First, read served, measurable, viewable and unmeasurable before judging the rate
Start with the report, not the line item. The viewability and attention guide sorts every impressionOne display of an ad or search listing, as counted by the platform. into buckets: served, which the ad server counted; measurable, where the measurement script could run and decide; viewable, where the answer was yes; non-viewable, where it was no; and unmeasurable, where the script could not run or could not see the ad's container. The viewable rate is normally calculated on the measurable base, so two campaigns can show the same rate while one of them has most of its impressions in the unmeasurable bucket. The first number to read is therefore the measurable rate, measurable impressions as a share of served, and only then the viewable rate on top of it. A campaign with a small measurable base does not yet have a viewability problem; it has a measurement problem, and the fixes below are different.
The DV360 reporting guide shows where Active View's measurable and viewable columns sit, with average time in view beside them, and asks for them per creative size, per exchange and per inventory source rather than per line item, because viewability is a property of the placement far more than of the creative and a line item's average hides which sources are pulling it down. Pull that breakdown for the period in question and read it in two cuts. By environment first: a large unmeasurable share concentrated in app inventory points at publishers without a measurement integration. Then by size and position: a healthy measurable rate with non-viewable impressions clustered in one size or one set of placements points at the slot, the creative or the bid goal rather than at measurement.
Two cautions before the table. The standard defines an opportunity to see, as the viewability glossary entry puts it, and the Media Rating Council, as the site's guides state it, publishes the definition and accredits the vendors that measure against it; it says nothing about whether anyone looked, so a higher rate is a floor cleared, not a result delivered. And a floor that becomes a target distorts behaviour. The Goodhart's law page lists the viewability rate among the targets that drift when chased, because inventory built to score well, sticky units, refreshing slots and made-for-advertisingA site built to sell ad impressions rather than serve readers: thin or scraped content, a high ratio of ads to content and traffic bought rather than earned. pages, is exactly what a campaign optimised on the rate alone finds. The aim of the diagnosis is a rate that reflects the plan's inventory, not the highest rate available. To see what the current figure implies for the price of exposure, enter the all-in cost and the viewable rate into the impression calculator and read the result as the viewable cost per thousand the plan is really paying.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| The viewable rate looks low, or oddly high, and the measurable share is a fraction of served | Measurable impressions as a share of served, broken down by environment and inventory source | A large unmeasurable share, so the rate is read on a small base and the real picture is hidden | Report the measurable rate first; exclude or renegotiate the placements that cannot be measured |
| The measurable rate is healthy and non-viewable impressions cluster in particular sources or positions | Non-viewable impressions by inventory source, exchange and position on the page | Placements and sizes that sit below the fold or in slots the reader never scrolls to | Apply pre-bid viewability targeting at the level the objective needs and prune the worst placements from the placement report |
| A cheap cost per thousand, a high win rate, a long tail of unfamiliar domains and a low rate together | The bid strategy and goal on the line item | A bid goal that counts served impressions, so cheap unseen slots win the auction | Move the line item to the viewable goal documented for its type, or to a fixed bid with a viewability filter |
| One creative or one size drags the line item's average down, or served and rendered counts diverge | Active View by creative and by size; the ad server's count against the DSP's | Heavy or slow creative that renders after the reader has moved on, or a size that serves mostly below the fold | Rebuild the creative within the documented specifications and retire the size that serves low, tested under even rotation |
| Two figures for the same campaign disagree and each side quotes its own | Each vendor's measurable base and the environments it can measure | Two vendors with different measurement footprints compared as if they measured the same impressions | Report each rate on its own base and name the counting source in the contract |
| A deal or line item delivers far below the rate the plan expected and the publisher disputes it | The line item's viewability targeting and the deal's written terms | No predicted-viewability filter or deal terms that never agreed a viewability standard | Agree the standard, the measurement source and the remedy before launch; add the filter only where the objective needs it |
Fix, verify and prevent each cause
A large unmeasurable share hides the real picture
The viewability guide is blunt about the bucket most reports skip: unmeasurable impressions are excluded from the rate entirely, and a large share of them hides what really happened. They are neither viewable nor non-viewable; the script could not run, or could not see the ad's container. The guide names the usual reasons. In-app measurement depends on the publisher integrating a measurement interface, without which the ad is simply unmeasurable, and an ad inside a frame that cannot be inspected becomes unmeasurable rather than viewable or not. Connected television and some audio environments cannot be measured in browser terms at all.
