Modelling PDF usage on UK local government websites

This article models PDF usage on UK local government corporate websites, using published council disclosures as its evidence base and triangulating the result against independent analytics sources. For the common case of a council hosting about 1,000 PDFs, the supporting evidence produces an estimate of 270 to 710 PDF requests served per day, with a central working figure of 450.

Executive summary

A practical, evidence-led estimate for a typical UK council corporate website hosting about 1,000 PDFs (excluding large committee-management platforms such as Modern.Gov) is roughly 270 to 710 PDF file requests served per day, with a central working estimate around 450 PDF requests per day.

This range is derived by scaling a rare, explicit council disclosure of file usage bands. In one large metropolitan council's main website estate (explicitly excluding its committee-management platform), there were 1,440 downloadable files (PDF, CSV, Word) and, in the past year, 67 percent were viewed fewer than 100 times, while 4 percent were viewed more than 1,000 times. Interpreting those bands conservatively produces an implied daily volume that, when scaled to 1,000 PDFs, lands in the 270 to 710 served-per-day range.

Across councils, multiple independent disclosures show an extreme long-tail pattern where most PDFs are barely used and a small minority drives most consumption. One unitary council reports a recent review finding over 5,800 PDFs, with 77 percent viewed 50 times or less and 49 percent viewed fewer than 5 times. A district council reports 78 percent of PDFs were not viewed in the last 30 days. A Scottish council area states it updated 363 PDFs, described as the only PDFs viewed or downloaded from the corporate site over the past 13 months.

These statements collectively support a modelling assumption that the median PDF receives close to zero usage, while the mean is pulled up by a small number of highly accessed documents.

Where councils publish page-level analytics that include document downloads, the conversion from page views to document downloads varies widely by context, typically reflecting high-intent pages such as consultations, toolkits, and route guides. Examples include 75,879 page views and 9,224 document downloads on a county consultation page (about 12 percent downloads per page view), and 7,084 page views with 158 toolkit downloads (about 2.2 percent). These ratios should not be applied to a whole council site without adjustment, but they provide grounded bounds for modelling the parts of a site where PDF downloads actually occur.

Any estimate is sensitive to measurement design. Many councils explicitly note that analytics counts are impacted by cookie consent choices. A county council performance report states that cookies are needed to count website traffic and that making "Reject cookies" more prominent likely reduced measured sessions without reducing actual usage. A Scottish council performance report similarly notes visits are based on users accepting analytics cookies and that actual visits are suspected to be higher.

Scope and definitions

This article focuses on PDF usage modelling for UK local government public web estates, specifically the common case where a council has a corporate content site hosting roughly 1,000 PDF files alongside HTML pages.

PDF served is defined here as an HTTP request that returns the PDF content to the requester (for example status 200 or 206), which includes in-browser viewing as well as saving a file. This is the natural definition for server logs and CDN logs, and it is the closest operational equivalent to how many PDFs are served per day.

This definition differs from a narrower analytics definition of download event, which often triggers only on a link click and may miss PDFs opened directly, cached-request patterns, or requests from third-party domains. For Google Analytics 4, file download measurement is typically implemented via enhanced measurement events and surfaces as an event such as file_download.

In this article, corporate website PDFs means PDFs hosted under the council's main site and standard subdomains. Committee-management PDFs means documents served via platforms such as Modern.Gov that can generate large and constantly changing collections of meeting papers, agendas, and reports. A London borough's accessibility statement describes its democracy portal as provided via a third-party platform (Modern.Gov) where the volume of PDF documentation changes daily.

The 1,000-PDF scenario should be read as the corporate site inventory case, not the committee-management case.

Evidence base

The UK local government sector does not publish a standardised, comparable set of PDF download metrics. This article therefore uses a best-available-evidence strategy built around three high-signal source categories:

• Publicly published analytics pages and performance reports (site users, visits, page views).
• Council accessibility statements and disproportionate burden assessments that disclose explicit PDF counts and view distributions.
• Published consultation, scrutiny and JSNAA reports that report page views together with document download counts.

