From checking 1,043 URLs across two sites to verifying repairs in production
Netsujo Inc. ·

On November 19, 2026, Netsujo checked 1,043 sitemap-listed URLs across our two websites. We then repaired two of the issues we found: a link to an aggregate CSV from our original research, and the HTML structure of a sales page for agency partners. We checked both changes at their production URLs.
This article connects the broad URL check with the two repairs we investigated in depth. It offers a practical way to define a website improvement as one piece of work, from the target URL and proposed change through to checking the published result.
Restore access to material readers can examine and share
Readers who want to examine a research article's figures need a working route to the underlying data. In this case, the data page opened, but its CSV download link returned 404. The page was available while the material needed for the reader's next task was out of reach.
The table shows the two repairs and what we confirmed in production. If it does not fit on your screen, you can scroll it horizontally.
| Target | What the check found | Confirmed cause and change | What we confirmed in production |
|---|---|---|---|
| Aggregate CSV for the original research | The link from the data page returned 404 | The destination had no route and was outside the configured delivery routing. We implemented the CSV under an existing delivery path and used the same aggregation definitions as the page's table | HTTP 200, with 1 header row plus 26 data entries. The data page also linked to the new URL |
| Sales page for agency partners | The HTML contained 13 h1 elements and no OG image declaration | We separated the screen content from its print copies, retaining h1 for the screen title and using h2 for the print titles. We also set an OG image default | The change took the page from 13 h1 elements to one, with an OG image declaration. A further check on November 20 confirmed that state remained in place |
The verified results concern access to the CSV and the published structure of the sales page. Effects on search rankings or inquiries remain unmeasured. The sales page still had zero hreflang declarations for language alternatives, and work remained on language support and other broken links.
The affected resources are the original research data page and the sales page for agency partners.
Check the 1,043 HTTP 200 responses and their onward links separately
The denominator was the set of URLs listed in the sitemaps when we retrieved them: 705 on netsujo.jp and 338 on www.miyakodeit.com, for a total of 1,043. This was not an inventory of every URL reachable on either site.
| Item | Scope of this check |
|---|---|
| Check window | November 19, 2026, 00:38–00:49 JST |
| Targets | Sitemap-listed URLs at retrieval time: 705 on netsujo.jp and 338 on www.miyakodeit.com |
| Retrievals | 1,043 body GET requests, 176 HEAD requests for internal links outside the sitemaps, and 4 sitemap and robots retrievals |
| Data examined | Server-returned HTML, HTTP responses, links, metadata and related structure |
| Method | Observed robots.txt; sequential retrieval per host, with a 500 ms interval and a 15-second timeout. Redirects were recorded without being followed |
| Not performed | AI semantic evaluation, checks of the page after JavaScript execution, form submission, visual layout or readability checks, assessment of content meaning and quality, or measurement of business effects |
All 1,043 body GET requests returned HTTP 200 and HTML. That establishes that HTML was retrievable at those URLs. It does not establish the quality of the content or that every link within each page was usable.
We separately checked 176 internal links outside the sitemaps using HEAD, which does not retrieve the response body. Those 176 checks cannot be added to the 1,043 successful body GETs and presented as a single count of pages examined. Recording the target set and method separately preserves what each result means.
The CSV was a concrete example of an issue that checking the page's HTTP 200 response alone would leave unresolved. Opening the page and obtaining the supporting material linked from it needed separate checks.
Trace the CSV failure through the resource and delivery path
Investigating the 404 showed that the old destination had no corresponding file or route. Its path was also outside the routing that directed requests to SIGNAL. A repair therefore needed both a resource to return and a delivery path that could reach it.
We added a route that returns the CSV under the existing delivery configuration, then changed the data page's link to that URL. The repaired aggregate CSV is linked from the data page.
We also aligned the meaning of the figures in the table and the CSV. The CSV uses the underlying aggregate data and shares the numerator definitions used to calculate proportions in the data page's table. This lets readers compare the published table with the downloaded data on the same terms.
