30,800
Branch copies of every test page, each self-canonical or with no canonical at all. Google kept a branch version as the canonical for 61% of the tests it could read, often the branch furthest from the searcher.
A fictional business. The figures are a modelled scenario, not a client result.
PULSO LABORATORIOS · LIMA · PERU
+163% organic visits to test pages
13 months · Latin America · Signal plan · Diagnostic lab chain, 22 collection points · Lima
Canonical consolidation · robots.txt rewrite · MedicalTest schema · Segmented XML sitemaps
92% of the indexed catalogue is clear plasma
THE CHALLENGE
1,400 tests copied onto 22 branch pages, 610 blocked by robots.txt, no MedicalTest markup: searches for “prueba de glucosa precio” went to competitors.
Pulso had grown branch by branch, and so had its website. Each of the 22 collection points published its own copy of the full catalogue, a catalogue migration had left 610 tests behind a robots.txt rule, and nothing on any page told a search engine that this was a medical test with a price, a sample type and a fasting time. People in Lima searching for the price of a glucose test found competitors, aggregators and, now and then, a Pulso branch page with the wrong address on it.
THE SAME TEST, AS GOOGLEBOT FOUND IT
/examenes/glucosa-en-ayunasthe page that should rank/sedes/miraflores/examenes/glucosa-en-ayunas/sedes/san-isidro/examenes/glucosa-en-ayunas/sedes/surco/examenes/glucosa-en-ayunas/sedes/los-olivos/examenes/glucosa-en-ayunas/examenes/glucosa-en-ayunas?sede=jesus-maria/examenes/glucosa-en-ayunas?utm_source=wa+ 16 more copies of this test
1,400 tests × 22 branches
30,800 duplicate URLs
DIAGNOSIS
30,800
Branch copies of every test page, each self-canonical or with no canonical at all. Google kept a branch version as the canonical for 61% of the tests it could read, often the branch furthest from the searcher.
610
Tests left under /perfiles/ and /examenes/especiales/ after a catalogue migration, both disallowed. Hormone panels and urine tests were the worst hit: 150 of 240 and 90 of 180.
0
Test pages with MedicalTest markup. Price, sample type, fasting time and turnaround were plain text in an accordion, so neither search engines nor AI answers could read them as facts.
85%
Share of Googlebot requests spent on branch copies and ?sede= or tracking parameters, measured over ten weeks of server logs. New tests waited a month to be crawled.
THE PROTOCOL
Built on our Signal plan: the technical engagement for sites where crawl, indexing and architecture decide the outcome.
One canonical page per test at /examenes/. The 30,800 branch copies return a 301 to it, and the test page lists the 22 branches that take the sample, with opening hours. The 22 branch pages stay, as location pages that link to their tests.
The two legacy rules came out and the 610 tests moved back under /examenes/. The new file blocks only patient results, the ?sede= parameter and tracking tags, and names the test sitemap. Every change was tested against the live crawl first.
Each test page now declares a schema.org MedicalTest with the sample, preparation, fasting time and turnaround, plus an Offer with the price in soles. Generated from the lab's own catalogue, so a price change reaches the markup the same day.
One sitemap per test family, with lastmod tied to real catalogue changes, plus log monitoring by family. A stalled family showed up in days, and new tests were crawled within 48 hours instead of a month.
RUN THE CENTRIFUGE
Each tube holds every indexed URL for one test family. Canonical test pages rise as clear plasma; duplicates and blocked pages sink as sediment. Change a setting and spin again.
Settings
No canonical: Google chooses among 23 near-identical URLs per test.
Blocks /perfiles/ and /examenes/especiales/: 610 tests Google cannot read.
Price and preparation are plain text in an accordion.
robots.txt
User-agent: *Disallow: /resultados/Disallow: /perfiles/Disallow: /examenes/especiales/Sample URL
GET /sedes/miraflores/examenes/glucosa-en-ayunas<!-- no canonical tag -->Load a sample
Plasma: indexable canonical test pages
412
of 1,400
Sediment: duplicate and blocked URLs indexed
12,058
11,448 duplicates · 610 blocked
Googlebot requests on sediment
85%
Separation · 3%
Modelled position for “prueba de glucosa precio”
#14
Test pages with MedicalTest · 0
A simplified model of this fictional catalogue, not a crawl of a real site.
MONTH BY MONTH
Spin up while the audit and the rules are prepared, full speed through consolidation and markup, then a long deceleration while the index settles.
1 · Months 1–2
Full crawl of 38,600 URLs, ten weeks of logs, Search Console coverage by family, and revenue per test from the booking system.
2 · Months 3–4
robots.txt rewritten and tested against the live crawl; 610 tests moved back under /examenes/; canonical and redirect map signed off with the lab.
3 · Months 5–7
30,800 branch copies redirected in three waves, biochemistry first. Branch pages rebuilt as location pages with their own test lists.
4 · Months 8–9
MedicalTest and Offer markup generated from the catalogue for all 1,372 live tests; validated nightly against the price list.
5 · Months 10–13
Sitemaps per family, log alerts, internal links from the 22 branch pages. Duplicates drained from the index; the plasma kept rising.
RESULTS
Organic visits to test pages
+163%
18,200 → 47,900 organic visits a month
Indexable canonical test pages
412 → 1,372
Indexable canonical test pages, month by month
Duplicate URLs indexed
11,448 → 123
Branch copies and parameters still indexed at month 13: redirects not yet recrawled.
Position for “prueba de glucosa precio”
#14 → #2
Month 0 → month 13
Modelled scenario for a fictional business — not a client result. Rankings and traffic are never guaranteed; see our Terms.
| QUERY | MONTH 0 | MONTH 13 |
|---|---|---|
| prueba de glucosa precio | #14 | #2 |
| perfil lipídico precio lima | #19 | #3 |
| hemoglobina glicosilada examen | #23 | #4 |
| examen tsh precio | #17 | #3 |
| examen de orina completo precio | #12 | #2 |
| prueba de embarazo en sangre lima | #26 | #5 |
Queries as searched in Lima, in Spanish. Positions are modelled monthly medians.
A large catalogue, many locations, a robots.txt nobody has read in years? Start with the Signal plan, or tell us what your site looks like.