Examples of usable traffic disclosures include Midlothian Council (346,508 website visits and 1,014,733 page views in a quarter, with an explicit note that visit counts depend on accepting analytics cookies); Rochdale Borough Council (monthly users and page views with commentary about cookie rejection reducing the measurable dataset); Lincolnshire County Council (a business toolkit instance with 7,084 page views and 158 toolkit downloads); and East Sussex County Council JSNAA annual report (grouped document access counts including 649 documents accessed 6,399 times in 2015/16).

Measurement distortions to model

Cookie consent and measurement loss is not a theoretical issue. Councils explicitly report that changes to cookie controls reduce measured traffic, and that cookies are needed to count some analytics datasets.

Caching and CDN layers distort the relationship between served to a user and seen in origin logs. Web request popularity is often Zipf-like, which is precisely the condition under which caching yields substantial offload and complicates naive counting.

Each of three signals (client analytics events, edge logs, origin logs) feeds a different downstream report, and each misses a different slice of reality. Client analytics miss non-consented users. Origin logs miss cache hits. Reports bounded to the corporate domain miss PDFs served from third-party hosts. Any defensible figure reconciles across those signals rather than trusting one.

Statistical model

A council PDF usage model needs two layers. A site layer converts overall site activity into a pool of document-serving opportunities. A document layer allocates total PDF requests across individual PDFs, which is highly skewed.

Because PDF popularity is long-tailed, the more informative object is a ranked popularity model rather than a mean. A ranked popularity model assigns a predicted annual request count to each PDF according to its rank in popularity, with the most popular PDF at rank 1. Annual requests fall away sharply as rank increases, following a power-law form that is characteristic of web object accesses.

The most operationally valuable disclosed parameter set is the banded distribution of annual views from the metropolitan council example:

• 4 percent of documents viewed more than 1,000 times per year.
• 4 percent viewed between 500 and 1,000 times.
• 25 percent viewed between 100 and 500 times.
• 67 percent viewed fewer than 100 times.

Additional disclosures tighten the lower tail of the distribution:

• 77 percent of PDFs viewed 50 times or less, and 49 percent viewed fewer than 5 times (unitary council).
• 78 percent of PDFs not viewed in the last 30 days (district council).
• Only 363 PDFs viewed or downloaded over 13 months (Scottish council area corporate site).

Scenario estimates for 1,000 PDFs

Using the metropolitan council's annual view bands as the empirical template, scaled from 1,440 files down to 1,000, gives a directly defensible estimate of PDFs served per day for a corporate estate. To convert the disclosed bands into totals, an average must be assigned within each band. The values below are intentionally conservative for the lowest band, consistent with other councils stating that many PDFs receive very low view counts.

Assumed mean annual views per PDF by band (illustrative):

• Over 1,000 views band: 1,100 (low), 1,600 (central), 3,000 (high).
• 500 to 1,000 band: 550 (low), 750 (central), 900 (high).
• 100 to 500 band: 120 (low), 250 (central), 350 (high).
• Under 100 band: 2 (low), 10 (central), 25 (high).

This yields the headline estimate of about 270 to 710 PDFs served per day for a 1,000-PDF council corporate estate, with a central estimate around 450 per day.

Even with conservative within-band means, a small minority of PDFs accounts for a majority of views. Under plausible within-band assumptions, the top 8 percent of PDFs (those above 500 views) can account for roughly 60 to 70 percent of annual file views.

Bounding via page-view-to-download ratios

A second, independent angle is a site-to-download conversion bound, constructed using explicit page-view and document-download pairs from high-intent pages: consultation pages around 12 percent document downloads per page view, and toolkit pages around 2.2 percent.

A council's full-site ratio will generally be lower than a consultation landing page, because only a subset of site journeys involve documents. But these disclosures credibly demonstrate that a 1 to 10 percent download intensity is realistic for document-heavy sections. If a typical council site has 5,000 to 12,000 page views per day, then even a 2 percent site-level PDF request intensity would imply 100 to 240 PDF serves per day, while a 5 percent intensity implies 250 to 600.

The band-based estimate of 270 to 710 PDFs served per day is therefore consistent with a world where PDFs are a small but meaningful slice of a council's daily content delivery, even before adding democracy portal traffic.

Served requests vs download events

Councils and analytics platforms often mix opened/viewed with downloaded. A key metropolitan disclosure states it is unable to track downloaded files and expects download counts to be significantly lower than view counts.

