How to audit a UK podcast network's audience and attribution claims

What to take away

What an audience and attribution audit must prove before it starts

An audit is a test of a claim, not a general inspection. Write the claim down first. If a network says it reaches 400,000 monthly listeners in the North West, that is the claim. Everything you collect either supports it, contradicts it or leaves it unproven.

Fix three things in writing before you request data: the audit period, the metric definitions and the acceptable evidence. A download, a stream start, a completed listen and a unique listener are four different numbers. Networks rarely lie about which one they mean, but they choose the flattering one.

Set the audit period to a closed quarter. Live dashboards move, and a seller can regenerate a screenshot on the day you ask. Closed periods match RAJAR survey windows and give you a stable comparison.

The scope should cover the network, its sales house and any agency in the chain. Most disputes in UK podcast buying are not about the audience. They are about which entity contracted to deliver it and which entity holds the data.

Agree the evidence standard up front. Dashboards are useful, invoices are better, and raw log files with timestamps are best. If a network will not provide logs, say so in the report rather than treating a screenshot as proof.

Decide who signs off. Auditors who send findings to a junior planner get a defensive reply. Auditors who send findings to the commercial director named on the insertion order get data.

The four questions the audit answers

Using Podsights and Spotify Ad Analytics as the measurement spine

Podsights is the measurement layer most UK networks bolt onto their hosting. It works by placing a pixel on the advertiser's site and matching it against the IP addresses and user agents that requested the podcast ad file. That is a probabilistic match, not a census.

Your first check is whether the pixel fired at all. Ask for the raw event count, not the dashboard total. A pixel that fires on every page view inflates conversions, and a pixel behind a consent banner may fire on almost nothing.

Spotify Ad Analytics UK reporting gives you a second view for inventory bought through Spotify's own ad platform. It reports impressions, completion rates and, where the advertiser has connected it, conversion data. Treat it as the platform's own account of its own delivery.

Compare the two where they overlap. If Podsights reports 60,000 impressions and the platform reports 95,000 for the same flight, you have a definitional gap or a duplicated delivery. Both are worth writing down.

The IAB UK measurement hub sets out the industry position on digital ad measurement standards, including the expectation that metrics are defined and comparable across buys. Use it as the reference when a network argues that its own definition is normal practice.

Podsights measurement checks

Spotify Ad Analytics UK reporting

Spotify's own reporting is strongest for reach and frequency within its logged-in audience. It is weaker for open podcast feeds, where the platform is one of several apps serving the same episode.

If a network sells Spotify inventory and open-feed inventory together, ask for the split. A blended completion rate hides the fact that one half of the buy performed and the other half did not.

For a fuller picture of how each model behaves, see what each podcast advertising attribution methods approach means for your numbers, particularly the difference between platform-reported and third-party-measured delivery.

Checking network claims against dated RAJAR comparisons

RAJAR is the joint industry currency for UK radio audience measurement, run by Radio Joint Audience Research and published quarterly. It is the closest thing the market has to an agreed number for audio reach.

Podcast networks are not measured by RAJAR in the same way radio stations are, so you cannot read a podcast figure straight off a RAJAR table. What you can do is compare the network's claimed reach in a region with the total audio audience RAJAR reports for that region in the same quarter.

If a network claims 300,000 weekly listeners in Wales and RAJAR's quarterly release shows the total radio audience in Wales is a fraction of that, the claim needs explaining. It may be cumulative, it may be global, it may be a download count. Ask which.

Use the dated RAJAR quarterly listening data and record the publication date and survey period in your working papers. A comparison without a date is not evidence.

Regional splits matter more in the UK than in most markets. Greater London, the North West, Scotland and the West Midlands all have distinct listening patterns, as do Yorkshire and the Humber, the South East, Wales and the North East. A national average tells a buyer almost nothing about a regional test.

Be careful with age and demographic cuts. RAJAR reports by broad age band, and a network's claim of a "hard to reach 25 to 34 audience" is only checkable if the network will share its own demographic source.

How to run the comparison

Companies House filings as a due diligence layer on networks and agencies

Audience data tells you whether the buy worked. Company data tells you whether the counterparty can honour the rest of the contract.

Start with the registered name on the insertion order. Buyers frequently contract with a trading name that belongs to a different legal entity, which matters if you need to enforce anything.

Check the filing history for late accounts, changes of name and dormant periods. A network that was dormant eighteen months ago and now claims a large audience is not necessarily dishonest, but the gap needs an answer.

Look at the directors and the persons with significant control. Shared directors between the network and the sales house are common and legitimate, but they change how you read an arm's length negotiation.

The Companies Act 2006 sets the disclosure and reporting duties that produce these filings, so you can rely on them as a statutory record rather than a marketing document. The full text sits at Companies Act 2006 if you need to cite the duty in a report.

