com/search/docs/fundamentals/do-i-need-seo) suggests asking prospective SEOs for examples of previous work, what results they expect in what timeframe, how they measure success, and whether they will share every change they make with reasoning. A pilot turns those questions from talk into evidence. You see the change log, the reasoning, and the measurement in your own account.
There is also a timing problem that a pilot respects honestly. Ahrefs analyzed a large sample of pages and found that [only 1.74% of newly published pages reached the top 10 within a year](https://ahrefs.com/blog/how-long-does-it-take-to-rank/), while the average page ranking number one was five years old. That study measured new URLs from a crawl sample, not agency client work, so it does not prove how your campaign will go. What it does show is that ranking gains on competitive terms are slow, and that you should not judge a pilot by whether it hits position one in 90 days.
That is the key design principle. A good pilot does not measure rankings alone. It measures leading indicators the agency can actually control, and it checks the quality of the work itself.
What to measure in 90 days (and what not to)
Split pilot metrics into three groups.
Group one: work quality (fully in the agency's control)
- A written technical audit with prioritized findings, each tied to evidence from your crawl, Search Console, or server logs
- A change log listing every recommendation, who owns it, its status, and the reasoning
- Implemented fixes that you can verify yourself, such as corrected canonical tags, cleaned redirect chains, or valid structured data
- Content briefs that cite real sources and reflect what you can credibly say, not generic outlines
Group two: leading indicators (influenced, not controlled)
- Indexed page count and crawl coverage moving in the right direction in [Search Console](https://developers.google.com/search/docs/monitor-debug/search-console-start)
- Impressions and average position for a defined set of target queries
- For AI search visibility, whether your pages are cited for a fixed panel of prompts you agree on up front
Group three: lagging outcomes (do not hold the agency to these at day 90)
- Organic revenue, signups, or qualified leads
- Top 3 rankings on competitive head terms
You can track these, but treat them as context. Penalizing an agency for outcomes that depend on a six month index and ranking cycle will push good agencies away and attract the ones willing to promise anything. Google says plainly that [no one can guarantee a number one ranking](https://developers.google.com/search/docs/fundamentals/do-i-need-seo). A pilot scored on guaranteed rankings selects for exactly the behavior you want to avoid.
The 90 day pilot structure
Here is a simple three phase structure you can copy.
Days 1 to 30: audit and baseline
The agency delivers a technical and content audit and agrees a baseline with you. The baseline includes current clicks, impressions, indexed pages, and the AI citation panel. Without a baseline, every later number is an argument instead of a fact.
Checkpoint question: could someone on your team explain the top five findings in their own words? If the audit reads like a tool export with a logo on it, treat that as data. Our own [AI SEO audit checklist](/blog/ai-seo-audit-checklist-diy-2026) shows what a useful audit should cover, so you can compare.
Days 31 to 60: implement the highest leverage fixes
Pick no more than five fixes. For an ecommerce store, that might be faceted navigation crawl waste, product structured data, and thin category pages. For a SaaS company, it might be pricing and integration pages that answer comparison queries, plus fixing rendering issues that hide content from crawlers.
The point is not to fix everything. It is to see how the agency handles a real implementation: how it writes tickets for your developers, how it follows up, and whether it verifies the fix afterward.
Days 61 to 90: prove one content or visibility win
Ship a small, well sourced set of pages, or fix a defined group of existing ones, and measure early movement against the baseline. Keep the volume low. Google's [spam policies on scaled content abuse](https://developers.google.com/search/docs/essentials/spam-policies) target producing many pages primarily to manipulate rankings without adding value. A pilot that publishes fifty AI generated pages in month three is a risk to your domain, not a proof point.
At day 90 you hold a decision meeting with a written scorecard.
Set access rules before the pilot starts
A pilot is also a security test. Agree these rules in writing:
1. **Ownership.** Your company owns the Search Console, analytics, and tag manager properties. The agency is added as a user, never the owner. Our guide on [data ownership at offboarding](/blog/seo-agency-offboarding-data-ownership-2026) explains why this matters when you leave.
2. **Least privilege.** Google advises granting only read access to Search Console when you want an audit. Start there and expand only when implementation requires it.
3. **Change approval.** Nothing ships to production without a ticket and a named approver on your side.
