Connected television has grown rapidly and sits awkwardly between two disciplines with different assumptions. Most of the difficulties in the channel come from applying the wrong set.
The viewing context is television
The thing digital buyers underestimate.
It is a large screen, viewed at a distance, frequently by more than one person, in a lean-back mode, with the advertisement unskippable and full screen.
Which means the attention profile is completely different from a video unit in a feed. Exposure is longer, the screen is not shared with other content, and the sound is generally on.
Creative made for feed placement — fast cuts, burned-in captions, square framing, the message in the first two seconds — looks wrong on a television and performs accordingly.
The creative that works here is television creative, which is a different craft with a different budget expectation, and it is the most common mistake I see.
The buying mechanism is digital
The thing television buyers underestimate.
Inventory is bought programmatically, in auctions, with audience targeting, at an impression level. All the machinery of digital buying applies, including the intermediary layers, the fee stacking and the supply path issues.
Which means television buyers arriving in the channel encounter problems they have never had to think about — undisclosed fees, resold inventory, duplicate paths, and inventory of uncertain provenance.
Fraud exists here too, and has taken forms specific to the channel, including spoofed device and app identifiers claiming to be premium inventory.
The verification and supply path hygiene that digital buyers have developed is directly applicable and frequently not applied, because the channel is perceived as premium.
The measurement fits neither
Where the genuine difficulty sits.
Television measurement is panel-based and reports reach and frequency against a population. Digital measurement is census-based and reports impressions and attributed outcomes.
Connected television produces impression-level data, which invites digital-style measurement, and the outcomes it drives are frequently not trackable to the device, because the person then acts on a different device or in a shop.
The result is that attributed conversions substantially understate the effect, while reach and frequency figures are difficult to reconcile with linear television data because the identifiers are different.
Practically, this means the channel is best assessed with television methods — reach, frequency, brand measurement, incrementality tests — rather than with click-based attribution, and buyers accustomed to the latter find that unsatisfying.
The frequency problem specifically
The most common complaint from viewers and a real operational failure.
Because inventory is bought across many apps and platforms, each with its own frequency management, there is no single view of how often a household has seen an advertisement.
The consequence is the widely reported experience of seeing the same advertisement repeatedly within one viewing session, which is worse than wasted spend because it produces active annoyance.
Partial solutions exist — buying through fewer paths, using platforms that offer cross-app frequency control, negotiating caps directly with publishers — and none of them fully solve it in an open programmatic environment.
This is a strong argument for consolidating spend with fewer suppliers in this channel, at some cost in reach.
Household versus individual
A conceptual point that affects targeting accuracy.
The identifier available is generally a device or household, not a person. Several people watch the same television.
Targeting that assumes an individual is therefore approximate, and audience segments applied at household level are broader than they appear.
This is well understood in television and frequently forgotten by digital buyers used to individual-level targeting, who then over-narrow their audiences and pay premiums for precision that is not achievable.
What I would advise
Make the creative for a television, not for a feed.
Apply digital supply path discipline, because it is a programmatic channel whatever it looks like.
Measure it as a brand channel with incrementality testing, not as a performance channel with attribution.
Consolidate suppliers for frequency control, accepting narrower reach.
And target more broadly than instinct suggests, because household-level identification does not support fine segmentation and the premium for it is real.
The inventory quality spread
A final caution that applies specifically to open programmatic buying here.
The channel contains everything from premium broadcaster inventory to ad-supported apps of very low quality, and both are available through the same pipes with similar descriptions.
App-level reporting is the minimum diligence, and it is frequently not provided by default.
Requesting it, reviewing the tail, and excluding aggressively is the same discipline as display placement reporting, and it is skipped more often here because the channel is assumed to be premium by nature.
Reconciling with linear
The planning problem for anybody buying both.
Linear and connected inventory reach overlapping audiences and are measured by different systems with different identifiers, so total unduplicated reach across the two is genuinely difficult to establish.
Cross-media measurement initiatives exist in several markets and adoption is partial.
Until that resolves, the practical approach is to plan them together against a single reach target using whatever panel-based estimate is available, and to accept that the number is approximate rather than pretending precision that does not exist.