Fix: calculate the measurable rate for the period, break the unmeasurable impressions down by environment, inventory source and publisher, and make a buying decision for each cluster: exclude the placements that cannot be measured, renegotiate them with measurement as a condition, or keep them knowingly and report the viewable rate with its base beside it. Do not raise the viewability threshold to fix an unmeasurable share; a filter cannot act on impressions the measurement never reached.
Verify: in the next period the measurable rate rises, the unmeasurable bucket shrinks to the placements that were kept on purpose, and where a second vendor measures the same campaign the two viewable rates move closer together as the unmeasured inventory leaves the buy.
Prevent: put the measurable rate in every report template beside the viewable rate, so a change in inventory mix shows up as a change in the base before it shows up as an argument about the rate, and ask every new app or connected television source for its measurement integration before it is added to a line item.
Placements and sizes that sit below the fold
The pixel share in the standard is measured against the viewport, not the page, so an ad far down a long article is not viewable until the person scrolls to it, and the time condition is continuous rather than cumulative, so an ad that flashes in and out of view during fast scrolling never accumulates its way to viewability. The brand safety and viewability guide owns the controls that act on this before the bid. Pre-bid, the DSP or a verification vendor predicts the viewability of an impression from the placement's history and bids only where the prediction clears a threshold the buyer sets; DV360 documents this as viewability targeting using Active View's predicted rate, and the Display & Video 360 Help Centre is the documentation for the setting itself. Verification vendors supply equivalent pre-bid segments built from their own measurement.
Fix: read the placement report by spend share rather than by impressions, as the brand safety guide asks, and remove the sources where non-viewable impressions concentrate; then apply viewability targeting at the level the objective needs rather than the strictest level available, because the inventory immediately below a strict threshold is often ordinary editorial placement and the prediction is made from history, so new publishers are under-bought until they have one. Keep a suitability control beside the viewability filter: the guide warns that viewability can be gamed by sticky units, auto-refresh and ad stacking, and a viewability filter without a suitability filter drifts towards exactly the made-for-advertising inventory the brand safety guide describes.
Verify: the non-viewable share by source falls in the next report, and the ReachUnique people who saw an ad. and price per thousand move in the direction the trade-off predicts; if reach collapsed or the cost per thousand jumped, the threshold removed more than the objective needed.
Prevent: decide the threshold from the objective before launch and record the reach and price the plan expects under it, review the placement report monthly by spend, and read a very high rate as a prompt to check where it came from rather than as a success.
A bid goal that pays for served impressions
The bidding and pacing guide describes every automated strategy in DV360 as a goal plus a constraint, and lists viewability and completed views among the signals a goal can use, with the Help Centre as the reference for which goals exist for each line item type. A line item whose goal is served impressions at the lowest cost, or a fixed CPMCost per 1,000 impressions. with no viewability filter, is told to buy the cheapest matching impression, and the brand safety guide explains what that finds: inventory that is cheap, highly plentiful and unseen, and, where the goal is cheap viewable impressions alone, the made-for-advertising sites that score well on the rate while delivering nothing.
Fix: move display line items to the automated goal built on viewability and video line items to the goal built on completed views, as the DV360 brand awareness setup recommends, and keep fixed bids for deal line items where the price is agreed in advance; where a fixed bid stays, add the viewability filter so the cheapest slot is no longer the winner by default. Change the strategy on its own, because the bidding guide warns that each change restarts learning and two changes in one week cannot be attributed.
Verify: once the daily figures stop swinging, read the viewable rate and the viewable cost per thousand per line item against the period before the change; the impression calculator turns the all-in cost and the new rate into the viewable impressions the budget now buys, which is the figure to compare rather than the raw cost per thousand, which always looks better without the control.
Prevent: write the bid goal into the plan beside the viewability expectation it is meant to deliver, so the next trader inherits a decision rather than a default, and treat a goal change as a logged change that the next report must explain.
