#web performance

Si tu medio digital va lento, los lectores se van antes de que el contenido tenga chance de engancharlos. Y sí, eso también puede estar frenando tus ingresos más de lo que imaginas.

¿Cuánto se pierde por una web lenta y anuncios que bloquean el render? Medios y tiendas online con LCP por encima de 2,5 s ven menos páginas vistas, peor SEO y conversiones que caen; responsables con conocimientos medios suelen estar frustrados por inversiones en marketing que no compensan. Se ofrecen soluciones técnicas concretas y priorizadas para decidir o supervisar desarrollos sin jerga innecesaria.

Optimización para Core Web Vitals para medios digitales: optimizar Core Web Vitals en medios digitales requiere priorizar LCP, CLS e INP en producción, controlar recursos de terceros (anuncios, iframes) y aplicar soluciones prácticas como preload, preconnect, lazy-loading y critical CSS. Incluye un checklist priorizado por impacto/esfuerzo, comandos y ejemplos reales para mejorar experiencia y SEO sin sacrificar ingresos publicitarios; ponerlo en práctica puede dar mejoras visibles en horas en staging y en días en producción.

Resumen del proceso y resultado en horas

Aplica los pasos enumerados y verás mejoras de LCP y CLS en horas en entornos de staging y en días en producción. Las acciones clave: controlar terceros, reservar cajas, preload de recursos críticos, timeouts y medición de campo.

1. Controla anuncios y terceros sin cortar ingresos. 2. Reduce LCP: preload, imágenes adaptativas y CDN. 3. Mide INP y fija SLOs; añade budgets en CI. 4. Ajusta CMS y edge rendering. 5. Vigila con RUM y LHCI.

Pero cuando empiezas a ajustar lo que de verdad importa, el cambio puede ser mucho más grande de lo que parece…

We run sitemap indexation audits as the first check on almost every new SEO engagement. It is fast, it is quantitative, and it produces a scoping decision within the first day of the engagement. Here is the workflow, top to bottom, as we actually execute it.

Photo by ROMBO on Pexels

Step 1: get access

Before anything else, we need three things from the client:

Read access to Google Search Console for the property

The sitemap URL (usually /sitemap.xml or /sitemap_index.xml)

Read access to the CMS or at least a screenshot of the SEO plugin configuration

The first two are the audit inputs. The third tells us whether the fixes we will recommend are executable at the CMS level or will require developer time. That distinction shapes the scope-of-work document.

Step 2: pull the sitemap and count it

Fetch the sitemap. If it is a sitemap index that references child sitemaps, recursively fetch every child. Deduplicate URLs across children (some CMSes emit the same URL in multiple children).

Break the URL list down by pattern:

Posts (usually /blog/*, /articles/*, or similar)

Pages (top-level marketing pages, service pages)

Tag or taxonomy archives (/tag/*, /category/*)

Author archives (/author/*)

Date archives (/YYYY/*, /YYYY/MM/*)

Paginated versions of any of the above (?page=2, /page/2/)

Search or filter URLs

Legacy patterns (URLs from a previous site structure that are still generated)

The URL pattern breakdown is one of the most useful artifacts of the audit. It tells us at a glance where the noise is concentrated.

Step 3: pull the indexed URLs from Search Console

Search Console has a Pages report that can be filtered to a specific sitemap. Export the "All Pages" data as CSV. This gives us Google's view of what is indexed vs excluded, by URL, along with the specific exclusion reason for each.

The Google Search Central developer docs cover the API endpoints for URL inspection that can accelerate this on large sites, but the API is not necessary for most audits.

Step 4: diff sitemap against indexed

Set operations produce three URL groups:

Both (in sitemap and indexed): healthy. No action.

Sitemap only (in sitemap, not indexed): candidates for fix or removal.

Indexed only (indexed, not in sitemap): usually a canonicalization mistake or legacy URLs that are still being served.

Most of the interesting work is in the "Sitemap only" group. That is where the Search Console exclusion reason matters.

