Programmatic advertising

How a programmatic buy actually works

Programmatic means buying advertising through automated exchanges rather than by negotiating with publishers. A publisher makes its ad space available through a supply-side platform. Advertisers bid for it through a demand-side platform. When a page or app loads, an auction runs and a winner is served, all inside the time it takes the page to render.

Most of that volume is open auction, where anyone can bid on widely available inventory. Private marketplaces are invitation-only auctions with a specific publisher at an agreed floor price, which buys better quality and better placement. Programmatic guaranteed reserves fixed inventory at a fixed price, which is closer to a traditional media buy run through automated pipes.

The reason to understand this is that the deal type largely decides the quality you get. Open auction is cheap and variable. Private marketplaces cost more and behave far more predictably. For most brands the right answer is a mix, with the brand-sensitive spend going through private deals.

• Open auction: broad reach, lowest cost, most variable quality
• Private marketplace: invited access to specific publishers at a floor price
• Programmatic guaranteed: fixed inventory at a fixed price
• Deal type usually matters more than targeting for quality control

Connected TV and streaming: what you are really buying

Connected TV means ads served into streaming services on a television set. The distinction worth understanding is between that and over-the-top video watched on a phone or laptop, because completion rates and attention differ substantially. A television screen in a living room behaves like television. A video in a browser tab does not.

The commercial difference from broadcast is how it is bought. Broadcast sells slots by programme and audience estimate. Connected TV sells impressions and lets you target by region, household attributes and audience segments. That means a regional business can advertise on a television screen without paying to reach an entire country, which is what changed the accessibility of this channel.

Ads on Connected TV are generally unskippable and run at full screen with sound, which produces high completion rates. Take those completion figures for what they are. A high completion rate on unskippable inventory measures the format, not the interest of the viewer, so judge the channel on business outcomes rather than on completion.

Digital audio and out-of-home

Streaming audio ads run on music services, podcasts and digital radio. The format has a real advantage in that people listen while doing something else, so attention is available at times when no screen is. It also reaches a listener who cannot click, which means the creative has to work on memory rather than response.

That shapes the writing. Audio ads need one idea, a brand name said clearly and more than once, and a way to act that does not require hands. A memorable name is worth more than a URL. Podcast placement adds host-read options, where the credibility of the presenter carries more weight than the production quality.

Digital out-of-home covers screens in transport, retail and street locations, bought programmatically by impression rather than by fortnight. It supports contextual triggers, so a screen can serve a different message by time of day, weather or nearby event. It is a reach and recall channel, and it should be measured that way rather than by clicks that cannot happen.

• Audio: one idea, brand name repeated, no visual dependency
• Podcasts: host-read placements carry credibility beyond the format
• Out-of-home: bought by impression, triggered by time, weather or location
• All three are recall channels, judge them on lift not on clicks

Brand safety, fraud and inventory quality

Open programmatic contains genuinely poor inventory: made-for-advertising sites built only to carry ads, apps with invisible ad slots, and traffic generated by scripts rather than people. Nobody buys this deliberately. It arrives when a campaign is left on broad settings and told to find the cheapest impressions.

The controls that matter are set before launch, not after. An inclusion list naming the sites and apps you are willing to appear on is the strongest single control, and it is the opposite of the usual approach of blocking things after they show up in a report. Exclusion lists handle categories and specific domains you never want. Third-party verification checks viewability and invalid traffic independently of the platform reporting it.

Then there is the quality standard you set. Viewability thresholds, minimum completion requirements on video, and rules about the content categories your ads sit beside. Enforcing these costs money in the sense that quality inventory is more expensive. It saves money in the sense that impressions nobody saw were never worth anything.

• Build an inclusion list before launch rather than blocking after the fact
• Use third-party verification, not just the platform's own reporting
• Set viewability and completion standards explicitly
• Review the placement report weekly and prune it

Audience building without third-party cookies

Third-party cookies, which programmatic audience targeting relied on for years, are restricted across most browsers. That has pushed the whole channel towards methods that do not depend on them, and it changes how audiences should be planned.

First-party data is the strongest foundation. Your customer list, matched into the platform, becomes both an audience and a seed for finding similar households. Contextual targeting, which matches ads to the content of the page rather than the history of the person, has improved considerably and now handles meaning rather than keywords. Geographic and household-level targeting works well for Connected TV, where cookies were never the mechanism anyway.

Identity solutions based on hashed email addresses exist and work where you have consented first-party identifiers to match. They are useful, but they are not a like-for-like replacement, and any plan that assumes cookie-era targeting precision will disappoint. Planning around contextual, geographic and first-party approaches produces a more honest forecast.

Measuring channels that do not produce clicks

Click-through rates on display, Connected TV and audio are close to meaningless. Nobody clicks a television. The few clicks display generates are frequently accidental. Judging these channels by clicks will lead you to switch off the ones that work and scale the ones serving worthless inventory.

The measures that hold up are different. Incrementality testing, where a comparable group is deliberately excluded and outcomes compared, gives the cleanest read on whether the spend caused anything. Geographic holdout tests do the same at region level and are practical for smaller budgets. Brand lift studies measure awareness and consideration changes directly.

Alongside those, watch the business signals. Branded search volume, direct traffic and enquiry volume during and after a flight. None of these alone proves anything, but together and across several campaign periods they build a reliable picture. We agree the measurement approach before the campaign runs, because deciding how to judge a reach channel after it has finished never ends well.

• Ignore click-through rate as a performance measure on these channels
• Use geographic holdouts, they work at modest budgets
• Track branded search and direct traffic through the flight
• Agree the measurement plan before launch, not after