Where a proportion has a zero denominator, the CSV leaves it blank. An absent denominator gives no basis for calculating a percentage; writing 0% would give it a different meaning. If a required variable definition is missing, generation stops instead of filling the gap with an assumption.
In production, we checked HTTP 200 alongside the text/csv Content-Type, an attachment declaration for downloading, and a noindex directive to keep this distribution file out of search indexing. The latter is the setting chosen for this file, rather than a recommendation to apply it to every CSV.
Separate the sales page's screen and print structure
The sales page for agency partners contained both the screen version and copies prepared for printing within the same HTML. Alongside the screen title, 12 print copies also used h1, bringing the total to 13.
We retained one h1 for the screen version and changed the print titles to h2. Styling is defined through classes, allowing heading hierarchy and visual presentation to be handled separately. The change makes the distinction between screen and print content explicit in the implementation.
We also added an OG image default, giving the page image metadata that sharing services can reference. We verified the declaration in production HTML. Display results within individual sharing services were outside this check's scope.
Reducing the count from 13 h1 elements to one was the outcome of organizing this particular page's structure. We have not used that count as evidence of better search rankings. Screen presentation, printed output and search performance each require observations appropriate to that question.
Define a bounded change and a concrete completion condition
The checks from this work can be organized into four parts of an implementation brief. They turn a broad inspection result into a target someone can work on and a condition both sides can use to judge completion.
- Target URLs and symptoms: Record which page leads to which resource, and what happens along that route. For the CSV, pair the data page with the destination returning 404.
- Confirmed cause: Identify the place to change: a file, route, delivery rule, or duplicated content in the HTML. Record any remaining uncertainty as well.
- Change and conditions to preserve: Align CSV delivery with the link and share the aggregation definitions. For the sales page, distinguish the screen and print content when changing the headings.
- What to read after publication: Check the HTTP response, CSV content and sales-page HTML at the same URLs. Add observations from production to the implementation record.
A request simply to “fix the material” leaves room for different interpretations of completion. Specifying the URL and expected content gives the requester and implementer a common result to examine. In this work, that included 1 header row plus 26 data entries in the CSV, and the h1 and OG image declarations in the sales page.
What we confirmed in production, and what remains to be measured
After the changes reached production on November 19, we verified that the CSV was retrievable and that the sales page had one h1 and an OG image declaration. On November 20 at 14:27 JST, we made unauthenticated GET requests to the target URLs again. The CSV returned HTTP 200 and 27 rows including the header. The sales page returned HTTP 200 with one h1 and an OG image declaration.
The November 20 checks covered the repaired URLs. They were not a repeat of the full 1,043-URL check. The production state described here is based on those recorded observations.
The sales page's hreflang remained unrepaired, with other work still open on language support and broken links. We also had no measurement of how the changes affected visits from search or inquiries. Assessing those effects requires observing access to and use of the material, and behavior along the inquiry path, over defined periods and under comparable conditions.
A repair record also provides a reference point for that later work. Knowing which state changed, at which URL and when, makes it possible to examine subsequent behavior alongside the change.
Apply the approach to your own website
Start with your page inventory and choose one onward link to a download or an inquiry route. Check whether the entry page opens and whether its destination can be used. If you find a problem, bring the cause, proposed change and repeat check at the same URL together in one implementation brief.
A page inventory provides a starting point for finding work. Connecting each change to what the reader needs to do next, and to a check after publication, makes that work easier for an internal team or an existing website partner to act on.
Netsujo SIGNAL is a continuous improvement service that develops a website into a sales foundation: a place where customers can discover, understand and compare an offering, then move toward a purchase or engagement. If you need to decide what to improve in a download or inquiry path, start with a target URL and the task that is proving difficult. Our service information explains how we help organize the evidence, changes and completion conditions.