Step 5: categorize the "Sitemap only" URLs

For each URL in the "Sitemap only" set, we bucket by the Search Console exclusion reason:

"Duplicate without user-selected canonical": canonical fix needed

"Alternate page with proper canonical tag": remove from sitemap (why include a canonical target's alternate)

"Crawled - currently not indexed": Google decided the content is not worth indexing - either fix content or remove

"Discovered - currently not indexed": Google chose not to crawl - usually a crawl-budget or discovery-authority issue

"Excluded by 'noindex' tag": remove from sitemap and question why the sitemap generator includes noindexed URLs

"Blocked by robots.txt": either unblock or remove from sitemap - blocking crawl is different from blocking indexation

"Soft 404" or "Not found (404)": remove from sitemap and either restore the URL or 301 it

"Server error (5xx)": fix the server error before doing anything else

Each bucket has a different recommended action. The scope-of-work document is built from the aggregate.

Step 6: sample the largest bucket

The largest bucket almost always tells the site's story. For most WordPress-based content sites, the largest bucket is "Crawled - currently not indexed" and the URLs in it are tag archives, paginated archives, and thin auto-generated pages. For most Shopify-based e-commerce sites, the largest bucket is "Duplicate without user-selected canonical" and the URLs are variant product pages.

We open 10 URLs from the largest bucket and look at:

Actual on-page word count

Canonical tag (does it exist, does it point where we expect)

Internal link count to the URL from the rest of the site

Whether the URL is genuinely useful to a visitor

This pattern-match usually takes 15 minutes and produces a clear picture of what is going on. From there, the recommendation writes itself.

The full walkthrough of this sampling and categorization step is in the sitemap audit article we publish for teams doing their own audits.

Step 7: write the scope document

Every audit produces a scope-of-work document that lists:

URL patterns to remove from the sitemap

URL patterns to noindex at the CMS level

URL patterns to 301 redirect

URL patterns that need actual content improvement

URL patterns to leave alone

Each row includes an approximate URL count and a difficulty rating (CMS-side change vs developer change). The client gets a proposal with clear scope and estimated impact.

Step 8: implement and monitor

Implementation is either the client's team executing our scope document or we handle it under a fixed-price scope. Either way, we monitor Search Console weekly for the first month to catch any indexation drops on URLs the client actually wanted ranked.

The two metrics we track:

Indexation rate (indexed / submitted)

Absolute indexed count of "kept" URLs (URLs we did not remove)

The first should climb steadily. The second should climb slowly or stay flat.

If the first is climbing but the second is dropping, we are deindexing URLs the client actually wanted ranked. The scope document was too aggressive. Back off.

If both are flat, the CMS configuration change did not stick. Check that noindex tags are actually rendering and that robots.txt is not overriding.

If both are climbing, the audit worked.

Photo by cottonbro studio on Pexels

Step 9: hand off

The audit engagement typically ends with a scheduled 90-day check-in and a delivered scope document, a fresh sitemap-diff script the client can re-run themselves, and a written playbook for what to do quarterly to prevent regressions.

Most clients see meaningful indexation and organic-traffic improvement within 6 to 10 weeks of implementation. A minority see nothing meaningful, which usually means the audit was not the highest-leverage intervention and we redirect the engagement elsewhere (content, internal linking, or backlink work).

Why this is the first thing we run

The sitemap audit is the first check because:

It is fast (one day of analysis for a mid-sized site)

It is quantitative (Search Console gives real Google data, not third-party approximations)

The scope decision is clean (either the numbers indicate leverage or they do not)

The fix is usually cheap (config-level changes, not content rewrites)

The results are measurable (indexation rate is a hard number)

Compare that to a content audit or a backlink audit - both are slower, more subjective, and produce less clearly measurable outcomes in the first six weeks. Not that content and backlinks do not matter, but the sitemap audit either produces a fast win or eliminates itself as the problem, and both outcomes are useful.

Where to learn more

The Google Search Central documentation is the canonical reference on how sitemaps and Search Console reports work. The Ahrefs blog has good coverage of specific indexation issues and how to interpret Search Console patterns. Moz has some of the clearer general-audience writeups on technical SEO fundamentals.

The 137Foundry technical SEO service is where we run this workflow end-to-end for clients. For sites doing it themselves, the detailed audit walkthrough covers the mechanics.