Also check the advertiser, not just the network. The ASA maintains a list of Non-compliant online advertisers - ASA | CAP, and a brand appearing there changes the risk profile of the campaign you are auditing.

For a structured approach to the wider supplier question, our guide to podcast advertising vendor due diligence covers the documents to request before contracting.

What to pull from the register

Reconciling attribution windows, postcodes and device data

Attribution window reconciliation is where most audits find their material error. A network reporting 900 conversions on a 30-day click window and an advertiser reporting 300 on a 7-day window are both telling the truth about different things.

The window must be agreed before the campaign, not after. If it was not, reconstruct it from the insertion order and the platform settings, and state clearly which window your numbers use.

Postcode data is the second check. Podsights and similar tools can return geography based on IP address, which is coarse and sometimes wrong. A campaign targeted at Greater London that reports 40 per cent of its listeners in Scotland has a targeting or a measurement problem.

Device data is the third. A large share of impressions attributed to desktop browsers is unusual for podcast listening, which is overwhelmingly mobile. Ask what a desktop impression means in this dataset.

Watch for double counting across apps. A single listener using Apple Podcasts and Spotify will appear twice if the network counts downloads per app rather than per person.

Ask for the reconciliation between platform-reported impressions and Podsights-matched impressions, with the gap explained line by line. A gap of a few per cent is normal. A gap of 40 per cent is a finding.

Under UK GDPR, podcast ad measurement that involves tracking individuals engages the rules on online tracking. The ICO's guidance on Online tracking | ICO explains the consent and transparency expectations that apply. An audit that ignores this is incomplete, and a network that cannot describe its lawful basis is a risk.

Our piece on how to spot podcast advertising measurement mistakes lists the specific patterns that show up most often in UK network data.

A worked reconciliation

Take a six-week flight with 1.2 million platform-reported impressions and 780,000 Podsights-matched impressions. The gap is 35 per cent. Split it: 12 per cent from app-level duplication, 15 per cent from consent-blocked measurement, 8 per cent unexplained.

The unexplained 8 per cent is your finding, and it is small enough to be a data quality note rather than a fraud allegation.

Writing the audit report so a buyer can act on it

The report has one job: let a planner decide whether to renew, renegotiate or walk. Structure it around the claims you tested, not around your method.

Open with a one-page summary that states each claim, your finding and the confidence level. Use plain language. If the audience claim is broadly supported but the regional split is not, say exactly that.

Put the evidence in an appendix with file names, dates and the person who supplied each item. An audit that cannot be reproduced by a colleague is an opinion.

Separate measurement error from commercial risk. A 10 per cent impression gap is a measurement note. A sales house that is not the contracting entity is a commercial risk. Buyers act on the second and file the first.

Give a clear recommendation per line item: continue, continue with conditions, renegotiate or stop. Auditors who hedge get ignored.

If the client needs ongoing visibility rather than a one-off report, the principles in our guide to building a podcast advertising reporting dashboard translate directly into the monitoring section of the audit.

Report structure

Red flags that end a network conversation early

Some findings are not worth negotiating. Treat these as terminal unless the network can produce documentary proof on the spot.

Refusal to share raw logs is the clearest signal. A network that will only send dashboard screenshots is asking you to trust a system you cannot inspect.

Audience numbers that change between meetings without a new RAJAR release are another. Numbers move when data is published, not when a buyer asks a hard question.

A mismatch between the contracting entity and the entity holding the measurement account is a structural problem. It usually means the data is not yours to audit and never will be.

Claims of regional targeting that the geography data contradicts, particularly a London-targeted campaign reporting most of its audience outside the South East, should stop the buy until explained.

Attribution windows that quietly extend between the proposal and the report are the most common form of inflation. If the proposal said seven days and the report uses thirty, the conversion number is not comparable.

Finally, a network that cannot name its lawful basis for tracking under UK GDPR, or that dismisses the question, is not ready for a brand campaign. The ICO position is public and the answer should be immediate.

Common questions

How long should a podcast network audit take?

Four to six weeks for a single network with one campaign in scope. Most of the time goes on waiting for data exports and company records, not on analysis.

Can I audit a network without Podsights access?

Yes, but you will rely on platform reporting alone. Ask the network to grant read-only access to the measurement account, or to export raw events, before you accept a screenshot.

Is RAJAR data valid for podcast-only networks?

Not directly. RAJAR measures radio listening, so use it as a regional benchmark for total audio reach rather than as a like-for-like podcast figure.

What attribution window should we standardise on?

Seven days for click-based measurement and one day for view-based, unless the campaign has a long consideration cycle. Whatever you choose, apply it consistently across all networks in the buy.

Do we need to check Companies House for small networks?

Yes. Small networks are where late accounts and dormant periods show up most often, and the check takes minutes.

What if the network disputes our findings?

Send the comparison and the raw files, ask for a written response within five working days, and record their answer in the appendix. Disagreement is normal; refusal to engage is a finding.