Heavy or slow creative, or a size that serves mostly below the fold
The standard requires the ad to have rendered: a slot that was filled but never drew the creative does not count as viewable, and the DV360 creatives guide notes that a large gap between the ad server's count and the DSP's points at a tag firing before the ad renders. In general practice a creative that takes long to draw, because its bundle is heavy or its assets load from a slow source, is counted as served by the time the reader has scrolled past it. The creatives guide also makes the size point directly: Active View viewability should be read by creative and by size, since a size that serves mostly below the fold drags the line item's average down, and the viewability guide adds that standard sizes that fit above-the-fold slots are viewable more often than tall units that need a scroll.
Fix: pull the creative report with size and type as dimensions and find the creative or size that sits far below the line item's average; rebuild a heavy or slow creative within the documented specifications the Help Centre gives for each format, and retire or restrict the size that serves low rather than removing it from every line item, since the creatives guide also warns that a line item cannot bid on slots it has nothing to fill.
Verify: run the comparison under even rotation for a test flight, because optimised rotation gives the winning creative the better placements and biases any size comparison, then confirm that the rebuilt creative's viewable rate matches its siblings in the same size and that the served-to-rendered gap narrows.
Prevent: make size-level viewability a line in the creative QA before every flight, plan the size set from the sizes the inventory actually sends, and keep the creative size dimension in the viewability report template so a new size that serves low is seen in its first week.
Two vendors compared as if they measured the same impressions
Active View and an independent verification vendor will disagree on the same campaign even when neither is wrong, because vendors differ in which environments they can measure: one measured impressions the other could not, the unmeasurable share moved, and the rate calculated on each base moved with it. The viewability guide calls the gap a footprint difference rather than a dispute over any individual impression, and the reporting guide places the Active View columns so that both figures can be read together.
Fix: put the two reports side by side with their measurable bases, calculate each viewable rate on its own base and again on served impressions, and break the unmeasurable impressions down by placement to find where the footprints diverge; then decide which source counts for what. The media auditing and verification guide sets the rule for the contract: name an accredited counting source for billing and make-goods, because where the contract names a source that source wins disputes, and where it is silent disputes become negotiations.
Verify: the difference between the vendors is explained by named placements, and once the unmeasured inventory leaves the buy the two rates sit within a tolerance the team has written down.
Prevent: carry both figures and both bases in the report template, record which source is the counting basis for each deal and for the campaign as a whole, and treat a widening gap as a change in inventory mix to investigate rather than a vendor to argue with.
No predicted-viewability filter or deal terms that never agreed a standard
The DV360 inventory and deals guide is explicit that a cost per thousand agreed without agreeing on viewability or environment is a frequent source of disputes after launch, and asks for the line item type, the creative sizes and formats, the geography and any brand-safety requirement to be stated in the proposal. The same guide notes that viewability targeting on the line item applies to deals as well as to the open auction, so the filter cuts both ways: absent, a deal can deliver whatever the publisher's slots happen to achieve; inherited from prospecting defaults, a strict one can stop a deal delivering at all.
Fix: for a running deal, read its delivery per publisher through Active View and compare it with what the proposal said, then renegotiate with the standard, the measurement source and the remedy written in, as the viewability guide asks for any deal that guarantees a viewable rate or sells on a viewable cost per thousand; for an open-auction line item with no filter, add viewability targeting at the level the objective needs. Where a deal line item carries a filter it was never meant to have, lift it for that publisher deliberately and record why.
Verify: the deal's detail page and the line item troubleshooting view show bid requests, bids, wins and served impressions in order, so a filter that is too strict shows up as requests without bids, and the per-deal Active View figures settle at the agreed level within the first weeks.
Prevent: give every deal a row in the build sheet with its viewability term, counting source and remedy beside the deal ID, make that row part of the pre-flight check, and review deal line items against each publisher's actual inventory before inheriting any default.
Prevention: a viewability standard agreed per deal and a measurable-rate check in every report
Every cause above was visible before launch to someone who looked in the right place: the measurement integration a publisher lacked, the slot a size would land in, the goal a line item was left on, the base a vendor's figure was calculated on, the term a proposal left out. The routine that prevents the next low rate is therefore a set of decisions made before the flight and a short reading habit during it.