At 137Foundry we run a lot of technical SEO audits, and the sitemap-vs-indexation gap analysis is almost always the first check we do on a new engagement. Most of the time it finds something worth fixing. Sometimes it finds something that reshapes the whole engagement.

The audit we ran for a mid-sized publisher last quarter is the extreme end of that distribution. Here is what we found and what it meant for the client's organic traffic.

Photo by Pixabay on Pexels

The starting picture

The client had 8,400 URLs in their sitemap. Google Search Console reported 3,441 indexed. That is a 41 percent indexation rate, which is low even by "content site" standards - most healthy sites sit in the 70 to 85 percent range.

Organic traffic had been flat for about 18 months despite the client publishing an average of 12 new long-form posts per month. Their instinct was that Google had penalized them somehow. Their previous SEO consultant had focused on backlink acquisition and on-page optimization. Neither had moved the needle.

The client's own diagnostic hypothesis was "Google does not like our content." The actual diagnosis turned out to be "Google would like your content fine if it could find it under all the noise."

What the sitemap contained

We pulled and inventoried the sitemap. The 8,400 URLs broke down roughly like this:

1,900 published posts (which is what the client actually wanted ranked)

2,400 tag archive pages

1,600 author archive pages (from 40 authors, each with a paginated archive going back years)

1,300 date archives (year, month, and day pages)

900 paginated pages of the main blog index

300 legacy URLs from a 2019 CMS migration

The published posts were 22 percent of the sitemap. Everything else was CMS overhead.

The client's own view of their site was "we have almost 2,000 posts." Google's view of the site, based on the sitemap, was "you claim to have 8,400 pages worth indexing."

What was actually indexed

The 3,441 indexed URLs broke down as:

1,510 published posts (80 percent of posts were indexed - not bad)

720 tag pages (30 percent of tag pages were indexed - the ones with more posts under them)

400 author archives (25 percent)

380 paginated blog index pages

250 date archives (mostly year-level)

181 misc

So Google was indexing about 30 percent of the CMS-generated overhead and about 80 percent of the posts. The overhead was contaminating the site's crawl profile and, we suspected, its quality signal.

The full workflow for producing this inventory is written up here - the client-specific version is just the same audit at a larger scale.

The plan

We wrote a scope-of-work document with four changes:

Remove all CMS-generated archive URLs from the sitemap (tag pages, author pages, date archives, paginated blog index).

Add noindex, follow to those same URL patterns at the CMS level (so Google would deindex the ones it had already indexed and stop bothering with the rest).

301 the 300 legacy URLs to their current equivalents. Delete the ones that had no equivalent.

Trim the tag chip links at the bottom of each post template so posts stopped voting for zombie tag pages.

Total implementation was one day of technical work on the client's side plus configuration changes in Yoast.

The result

Six weeks after the changes went live:

Sitemap dropped from 8,400 URLs to 3,200 (62 percent reduction).

Indexation rate climbed from 41 percent to 88 percent of the new sitemap.

Absolute indexed URL count climbed from 3,441 to 2,816.

Wait - the absolute indexed count dropped. That is expected. Google deindexed the tag pages, author pages, and paginated archives as requested. That was the point.

The interesting number is what happened next: organic traffic on the client's actual posts climbed 23 percent in the same six-week window. Because Google's crawl budget was no longer being spent on 5,000+ URLs of CMS overhead, it was reaching new posts within days instead of weeks, and it was recrawling old posts more often (which lets ranking updates surface faster).

Rankings on cannibalized keywords - queries where a tag archive had been outranking the post that actually deserved to rank - shifted. In several cases the post moved into position 1-3 within four weeks of the tag archive being deindexed.

What we did not do

Notable omissions in the scope:

We did not delete any posts.

We did not rewrite any content.

We did not acquire any backlinks.

We did not change on-page SEO on individual posts.

All of the traffic gain came from letting Google concentrate on what already existed. This is the pattern we see repeatedly at 137Foundry - the biggest quiet SEO win most sites can make is not building more content but removing the URLs that Google is spending crawl budget on and not ranking.

The Google Search Central documentation covers the mechanics of crawl budget in general terms; the specific mechanics of how sitemap cleaning affects it are best learned by doing the audit and watching the numbers.

What made this client's case extreme

Two things made this engagement produce more visible movement than a typical sitemap audit.

