The first 90 days of an SEO engagement are unusually heavy on discovery. Teams validate access, crawl the site, repair measurement, map search demand and establish a baseline. After that, the campaign should stop behaving like a prolonged audit and become a repeatable decision system.
The strategy after day 90 is not “publish more.” It is to use early evidence to decide which constraint deserves the next unit of engineering, editorial and promotion capacity. That shift separates an operating campaign from a stack of recommendations.
Use a post-90-day decision loop
This original SEO Companies Hub contract converts setup evidence into a repeatable operating cycle.
| Gate | Required decision | Evidence |
|---|---|---|
| Setup closure | Implemented, accepted-risk or still-blocking status for each foundation | Real URL/system verification, not audit delivery |
| Cohort | Bounded changed pages/templates and untouched comparison where possible | Version, release date and included URL list |
| Bottleneck | One dominant constraint for the next capacity unit | Discovery, indexing, relevance, quality, authority, conversion or measurement diagnosis |
| Intervention | Specific change, owner, dependency and acceptance rule | Ready brief and implementation commitment |
| Observation | Leading search signal, task outcome and guardrail windows | Search, site and business systems kept distinct |
| Decision | Scale, modify, hold, rollback or stop | Predeclared threshold plus documented uncertainty |
The cycle repeats by portfolio cohort; a quarterly calendar alone does not make work strategic.
Close the setup phase explicitly
Before changing cadence, confirm that the initial controls are actually working. A credible day-90 review should answer:
- Can Google discover, crawl and index the intended page set?
- Do canonical URLs, redirects and status codes match the desired architecture?
- Are Search Console and analytics collecting usable data?
- Can the team trace priority conversions beyond a form submission?
- Does every target query group have one intended page or a documented gap?
- Are owners and acceptance criteria assigned for outstanding work?
Do not declare setup complete because an audit document was delivered. The gate is implementation or an explicit business decision to accept the risk. Unresolved critical defects should remain visible in the roadmap.
Google's SEO Starter Guide emphasizes organizing a logical site, using descriptive URLs, writing useful content and helping search engines understand pages. Those fundamentals remain part of the operating system after the audit. They are not one-time boxes.
Replace the baseline with cohorts
The opening baseline usually describes the whole site. It is useful for orientation but weak for deciding what to do next. After 90 days, track cohorts tied to actual interventions.
A cohort might include pages published in April, product pages whose internal links changed, templates that received structured-data fixes or articles refreshed after losing clicks. Record the change date, included URLs, expected leading indicator and business outcome.
This allows a fairer comparison than a sitewide chart. A new content cohort should not be judged against ten-year-old branded pages. A technical change should not receive credit for demand generated by a seasonal campaign elsewhere on the site.
Keep untouched comparison groups where practical. SEO is not a laboratory, but a stable comparison makes conclusions less reckless.
Choose the current bottleneck
Every quarterly cycle should name the dominant constraint. Common categories include:
- Discovery: important URLs lack internal links or sitemap coverage.
- Indexing: duplicates, canonical errors or low-value page sets obscure the intended inventory.
- Relevance: no page fully answers the query and decision task.
- Quality: the page is shallow, outdated or unsupported.
- Authority: useful pages lack independent awareness and references.
- Conversion: qualified visitors arrive but cannot complete the next step.
- Measurement: teams cannot connect search behavior to business outcomes.
Select work that attacks the named bottleneck. Publishing ten articles does little when the template accidentally canonicalizes them to the category page. A redesign is not the priority when the actual gap is missing comparison content.
Move to a portfolio roadmap
After day 90, maintain a balanced portfolio instead of a single backlog. A practical capacity model reserves effort for four streams:
- technical health and incident response;
- existing-page improvement;
- new demand coverage;
- measurement and learning.
Authority building and conversion work can be explicit streams when they require separate teams. The exact percentages should change with evidence. A migration quarter may be technically heavy; a stable site with weak category coverage may direct more effort to content.
