The parts of UK media nobody writes down.
- guides, each naming its own sources and its own limits
- 16
- words, counted from the guides rather than claimed
- 23.5k
- questions answered across them
- 31
- campaign outcomes cited, because there are none yet
- 0
Three jobs, 16 guides
Each guide states the thing it will not claim as plainly as the thing it will, and names the code, the data module or the research report every figure came from. None of them describes a campaign that has run, because none has.
Buying the awkward channels
The UK channels whose buying process is not an API, and the vocabulary that is not the one you asked in. What you are actually selecting, what it costs, and what the paperwork does to your timeline.
How UK addressable TV targeting actually works
You are buying households whose profile over-indexes on your audience. That is worth knowing before you write the brief.
Anyone weighing addressable TV for a UK campaign for the first time, and anyone who has been shown an addressable proposal that implies job-title targeting and wants to know what the deck actually means.
- Fidelity
- 55 / household, fixed
- AdSmart minimum
- £3,000 per burst
How UK cinema advertising is really booked
Two sales houses, no buying API, and a calendar that starts on a Friday.
Anyone who has been quoted for cinema and cannot tell what they are being sold, and anyone building a multi-channel plan who wants to know what cinema adds to it and what it costs in process.
- Fidelity
- 40 / venue, fixed
- Lead time
- ~18 days incl. clearance
Buying UK DOOH without a movement-data licence
A billboard cannot know who you are. It can know who tends to walk past it, and when.
Anyone reading a DOOH proposal that claims to target a professional audience, and anyone planning out-of-home who wants to know which parts of the pitch survive scrutiny.
- Fidelity
- 50 / geo-cohort, fixed
- Screen ranking
- Empty by design
What a DSP is, and whether you need one
The fee is the easy part to compare. The harder question is whether you have anyone to drive it.
Marketing leads being told they should have a DSP, and anyone comparing a platform fee without a clear view of what the platform is for.
- Observed fee range
- 5%–20% of spend
- The Trade Desk
- ~20.3% in 2024
UK media buying for B2B marketers
The committee is six to ten people. You are reaching two of them.
B2B marketers with a working LinkedIn and search programme who have been told to find growth beyond it, and who need to know what the next channel actually delivers.
- Best B2B proxy
- CTV, 85 / person
- Addressable TV
- 55 / household
How to read a media plan
The interesting fields are the ones most plans leave out.
Anyone approving a media plan they did not build, and anyone comparing two proposals that look similar and are not.
- Fields to demand
- Six
- Most often missing
- Data cost per line
Choosing between cinema and CTV
They are not the same buy, and the deciding factor is rarely the price.
Anyone with a video budget deciding where it goes, particularly on a launch where both are on the shortlist.
- CTV fidelity
- 85 / person, best case
- Cinema fidelity
- 40 / venue, fixed
Sky AdSmart explained
Household attributes, postcode districts, and a booking process rather than an auction.
Advertisers considering addressable television for the first time, and anyone reading an AdSmart proposal that implies more precision than the vocabulary supports.
- What it selects
- Households, by attribute
- Fidelity
- 55 / household, fixed
Knowing what you bought
The two numbers a media plan should carry and almost never does: how far the targeting is from the audience you described, and whether the campaign can support a causal claim at all.
What a fidelity score is, and why anyone should publish one
One number, one question: how much of the audience you described survived the translation.
Anyone comparing multi-channel media proposals and trying to work out which of the targeting claims in them are real, and anyone who has to defend a channel mix to somebody more sceptical than they are.
- Range published
- 85 down to 40
- Grades
- Five, person to venue
Designing a geo-lift test that can actually be powered
The design decides the answer. Run the power calculation before the flight, not after it.
Anyone who has been asked to prove that media worked, and anyone about to commission an incrementality test who wants to know before signing whether it can produce an answer.
- Power gate
- α 0.05 · 80% power
- Markets in the table
- 12 UK conurbations
Why most incrementality tests are invalid
An underpowered test does not give a weak answer. It gives a random one.
Anyone who has been handed an incrementality result and is not sure how much weight it carries, and anyone about to commission one.
- Standard significance
- 5%
- Standard power
- 80%
Running it safely
What changes when an agent holds the keyboard, and which controls have to exist before that is a good idea rather than an expensive one.
Running paid media from an AI agent without losing control
The agent proposes. A named human holds the money. Everything else is implementation detail.
Anyone considering letting an agent operate an advertising account, and anyone who has to sign off on that decision for somebody else's budget.
- Human-only operations
- 11, with no agent tool
- Authority
- Mandate: ceiling, expiry
Frequency capping across channels, and why it usually is not happening
Seven platforms each capping at four is not a cap of four.
Anyone running three or more channels who has been shown a combined reach number, and anyone who suspects their launch is hitting the same people repeatedly.
- The problem
- Caps are per-seller
- What it needs
- One exposure record
What Clearcast and the CAA actually check
The clearance body is not judging your creative. It is asking for your evidence.