Before launch, agree the standard per deal and per line item. For each deal, the proposal states the viewability expectation, the measurement source that will count it and what happens if it is missed, and the build sheet carries those terms beside the deal ID so the pre-flight check can confirm them. For each open-auction line item, the plan records the bid goal, whether a viewability filter applies and at what level, and the reach and price the plan expects under those settings, so that a later shortfall in reach can be attributed to the control that caused it rather than blamed on the market. The brand safety guide's sequence is the safe one: monitor first, read a period of verification data to learn where the plan's inventory fails, then apply the filters for the failures that appeared, rather than stacking every control on the first day.
During the flight, read the buckets on a fixed cadence. The awareness setup reviews viewability by inventory source and the verification report every fortnight, and the reporting guide's scheduled offline reports make that cheap: one template per report type with the measurable rate, the viewable rate on the measurable base, average time in view and the creative size, exchange and inventory source dimensions, run the morning after the period closes. Beside the rate, keep the guardrail the Goodhart's law page asks for: an outcome on the advertiser's side, engaged visits, lift or conversions, and a check on where the impressions ran, so that a rising rate bought from refreshing units on unfamiliar domains is caught by the placement report rather than celebrated. The monthly placement review by spend share is where that check lives, and it is the same review that keeps the viewability filter from drifting towards inventory the suitability policy would reject. Where that review shows most of the spend on unfamiliar domains rather than low slots, the diagnosis of programmatic spend going to low-quality sites works back from the spend share.
When several of these habits are missing at once, a single diagnosis is the wrong tool. The programmatic advertising audit scorecard scores brand safety, viewability, deals, frequency and reconciliation in one pass and is the better starting point for a programme that has never agreed a standard. For the measurement questions that sit beside viewability, which impressions were seen by whom and how often, the programmatic measurement section orders them, and the frequency cap diagnosis picks up the complaint that usually follows a reach campaign whose viewability has just been fixed: the same people seeing the now-viewable ad far too often. The programmatic troubleshooting hub lists the other symptom pages for this root.
A hearing aid retailer finds half its problem was unmeasured (illustrative)
Umberhayes Hearing
Umberhayes Hearing, an invented hearing aid retailer with 23 branches, runs a 12-week display campaign in DV360 with a plan that promised a 65% viewable rate. The week-6 report shows 44%, and the agency proposes a strict viewability threshold on every line item.
The trader breaks the report into buckets first. Of 9.4 million served impressions, 7.5 million were measurable, so a fifth of the buy was never assessed; 3.3 million of the measurable impressions were viewable. The viewable rate on the measurable base is 44%, but on served impressions it is 35%, and the unmeasurable 1.9 million come almost entirely from three app publishers with no measurement integration. On the remaining web inventory the creative report shows a tall half-page unit at 31% viewable against 58% for the medium rectangle and leaderboard, because the tall unit serves mostly in slots below the first screen. A third figure had been circulating: the verification vendor's 52%, quoted against Active View's 44% as if the two measured the same impressions, when the vendor had measured only 6.1 million.
The retailer excludes the three app publishers until they can be measured, retires the tall unit on prospecting line items, and moves the open-auction display line items from a fixed CPM to the viewable bid goal, leaving the two preferred deals on fixed bids with a 60% viewability term written into each renewal. From week 7 the measurable rate rises to 94%, the viewable rate on the measurable base to 63%, and the two vendors' figures sit 3 points apart. The weekly report now carries the measurable rate, both viewable rates and the creative size dimension, and the plan for the next flight records a 65% expectation with the goal and the filter that are meant to deliver it.
Common mistakes
- Reading the viewable rate without the measurable rate. A rate on a small measured base describes a minority of the buy.
- Raising the threshold to fix an unmeasurable share. A filter cannot act on impressions the measurement never reached; exclude or renegotiate the placements instead.
- Reporting viewability per line item. It is a property of the placement; read it per source, exchange and size or the cause stays hidden.
- Chasing the rate without a guardrail. The cheapest way to a high rate is inventory built to score well, which is the drift Goodhart's law predicts.
- Comparing two vendors' rates without their bases. The gap is usually a footprint difference, not an error.
- Testing a size under optimised rotation. The winner gets the better placements and the comparison says nothing.
- Agreeing a price without agreeing the standard. A deal with no viewability term produces a dispute rather than a rate.
- Stacking every filter on day one. Reach falls, price rises and nobody can say which control did it.