Every roadmap item needs a hypothesis, affected page set, owner, dependency, risk, acceptance test and review date. “Improve internal linking” is an activity. “Add contextual links from 20 indexed guides to five orphaned service pages and confirm discovery in a fresh crawl” is controllable work.
Update the keyword map with real impressions
Early keyword maps depend on research tools and judgment. Once pages accumulate Search Console data, refine them with observed queries and page behavior.
Look for:
- queries where the intended page earns impressions but a weak click-through rate;
- multiple URLs alternating for the same intent;
- unexpected queries that reveal a missing section or separate task;
- high-visibility pages whose visitors rarely reach a meaningful next step;
- demand that the initial research underestimated.
Search Console reports search performance while analytics records on-site behavior. Google explicitly notes that the systems use different metrics and will not match perfectly. Use the connection to follow a decision path, not to force identical totals.
Change the content meeting
During setup, a content meeting often focuses on briefs and calendars. A mature campaign reviews evidence before assigning production.
For each proposed page, ask:
- What distinct search task does it satisfy?
- Which existing URL is closest, and why is expansion or consolidation insufficient?
- What original experience, data or expert input will improve the answer?
- What page should the reader visit next?
- Who will keep the claims current?
For existing content, diagnose the reason for change before rewriting. Lost clicks can reflect demand, SERP layout, technical problems, stronger competitors or stale information. A general rewrite may erase a page's strongest elements without fixing the cause.
Establish an implementation service level
Campaigns often stall between recommendation and release. Measure that delay.
Track median time from approval to implementation, the percentage of accepted recommendations released by the due date and the number blocked by each dependency. Segment high-risk technical changes from ordinary editorial updates so one does not distort the other.
If the client owns development, reserve sprint capacity and nominate a technical approver. If the agency implements, define repository access, QA environments, rollback requirements and production authority. Ambiguity here is a governance problem, not an SEO mystery.
Review outcomes at three speeds
Different decisions need different cadences.
Weekly reviews handle crawl incidents, broken releases, tracking failures and execution blockers. Monthly reviews examine cohorts, query patterns, conversions and completed experiments. Quarterly reviews reconsider the market, page architecture, resource allocation and the core bottleneck.
Avoid changing the strategy every week because rankings moved. Google advises that some changes can take hours while others may take months to be reflected. Record deployment dates and allow a sensible observation window.
Set scale and stop rules
Every material tactic needs a rule for expansion and a rule for stopping.
For example, a six-page comparison cohort might expand if pages are indexed, earn relevant impressions, attract qualified journeys and produce evidence of sales use. It might pause if the pages overlap, cannot obtain credible product data or attract only irrelevant queries.
Stop rules protect the budget from sunk-cost thinking. They also give experiments integrity: a team cannot quietly redefine success after a weak result.
Keep executive reporting decision-led
An executive review should state what changed, what was learned, what decision is required and what business risk remains. Rankings and traffic are supporting evidence, not the entire narrative.
A useful one-page summary contains:
- outcomes for the active cohorts;
- contribution to qualified leads, transactions or assisted pipeline;
- major technical or governance risks;
- completed and delayed roadmap commitments;
- the next bottleneck and proposed capacity allocation;
- decisions that require client approval.
Preserve detailed query, page and crawl evidence in an appendix or dashboard. The summary should make action easier without hiding uncertainty.
The post-90-day operating model
After the first 90 days, an SEO campaign should run as a cycle: observe, diagnose, prioritize, implement, verify and learn. Each cycle improves both the site and the team's ability to make the next decision.
If the agency is still producing audits without verified releases, or publishing content without cohort-level learning, the engagement has not matured. The goal is a controlled system that turns search evidence into durable product, content and customer-acquisition improvements.
Related decisions
- What Should Happen in the First 90 Days With an SEO Agency? — the adjacent hiring an seo agency decision.
- How to Turn Business Goals Into an SEO Roadmap — the adjacent seo strategy decision.
- How to Sequence Technical SEO, Content and Digital PR — the adjacent seo strategy decision.