SEO tools 2026: how to build an agency tool stack

An agency's list of SEO tools usually grows without a plan: a crawler is bought for one client, rank tracking is added for another project, and after a while subscriptions pile up that nobody can quite explain paying for. This article proposes a method that starts the choice of SEO tools not from individual products but from the work to be done. The goal is not to collect the most tools; it is to assign a data source to every question and an owner to every data source. Brand names appear in the examples, but this is not a ranking or a recommendation list; the same logic can be built with other products.

A tool stack starts from the work, not the products

When choosing SEO tools, the first question should not be "which tool is best" but "which decisions do we need to make". The decisions an agency makes regularly are more or less the same: which pages cannot be crawled or are not being indexed, which queries are gaining or losing visibility, which topics competitors are producing content on, which links have been won or lost, and how fast pages are for users. Write these decisions in a table and, next to each row, add the data source that supports that decision.

This table makes two things visible straight away. First, for some decisions more than one tool is doing the same job; that means unnecessary cost and conflicting reports. Second, some decisions have no data source at all; the team is working on guesswork there. The real value of building a stack is closing these two gaps.

First layer: Google's own data

The base of every stack is first-party data coming from the site itself. Search Console shows which queries earn impressions and clicks in Google Search, which pages are in the index and which problems have been detected. Google's own documentation positions the tool for monitoring search performance, checking indexing and receiving issue alerts. No third-party tool can reproduce this data one to one; at most it estimates it.

Page experience measurement sits in the same layer. PageSpeed Insights offers two kinds of data: field data from the Chrome User Experience Report and lab data simulated with Lighthouse. According to Google's explanation, lab data is useful for debugging, while field data reflects real user experience, and the two can contradict each other. We explained how we read this distinction in detail in our site speed test article. On the analytics side, no SEO report can be tied to business results without a setup that measures conversions.

This layer is cheap but labour-intensive: correct property verification, regular review of user access and data retention settings should be done in the first week. A property set up later does not bring back past data.

When setting up the first layer, think about the access structure from the start as well. The client's properties should live in the client's own account, with the agency given access as a user. A property opened in the agency's account becomes an asset that is hard to hand over when the engagement ends, and who owns the historical data becomes a matter of dispute.

Second layer: crawling and technical audits

The second layer is the crawlers that show how the site looks through a search engine's eyes. A desktop crawler can be enough for sites without many pages; for large or frequently changing sites, cloud-based and scheduled crawls are more practical. The criteria to look at when choosing are these: can it process content rendered with JavaScript, is it easy to export results, and can it compare two crawls and show what changed?

What matters in this layer is the habit of turning crawl output into a work list. A report with thousands of warnings moves nobody unless it is prioritised. In practice, producing a fix list of no more than a few items after each crawl and assigning it to an owner delivers more than long reports.

Third layer: market and competitor data

The third layer is the market tools, such as Semrush and Ahrefs, that maintain their own databases. These tools give information about keyword ideas, competitor visibility and link profiles. The point not to forget here is that this data is an estimate. Semrush's own help page also describes the organic traffic in its domain overview as "estimated". These numbers are valuable for finding direction and comparing; not for telling a client "your traffic is this much".

On the link profile side, we compared which tool shows what and the limits of free options in our backlink analysis article. Rank tracking needs its own decision too: if the keyword set, location and device settings are not fixed in advance, reports become impossible to compare over time. We walked through this setup step by step in our rank tracking guide.

When SEO checker tools are useful

SEO checker tools that give an instant score for a single URL are useful for a quick first impression. They are enough to spot obvious problems before a sales call or to check a page's title and description tags. But these scores are calculated differently from tool to tool and are not a measure Google uses. A low score from an SEO checker does not mean the page will not rank; a high score does not guarantee that it will.

We built our own free helpers on the same logic: the meta tag, schema and UTM helpers on our tools page are there to speed up a specific task, not to reduce a site's overall SEO health to a single number.