Where served requests must be translated into a narrower downloads concept, the relationship is a conversion factor applied to daily served requests to produce daily download events. A defensible sensitivity bracket for this conversion factor in local government settings is typically 0.3 to 0.7 if download is an event-based click measure and served includes direct opens, embeds, and repeat loads. The correct value is council-specific and should be estimated empirically using matched logs and analytics.

Recommendations

Build the denominator first

Modelling accuracy improves materially if 1,000 PDFs are not treated as a single pool. A practical categorisation that maps to council business functions is: service forms and transactional guidance; policy and strategy documents (including Local Plan and planning policy); consultations and evidence bases; statistics and dashboard reports; and committee and governance documents (if included). This is also aligned with the sector direction: councils routinely commit to moving to HTML-first and reducing the PDF footprint, while recognising that large back-catalogues and third-party documents create disproportionate burdens.

Instrument downloads in two independent ways

• GA4 file download events (captures click intent, depends on consent and configuration).
• Edge log counting of PDF responses (captures served reality, needs bot filtering and cache awareness).

Discrepancies are then reconciled:

• If event counts are far below edge logs, quantify consent loss and quantify direct-open behaviour.
• If edge logs are far below expectations, check whether a CDN cache is terminating requests outside the log boundary.
• Ensure third-party platforms are either included or explicitly excluded, but not mixed unknowingly.

Treat served PDFs per day as a distribution

• Total PDFs served per day (mean).
• Median PDF servings per day (likely near zero).
• 90th-percentile PDF (usage threshold for prioritisation).
• Share of all PDF serves accounted for by the top 1 percent, top 5 percent, and top 10 percent.

This is operationally aligned with how councils approach their website work: the most-used documents are addressed first, because resource is limited and the regulations recognise disproportionate burden where a council cannot reasonably act across an entire back-catalogue.

Practical bottom line

For a working planning number for a typical UK council website with about 1,000 PDFs, use:

• 450 PDFs served per day as a central estimate.
• 270 to 710 served per day as a defensible sensitivity range.
• An assumption that a small minority of PDFs accounts for most of that volume.

This estimate is suitable for capacity modelling, resource prioritisation, and setting measurement requirements. It is not a substitute for direct measurement on a specific council estate, especially where committee management, third-party portals, or consent-driven undercounting materially change the numerator or denominator.

Sources

All sources accessed April 2026.

Council accessibility statements and disproportionate burden assessments

• Manchester City Council. Disproportionate burden assessment.
• Warrington Borough Council. Accessibility statement.
• Wealden District Council. Disproportionate burden assessment for PDFs.
• Dumfries and Galloway Council. Disproportionate burden assessment.
• Medway Council. Accessibility regulations disproportionate burden assessment.
• London Borough of Bexley. Accessibility statement for www.bexley.gov.uk.

Council performance and traffic reports

• Midlothian Council. Corporate Solutions Quarterly Performance Report, Q2 2024 to 2025.
• Rochdale Borough Council. Website performance.
• West Lindsey District Council. Website statistics.
• East Devon District Council. Public reports pack, Scrutiny Committee, 10 December 2020.
• Epping Forest District Council. Report to Council (FED PH report).
• Kent County Council. Quarterly Performance Report, Q3 2023 to 2024 (Appendix 1).

Council committee, consultation, and scrutiny reports

• Kent County Council. Consultation Report (Appendix B).
• Lincolnshire County Council. Visitor Economy Update, Scrutiny Report, 28 February 2023.
• East Sussex County Council. Joint Strategic Needs and Assets Assessment (JSNAA) annual report, 2015 to 2016 (Appendix 1).

Analytics documentation and tooling

• Google Analytics. Enhanced measurement events, including file_download.
• Google Analytics. GA4 Data API.
• Matomo. Reporting API reference.
• Siteimprove. The Document Downloads group explained.
• GOV.UK Data Community. Find in GA4 guidance.

Regulatory, archival and methodology references

• GOV.UK. Accessibility requirements for public sector websites and apps.
• Breslau, Cao, Fan, Phillips and Shenker. Web caching and Zipf-like distributions: evidence and implications.
• The National Archives. UK Government Web Archive.
• The British Library. Cyber-attack statement and ongoing service disruption.
• UK Parliament. Deposited paper on server-log methodology.