First, the client's site had unusually low indexation to start (41 percent vs typical 60-70 percent for sites in this category). The gap between where they were and where they could be was bigger.

Second, they had genuinely high-quality posts that were being cannibalized by CMS overhead. Sites with weaker content do not see the same rebound because there is not as much authority waiting to be released.

On sites with typical starting conditions, we see indexation gains in the 15-25 percentage point range and organic traffic gains in the 5-15 percent range. This client was well past both.

Photo by Letícia Alvares on Pexels

What we learned about diagnosing this from the outside

Two questions distinguish sites where the sitemap audit will produce big results from sites where it will only produce marginal results.

Question one: what fraction of your sitemap is CMS-generated archives? If it is over 30 percent, the audit is going to move numbers. If it is under 10 percent, the audit is a cleanup that will produce marginal gains at best.

Question two: what is your current indexation rate? Below 60 percent means you have leverage. Above 80 percent means you are already close to what the site can achieve.

Both of those are visible in Search Console within 10 minutes of a diagnostic call. We know within that first call whether a sitemap-focused engagement is going to be transformative or incremental.

For sites whose numbers indicate a sitemap-focused engagement is worth running, the 137Foundry technical SEO service is where we handle the audit and fix end to end. For sites where the numbers indicate the leverage is elsewhere, the diagnostic call itself is often enough to redirect the SEO conversation productively.

For deeper technical context, the Ahrefs Site Audit documentation and Google Search Central cover overlapping ground.

The takeaway

The best-kept SEO secret is not a link-building tactic or a content framework. It is that most sites are asking Google to consider a much noisier set of URLs than they should be, and cleaning that set produces measurable ranking gains without touching a single piece of content.

Cannibalization audits are one of those SEO deliverables that clients often want and rarely understand. The word "audit" makes people picture a spreadsheet with hundreds of highlighted rows and a set of recommendations that will take six months to work through. The version that actually produces results is much smaller in scope, moves through a specific sequence, and produces a shortlist of decisions rather than a shopping list of tasks.

This is a walkthrough of what the audit actually looks like when we run it, why we structure it the way we do, and what the delivered output usually looks like at the end.

The intake question

Before we run any tool, we ask one question: what queries does the client actually care about ranking for. Not the queries their site currently ranks for. Not the queries their competitors rank for. The specific queries that, if the client were ranking at position 3 for them tomorrow, would materially change their business.

The answer is almost always a short list. Between 8 and 40 queries covers most B2B service businesses. E-commerce sites are longer, but even there the 80/20 rule usually surfaces 100 or so queries that matter versus the thousands of long-tail phrases the site touches.

Cannibalization audits that fail almost always start by casting too wide a net. If the audit tries to fix every query where more than one URL gets an impression, it produces a backlog nobody works through. Starting from the client's actual priority list produces a shortlist of high-impact decisions instead.

The tooling pass

Once we have the priority query list, we run the specific queries through Google Search Console's Performance report and record which URLs are getting impressions for each. This takes about 30 minutes for a 40-query list, done manually. For larger lists we use a Python script to pull the data through the Search Console API.

We also do a manual SERP inspection in a private browsing window for each priority query. This surfaces cases where a page is ranking that Search Console has not yet weighted heavily enough to appear in the "top pages" list. For high-volume queries this manual pass is faster than trying to configure the API pull to be sensitive enough.

The output of the tooling pass is a table with three columns: query, URLs on the client's site with impressions, and current ranking positions for each. Most rows will have one URL. The rows with more than one URL are the ones we investigate further.

The classification pass

For each query with multiple URLs, we classify the situation into one of four categories.

True cannibalization. Two pages targeting the same searcher intent, splitting the signal, both ranking below where either would rank alone.

Healthy topical clustering. Multiple pages on related but distinct sub-topics. The site's authority on the topic is spread across a cluster of related pages, which is often what a strong SEO strategy looks like. Not a problem.

Category or archive page interference. A tag page or category page from the CMS is competing with individual content pages. Usually fixable with a noindex or a canonical, not with a content change.

