Display Advertising

Where Display Fits

Search captures existing demand. Display creates and sustains it, and retargets people who have already shown interest.

Judging display on last-click conversions will almost always make it look like a failure. Its contribution is largely in assisted conversions, branded search lift and retargeting efficiency.

The Google Display Network

The GDN spans millions of sites and apps. Scale is not the problem; quality control is. Default settings will place ads across a very wide inventory pool including low-quality apps and made-for-advertising sites.

Practical controls: exclude mobile app categories unless apps are deliberately part of the plan, add placement exclusions and review the placement report weekly, exclude sensitive content categories, and set a frequency cap.

Targeting Options

Contextual — by page topic or keyword. Increasingly important as third-party cookies decline.

Audience — affinity, in-market, custom segments and remarketing lists.

Placement — specific sites or apps you select.

Demographic — layered on top rather than used alone.

Layering narrows reach quickly. Stacking three targeting types often reduces a campaign to almost no eligible inventory.

Creative That Works

Display creative is seen peripherally for a fraction of a second. It must communicate one idea.

What holds up: a clear brand identifier, one benefit or offer, a single call to action, and strong contrast. What fails: multiple messages, small text, and creative designed to be read rather than glanced at.

Responsive display ads assemble combinations from supplied assets. Provide the maximum number of headlines, descriptions and images and let the system test, but review the asset performance report — weak assets drag the whole set.

Measurement

Track view-through conversions alongside click-through, while treating view-through with appropriate scepticism — an impression served below the fold to a disinterested user is not causally responsible for a later purchase.

The most honest measurement is a holdout test: withhold display from a matched segment and compare outcomes. It is the only method that isolates genuine incremental effect.

Making Display Accountable

Display carries a reputation for being unaccountable, and most of that is a measurement choice rather than a property of the channel.

Separate view-through from click-through conversions in every report. Blended, they overstate the channel; separated, both numbers are usable. This single change alters most display debates.

Set a view-through window you can defend. The defaults are generous, and a long window on a high-reach channel credits purchases that would have happened regardless.

Exclude your own converters and recent visitors from prospecting campaigns, or prospecting quietly becomes remarketing with a prospecting label and excellent-looking numbers.

Run a holdout for anything substantial. A matched-region test is the only method that answers whether display created demand rather than observed it.

Done that way, display is a legitimate upper-funnel channel with honest numbers. Done the default way, it produces impressive reports that fall apart the first time anyone tests them.

Placement Hygiene as a Standing Task

The largest source of wasted display spend is inventory nobody chose. Automatic placement expansion will find cheap impressions, and cheap impressions are cheap for reasons.

A routine that works, run monthly rather than at launch:

• Sort the placement report by spend and exclude anything that has spent meaningfully with no conversions and no engagement.
• Exclude app inventory unless you deliberately want it. A large share of accidental display spend is mobile games, where the click is usually a mis-tap.
• Check content categories, not just individual sites — the blunt category exclusions catch what a site-by-site list never will.
• Look at the distribution, not the average. A handful of placements consuming most of the budget is normal; the question is whether they are the ones you would have picked.

And treat a placement that performs implausibly well as something to investigate rather than to scale. The cheapest, best-performing inventory in the report is where to look first for invalid traffic.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.