4. **Disclosure.** The agency tells you which people and tools touch your account, including any subcontractors. See [how to find out who is working on your account](/blog/ai-seo-agency-staffing-whos-working-on-your-account-2026).
5. **No link buying.** Any outreach or link acquisition method is disclosed in advance. Paid links that pass ranking signals violate Google's spam policies.
How to price a pilot fairly
Agencies will quote a pilot differently, so use principles instead of a target number.
- **Pay for the work, not a discount.** A heavily discounted pilot usually means the agency expects to recover cost in the retainer. A fair paid pilot lets either side walk away cleanly.
- **Fixed scope, fixed fee.** Define deliverables, not hours. An hourly pilot with no cap is a retainer in disguise.
- **Credit option.** Many buyers negotiate that some part of the pilot fee is credited toward the first retainer months if they continue. That is a reasonable ask, not a guarantee any agency owes you.
- **No auto renewal.** The contract ends at day 90 unless both sides sign again. Compare this against the [contract clauses we recommend checking](/blog/seo-agency-contract-clauses-2026).
If you are weighing a pilot against paying only for results, our breakdown of [performance SEO versus a retainer](/blog/performance-seo-agency-vs-retainer-2026) covers the tradeoffs.
The day 90 scorecard
Score each line from 0 to 2: 0 is missed, 1 is partial, 2 is fully met. A total under 60% means do not renew.
| Area | What a 2 looks like |
|---|
|---|---|
| Audit quality | Findings tied to your own data, prioritized, understood by your team |
|---|---|
| Communication | Questions answered within an agreed window, no surprises |
| Implementation | At least three fixes shipped and verified in the live site |
| Measurement | Baseline agreed, reports reproducible from your own accounts |
| Judgment | Agency pushed back on at least one bad idea, including its own or yours |
| Risk | No tactics that conflict with Google's published spam policies |
| Fit | Your developers and the agency's team work well together |
The "judgment" row matters more than it seems. An agency that agrees with every request is not exercising expertise. Google's guidance on [AI features in Search](https://developers.google.com/search/docs/appearance/ai-features) says standard SEO best practices remain relevant for AI Overviews and AI Mode, and that there are no special extra requirements. An agency selling secret AI tricks that contradict that is a signal to ask harder questions.
When a pilot is the wrong move
Be honest about the limits. A pilot is a poor fit when:
- Your site has no developer access, so no fix can ship in 90 days. Fix that first.
- You cannot track conversions. Without that, you cannot judge business impact later.
- The site is brand new. Results will take longer than any 90 day window can show, and the pilot should focus on foundations only.
- You need a large migration handled urgently. In that case, hire for a defined project with its own timeline.
FAQ
What is an SEO agency pilot?
A pilot is a short, paid, fixed scope engagement, usually around 90 days, used to evaluate an agency's process and work quality before signing a long retainer. It includes a baseline, a small set of implemented fixes, and a written decision point.
How long should an SEO trial last?
Ninety days is enough to judge audit quality, implementation discipline, and early indicators. It is not enough to judge revenue results. Ahrefs data shows most newly published pages do not reach the top 10 within a year, so avoid grading a trial on competitive rankings.
Should I ask an SEO agency to guarantee rankings?
No. Google states that no one can guarantee a number one ranking and advises finding someone else if an SEO promises it. Ask for a defined process, a measurement plan, and a timeframe for leading indicators instead.
What access should an agency get during a pilot?
Start with read access to Search Console and analytics, owned by your company. Add edit access only for specific tasks that require it, and keep a record of every change.
Can I run a pilot with more than one agency?
Yes, if the scopes do not overlap. Give each a separate section of the site or a separate workstream so you can attribute changes cleanly. Overlapping scopes make results impossible to read.
Key takeaways
- Buy a 90 day paid pilot before a 12 month retainer.
- Score work quality and leading indicators, not guaranteed rankings.
- Own your accounts, grant least privilege, and require a change log.
- Keep content volume low during the pilot to avoid scaled content risk.
- Decide at day 90 with a written scorecard, and do not auto renew.
If you want to see what a baseline and prioritized fix list looks like for your own store or product, run the [free OnyxRank audit](/free-audit). If you would rather compare scope and fees first, see our [pricing plans](/pricing).
Pro Intel subscribers get the full picture - proprietary analysis, keyword opportunities, tactical playbooks, and template downloads every week. $49/mo.
One email per week. Actionable, no fluff.