Intent mismatch. Two pages ranking for the same query but serving different searcher intents (transactional vs informational, for example). Usually a case where the query itself has multiple intents and Google is showing different pages to different searchers. Sometimes worth splitting into two queries with differentiated pages; often worth leaving alone.

The classification is where the audit turns from data-gathering into decision-making. Most of the raw candidates from the tooling pass end up in the "not actually a problem" categories. The real cannibalization is usually a much shorter list than the raw candidates suggest.

The decision pass

For each confirmed true cannibalization case, we decide the consolidation direction. Which URL should own the query, and what mechanism gets us there.

The default is that commercial-intent queries go to the service or product page and informational queries go to the blog post. Exceptions get discussed with the client, usually because the blog post has accumulated more backlinks than the service page or because the blog post is ranking meaningfully higher than the service page and the client wants to preserve the current traffic while consolidating.

The mechanism choice (301 redirect, canonical tag, rewrite to differentiate) follows the framework laid out in the detailed cannibalization guide our team maintains. Each case gets one specific recommendation, not a menu of options for the client to choose from. Menus produce indecision; single recommendations produce implementation.

The implementation pass

Depending on the client, we either implement the consolidations ourselves as part of the engineering team at 137Foundry engagement, or hand the specific implementation instructions to the client's development team.

The implementation is where audits most often break down when done poorly. A well-written recommendation says exactly which file to edit, exactly which redirect rule to add, exactly which internal links to update, and exactly what to test after deployment. A poorly-written recommendation says "consolidate the two pages" and leaves the mechanical details to the implementer, who often does something inconsistent with the intent.

For redirect rules specifically, the Apache mod_rewrite documentation and the corresponding Nginx rewrite directives are worth having open while writing the specific rules. Small syntax errors produce redirect loops or missed cases; a specific tested rule is worth much more than a general "add a 301" instruction.

The verification pass

Two weeks after implementation, we re-run the tooling pass on the same query list. For each consolidation, we verify that the intended URL is now the only one getting impressions for the query, and that the ranking position has moved up (or held steady if it was already good).

Cases where the consolidation did not take effect get re-investigated. Common causes: internal linking that still points at the redirected URL, canonical tag on the wrong page, sitemap still listing the redirected URL, or Google simply choosing to ignore the canonical based on other signals. Each has a specific remediation.

The verification pass is what separates audits that produce results from audits that produce reports. Without verification, we would not know which recommendations actually worked, and we would not learn which types of consolidations tend to succeed versus fail on this specific client's site.

What the client gets at the end

The final deliverable from a cannibalization audit is a short document with three sections.

The confirmed cannibalization list with the specific recommendation for each. Not a menu; a decision.

The implementation instructions for whoever will do the work, written specifically enough that a developer unfamiliar with SEO can execute them.

The verification plan, including what to look for two weeks after implementation and how to know whether each consolidation succeeded.

Total pages in the deliverable: usually 8 to 15. Total specific consolidation cases: usually 12 to 30 for a mid-sized site. Total implementation effort for the client: usually 4 to 8 hours of developer time, spread over one or two sprints.

Why the audit produces results this way

The audit works because it is scoped tightly, structured as a sequence of decisions rather than a menu of tasks, and verified at the end rather than shipped and forgotten. It is not glamorous. There is no "reveal" moment where the client sees a big red spreadsheet and has an epiphany.

What actually happens is that three months after the audit, the client's rankings for the priority queries have moved up, cannibalization has resolved on the specific queries that were consolidated, and the site has a documented process for preventing new cannibalization in the pages published going forward. The gain compounds because the fixes are permanent (URLs stay redirected, canonicals stay in place) and because the process change prevents new cases from being created.

That is a very different lived outcome from the "audit as report generation" pattern that produces long documents nobody acts on.

The one operational shift that most helps

If you are running cannibalization audits internally rather than contracting for them, the single most useful operational shift is to run the audit against a short priority query list every quarter, rather than against every query on the site once a year. Small, frequent, focused audits produce steady progress. Large annual audits produce backlogs.

The pattern that works: quarterly audit, 8 to 40 priority queries, 5 to 15 consolidation decisions, verified two weeks after implementation. Repeat forever. Over two years, the site's information architecture becomes noticeably cleaner and cannibalization becomes a rare event rather than a chronic condition.