Beyond search: three layers of one measurement stack

The layers above answer one question: how visible is the site in organic search, and what is technically in the way? That is necessary, but it stops at the click. Search tools cannot tell you which visits became qualified leads, which paid campaigns share the same audience, or whether the revenue a platform reports actually happened. For that the stack needs two more layers that connect to the first: analytics and attribution, and paid media. Treat the three as one workflow with shared definitions, not as three separate tool shopping lists.

What organic tooling can and cannot measure

Search Console, crawlers and market tools measure visibility, crawlability and competitive position. Search Console data is first-party but covers only Google Search, and market-tool traffic figures are estimates. None of them records what a visitor does on your site, so none can say whether organic traffic produced enquiries or sales. That boundary is where the next layer starts.

Layer B: analytics, attribution and business intelligence

The analytics layer turns visits into measured behaviour and, ideally, into business outcomes. It usually consists of a few connected parts rather than one product.

• Google Analytics 4 (GA4) records events and key events (conversions) on the site. It is free in its standard form, with a paid enterprise tier for very large properties.
• Google Tag Manager is the container that deploys and versions tags without code releases. Its standard tier is free; it is worth using mainly because every tag change becomes reviewable and reversible.
• Search Console integration in GA4. Linking the two adds the Google Organic Search Queries and Google Organic Search Traffic reports, which put query and landing-page data from Search Console next to engagement and conversion data from GA4. Google notes that the data can arrive up to 48 hours after collection, so treat the combined report as a weekly view, not a live one. Our Search Console guide covers which reports drive which decisions, and the GA4 setup guide covers the first reports.
• Looker Studio builds the reporting layer on top: one dashboard that joins Search Console, GA4 and ad-platform data. The core product is free, with a paid Pro tier for team workspaces and administration. A worked example is our Looker Studio SEO dashboard for clinics.
• A CRM or booking system holds what happens after the form: which lead was qualified, which became a customer and for how much. Without it, analytics ends at the enquiry. For a clinic example see source-to-booking attribution with a patient coordinator and CRM.

Consent, data quality and the limits of attribution

Measurement should be designed around consent from the start. Consent Mode and server-side collection reduce, but do not remove, the gaps that consent choices and browser restrictions create, and modelled conversions should be labelled as modelled in reports. Typical data-quality failures are duplicate events, a purchase event that fires again on refresh, missing UTM discipline, unfiltered internal traffic and conversion events nobody has defined. Attribution models assign credit using rules; they do not prove causation. Our articles on multi-touch attribution and cookieless attribution go deeper on what each model can support.

Layer C: paid media and cross-channel optimisation

The paid layer adds platforms where you buy attention, each with its own data and its own way of counting conversions.

• Google Ads provides campaign management and, through Keyword Planner, keyword ideas that are available inside the Ads account. Search term reports show the queries that actually triggered ads.
• Microsoft Advertising is worth including where Bing search share matters for your market. Its Google Import tool brings over campaigns and conversion goals from Google Ads, and it supports offline conversion goals. Imported settings still need checking, because the tracking tags on your site differ between the two platforms.
• Meta Ads Manager runs Facebook and Instagram campaigns. Meta's Conversions API sends server events that are processed alongside Pixel events, which is why event deduplication matters.
• Conversion tracking and offline imports. Google Ads supports enhanced conversions for leads, which uses the click ID and hashed first-party data from forms to connect a later CRM outcome to the ad. This is the mechanism that lets campaigns optimise toward qualified leads or sales rather than raw form submissions.

Using paid data to inform organic work, without treating them as the same thing

Paid search terms, ad copy results and landing-page tests show quickly which wording and offers earn clicks and conversions. That can inform which topics and titles organic content tests first. It cannot tell you organic demand or ranking potential: paid performance depends on bids, budget and auction competition, while organic ranking depends on different factors, so a keyword that converts well in ads may still be hard to rank for, and the reverse. Read the two side by side, and report them in separate lines. The relationship between the channels is covered in our guide to organic and paid sync in health tourism, and the weekly Google Ads optimisation routine shows the paid side in practice.

