Privacy-first measurement public sector Birmingham: principles

What to take away

Rule one: separate the public task from marketing consent

Birmingham City Council runs recycling, road safety and public health campaigns under public task. NHS trusts run vaccination and screening campaigns the same way. Under UK GDPR and the Data Protection Act 2018, public task covers processing needed to perform a task in the public interest. Marketing consent sits in a separate lane.

The ICO guidance on direct marketing and the public sector sets the boundary: promotional messages need consent or soft opt-in, service messages do not. That distinction shapes the whole measurement plan.

A bin collection reminder is a service message. A campaign urging residents to download a new app is closer to marketing. Mixing the two in one dataset contaminates the lawful basis.

Practically, Birmingham teams tag every data source with its basis before collection starts. The collection point decides the rules, not the reporting layer. If a source has no basis, it does not get collected.

That checklist keeps the two lanes from merging when a campaign grows. It also gives the information governance team something concrete to review.

Rule two: keep transparency notices specific to the campaign

A generic privacy notice does not satisfy the transparency duty. Birmingham City Council's privacy and data protection pages carry the corporate notice, but each campaign needs its own layer.

The planning privacy statement shows the pattern: it names the service, the data collected and the purpose.

NHS trusts add a clinical layer. A vaccination reminder notice explains that the trust holds the contact details, that the NHS Business Services Authority may process them, and how long they are kept. Residents can act on that. They cannot act on boilerplate.

Specificity also helps measurement. When the notice states the purpose, the analytics team can define a success event that matches it. A notice about recycling contamination supports a metric about bin checks. A notice about screening uptake supports a metric about appointment bookings.

Review notices at the same time as the campaign plan. If the plan changes, the notice changes. The connected disclosures that sit alongside the notice should cover the same ground.

Rule three: measure behaviour change without third-party cookies

Third-party cookies are unreliable for Birmingham campaigns. Chrome's Privacy Sandbox changes the mechanics, and many residents use browsers that block trackers by default. Public sector teams need server-side and first-party methods.

Matomo runs on Birmingham City Council infrastructure for some services, which keeps data on council servers. NHS trusts use the NHS website analytics stack for national pages and local dashboards for trust-level reporting. Looker Studio pulls the aggregates together.

Behaviour change needs a baseline. For a recycling campaign, the baseline might be contamination rates from bin audits. For a vaccination campaign, it might be uptake rates from the child health information service. The campaign then targets a shift against that baseline.

Attribution stays modest. A campaign can show that uptake rose in the postcode areas it covered, but it cannot prove that the leaflet caused the rise. Report the shift with its uncertainty, and keep the method in the reporting dashboard so others can check it.

A worked example: a Birmingham ward runs a recycling campaign for six weeks. Baseline contamination is 24 per cent. Post-campaign audits show 19 per cent. The team reports a five point shift, notes the audit sample size and does not claim causation.

Rule four: publish retention and sharing duties for partners

Campaigns rarely run on one organisation. Birmingham City Council works with contractors, NHS trusts work with voluntary groups, and both share data with evaluation partners. Each sharing arrangement needs a written duty.

Retention periods should be short and stated. Contact details for a campaign may be kept for the campaign plus a short review window. Aggregated results may be kept longer. The notice should say which is which.

Sharing terms should name the recipient, the purpose and the safeguards. If a partner receives personal data, the contract needs a data processing agreement. If a partner receives only aggregates, say so, because that changes the risk.

Birmingham's public campaign context includes a large and diverse population, so small-area data can identify people. Suppress small counts in public reports. The separate gates approach keeps privacy, security and accessibility reviews from collapsing into one approval.

Publish the retention and sharing schedule alongside the notice. Residents and partner organisations can then see the same rules, and the information governance team can audit against a public document.

Tool choice for public campaign measurement

Tool choice should follow the campaign purpose and the data involved. This is not a recommendation list: each body should assess its own lawful basis before adopting a tool.

Birmingham City Council's about Birmingham pages give the local context for these campaigns: a city of over a million residents, with varied language needs and a large retail and manufacturing base. That scale makes first-party data strategies more valuable than third-party tracking.

NHS trusts in the city add clinical governance to tool selection. A tool that touches patient data needs a data protection impact assessment. A tool that touches only campaign contact data may need a lighter review.

Whichever tool is chosen, the notice should name it. Residents who want to know where their data goes should not have to guess.

Common questions