Anyone writing television or cinema copy for the first time, and anyone whose flight date depends on a clearance turnaround they have not budgeted for.
- Broadcast code
- BCAP
- Non-broadcast code
- CAP
MCP for marketers: what the protocol actually changes
The protocol is a door. What matters is what is behind it.
Marketers being pitched agentic tooling, and anyone deciding how much authority to give an assistant over a media budget.
- What it is
- An integration standard
- What it is not
- A capability or a moat
First-party data in UK media buying
The bridge from your database to a television exists. Most of it is not built yet.
Marketers with a CRM they want to activate beyond the social platforms, and anyone evaluating a vendor's first-party data claims.
- Where it works easily
- Social platforms
- Where it is hard
- CTV, audio, OOH
If the guide you want is the one about your own campaign, that is a working session rather than a page.
Talk it through31 questions, and the guide that answers each
Every question these guides answer, listed rather than buried, because a title is a category and a search is a question.
The site-wide question index at /faq carries these alongside every other question this site answers, grouped by subject and deduplicated, so an answer appears in exactly one place.
Or describe the problem and let it point you at the guide
Grounded on these 16 guides and the pages behind them, so it answers from what is published rather than from what would be convenient. It will send you to the guide rather than summarising it away.
Prefer to talk it through? Book a working session→
Where every number in them comes from
These are written for the buyer rather than for the product. Every figure in them traces to this site's own data modules, to our code, or to our research reports, and each guide names its sources at the foot of the page. Where the honest answer is that we do not do something — book a cinema campaign, rank a screen by movement data, buy ITV inventory — the guide says so in the section where it comes up rather than in a disclaimer nobody reads. The test applied to each of these is whether it would still be worth reading by somebody who never books a session.
- This site's own data modules
- Fidelity scores, plan floors, the fee, the supply grades and the measurement model are published elsewhere on this site and quoted here rather than restated. A figure that lives in a data file is owned by that file, so a guide cannot drift from the page it describes.
- Our code
- Mechanisms — the power calculation, the market matching, the empty screen ranking, the minimum burst check, the workflow steps — are read off the code that implements them. Where our README and our code disagree, the code wins. It has disagreed at least once: the README listed a connector as having a live client, and every one of that client's live methods throws.
- Our research reports
- Market facts such as cinema rate structures, clearance costs and sales-house shares come from our own research reports, which carry inline citations to the published material they were taken from. Those were compiled in August 2026 and vendor terms move, so anything drawn from them is written as a dated survey and the guide says to check the current rate card.
- Nothing else
- No customer outcomes, no case studies, no lift figures, no invented statistics. Nothing here has spent money through a live rail, so any number that looked like a campaign result would be deterministic sandbox output, and none is printed.
Would this still be worth reading by somebody who never books a session?
And if a guide is wrong about something you know better than we do, that is worth forty-five minutes of anyone's time.
Talk it throughOr skip the reading and bring a real audience.
Forty-five minutes. Everything in these guides — the fidelity scores, the plan floors, the power calculation, the clearance steps — runs against your own brief in front of you rather than being described.
Book a working session45 minutes. Bring a real brief and we compile it live. · Design-partner phase · the sandbox needs no card and no credentials
Questions about the guides
Where the numbers come from, whether any of this describes a campaign that has run, and whether the guides are any use to somebody who never buys anything here.
Are these guides about AdBuyMCP or about the channels?
About the channels and the methods, with AdBuyMCP as the worked example. Each guide is written so that a buyer using a different platform, an agency or a sales house directly still gets the useful part: what an audience pack is, why a power calculation has to happen before the flight, what a venue type claims and what it does not. Where the product's own behaviour is the point — an empty screen ranking, a refusal to report an underpowered lift — the guide says which behaviour belongs to this platform rather than to the channel.
Where do the numbers in these guides come from?
Three places, and every guide names its own at the foot of the page. Figures this site already publishes elsewhere — fidelity scores, plan floors, the fee, supply grades — are quoted from the data module that owns them rather than restated. Mechanisms are read off our code, which is the authority whenever our README disagrees with it. And market facts such as cinema rate structures and clearance costs come from our own research reports, which were compiled in August 2026 with inline citations to the sales houses' published material, so they are written as a dated survey rather than as a standing fact.
Do these guides describe campaigns you have run?
No, and that would be the more flattering answer. AdBuyMCP has not spent money through a live rail, has no case studies and has produced no causal result. What these guides describe is how the channels work, what our code does, and where its limits are. Any number that looks like a campaign outcome would be deterministic sandbox output, and none is printed here for that reason.
Can I use these without buying anything?
Yes, and that is the intent. Nothing in them requires us: the geo-lift arithmetic works in a spreadsheet, the cinema calendar applies whoever books it, and the questions to ask an agent-operated media platform are worth asking of any vendor including this one. If a guide is only useful to somebody who has already bought the software, it has been written badly.