A practical workflow that connects the layers

1. Define outcomes first. Write down what counts as a qualified lead or a sale, and where each is recorded.
2. Instrument once. Define key events in GA4 through Tag Manager, with consent handled, and mirror the same definitions to each ad platform.
3. Connect the CRM. Pass lead source and click identifiers into the CRM and return qualification and revenue status to GA4 and the ad platforms through offline imports.
4. Report on one source of truth. Pick the system that drives budget decisions, usually GA4 or the CRM, and show platform-reported numbers beside it as diagnostics.
5. Review monthly. Compare organic and paid contribution by landing page and topic, decide where budget or content effort moves, and write down why.

Our paid media service and analytics and data service work from this same sequence, with measurement specified before campaigns change.

What to build at each maturity level

• Starting out: Search Console, GA4 with a handful of defined key events, Tag Manager, and one spreadsheet or Looker Studio report. Skip paid platform tooling until the outcomes are defined.
• Growing: add a crawler on a schedule, rank tracking with fixed settings, the Search Console link in GA4, UTM conventions, and one paid platform with conversion tracking verified.
• Established: add the CRM feedback loop and offline conversion imports, multi-platform paid media on shared definitions, and incrementality tests for large budget decisions.
• Advanced: add a warehouse such as BigQuery, server-side tagging and marketing mix modelling if spend and history justify the cost. These are not prerequisites for good results.

Free, freemium and paid: how to compare

Many of the tools in these layers have a free core: Search Console, GA4, Tag Manager and Looker Studio are free in their standard forms, and the ad platforms themselves are free to use because you pay for media, not the interface. Paid tiers and third-party platforms add scale, collaboration, support or data you cannot get elsewhere. Pricing and limits change often, so we do not publish figures here; check each vendor's current pricing page before budgeting, and judge a paid tool by the decisions it supports, as described above.

Common implementation mistakes

• Reading platform-reported conversions as additive across channels.
• Launching campaigns before conversion tracking is verified.
• Reporting organic and paid in one blended number so neither can be assessed.
• Treating a tool's score or estimate as a measured fact.
• Leaving the CRM disconnected, so optimisation stops at the form submission.
• Giving the agency ownership of the client's accounts and properties.

Documenting the stack: ownership, access and a cost line

A tool stack stays tied to individuals as long as it is not documented. Keep a simple record for each tool: which decision it is used for, whose account it is, how client access is granted, when it renews and which data is lost if it is cancelled. That last question matters; some tools keep past crawls or ranking history only while the subscription lasts.

When assessing cost, look not at the list price but at the number of decisions supported per tool. A subscription opened once a month can often be replaced with short, project-based access. The reverse is also true: running a tool the team uses every day on the limits of a free tier comes back as lost time.

An annual review keeps the stack healthy. Match every tool against the decision table again; a tool tied to no decision is removed, and a new tool is considered for any decision without a data source. We set up this framework in the first week of client projects as part of our SEO services; tool selection is treated as part of the measurement plan.

Sources

• Google Search Central — Using Search Console: Search Console's performance, index and alert functions.
• Google for Developers — About PageSpeed Insights: the difference between field data and lab data.
• Chrome for Developers — Introduction to Lighthouse: open-source page quality auditing.
• Semrush Knowledge Base — Domain Overview: estimated traffic data in the domain report.
• Google Analytics Help: Search Console integration: the two reports the integration adds and data availability.
• Google Ads Help: Enhanced conversions for leads and offline conversion imports: click ID and hashed first-party data.
• Meta for Developers: Conversions API: server events processed alongside Pixel events.
• Microsoft Advertising API guide: Google Ads Import: what Google Import brings over.