Display advertisers using AdRoll paid 45.0% less per thousand impressions to reach new audiences this summer than a year earlier, according to the company's quarterly State of Digital Advertising Report, which AdRoll announced on September 22, 2026. The figure covers July 1 through September 8, 2026. Over the same window, display retargeting CPMs fell 29.1% and account-based marketing CPMs rose 4.4%, the only one of the three categories to move upward.

The report page itself carries no publication date. It is labelled "2026 Q3 Report Highlights for the Holiday Season" and pairs the cost data with US inflation, consumer sentiment and card spending figures. Read next to a September 2 blog post from Pennock, a beauty and skincare ecommerce agency, it describes a market moving in two directions at once: cheaper display inventory on one side, and on the other, Meta auction prices that the agency says rose about 35% across its clients over July and August.

In Short

AdRoll, a company that places ads for businesses, found that showing display ads to people who had never visited a brand's website cost 45% less this summer than a year earlier. That matters because ad prices usually climb from October toward Black Friday, so a cheaper starting point changes how much reach a holiday budget can buy. But one agency working with beauty brands saw its Meta ad prices rise about 35% over a similar period, so cheaper display ads do not mean every channel got cheaper.

What AdRoll measured

AdRoll's report tracks one metric across three campaign types: average CPM, the cost of showing an ad one thousand times. According to the report's FAQ, the data comes from "AdRoll's performance statistics of more than 20,000 online businesses," and the report is updated quarterly. The three categories are display prospecting, which targets people who have not yet interacted with an advertiser; display retargeting, which reaches people who have; and account-based marketing, which AdRoll defines as "a business-to-business marketing strategy that targets a specific group of high-value accounts."

The window is described as "Q3-to-date." By PPC Land's count, July 1 through September 8 covers 70 of the quarter's 92 days, about 76%. Each category is compared twice: against the same dates in 2025, and against what AdRoll calls "the comparable Q2 period" or "the matched prior-quarter period." The report does not give the exact start and end dates of that Q2 baseline.

According to AdRoll, CPM "is determined by supply and demand," and "changes in CPM are often driven by advertisers' demand for ads." By that framing, a 45% fall in prospecting prices reflects some mix of fewer bids and more available impressions. The report does not say which of the two dominated.

Prospecting: the biggest fall, slowing pace

From July 1 through September 8, display prospecting CPMs fell 45.0% from the same period in 2025 and 25.5% from the comparable Q2 period, according to the report. AdRoll reads the gap between those two figures as a sign of deceleration: "The smaller sequential decline shows that the pace of change slowed after the much larger year-over-year shift."

The prospecting chart on the report page plots monthly values for January through August of both years without a numeric axis. The 2025 line climbs to a June peak and holds near that level in July and August. The 2026 line peaks lower in June, then drops sharply in July and slips further in August.

The decline did not start this quarter. In AdRoll's Q2 2026 edition, announced on June 9, prospecting CPMs for April and May were 48% below the prior year. In the Q1 2026 report, covering January and February, the year-over-year decline was 11%. Going back further, the Q4 2025 report had prospecting CPMs up 2% on average for September through November 2025.

Retargeting reverses

The retargeting figures mark a sharper change of direction. According to the report, display retargeting CPMs fell 29.1% year over year in the July 1 to September 8 window and 40.2% from the comparable Q2 period. A year ago, the trend ran the other way. AdRoll's Q4 2025 data showed retargeting CPMs up 11% for September through November 2025, and its Q1 2026 data showed them up 18% in January and February. In the Q2 2026 announcement, April and May retargeting CPMs were still 12% higher than a year earlier.

In March, AdRoll chief executive Vibhor Kapoor described the widening gap between retargeting and prospecting costs as a structural shift in how advertisers build demand. Six months later, both lines point down.

The report offers a narrow explanation for the retargeting move. "Retargeting prices against a finite, already-engaged audience," it says, and "part of this move likely reflects audience availability as much as broader market pricing." In other words, the size of the pool of people who have already visited an advertiser's site can push retargeting prices around independently of the wider auction. The report does not quantify that effect.

ABM moves the other way

Account-based marketing CPMs "rose 4.4% year-over-year, the opposite direction of the other two channels, though they were still down 15.1% from the matched prior-quarter period," according to the report. AdRoll attributes the different behaviour to audience structure: "ABM, like retargeting, is priced against a narrow, named-account audience rather than the open market."

The ABM chart shows 2026 running above 2025 for most of the first seven months, with a 2026 peak in June, before the two lines converge in August. That pattern also differs from earlier editions. The Q1 2026 report had ABM CPMs down 8.5% year over year, and in Q4 2025 the ABM decline had narrowed to 10% in November from 23% in August. AdRoll's ABM product was formerly a separate brand, RollWorks, until the company folded it into AdRoll ABM in August 2025.

What the report leaves out

Several elements that would allow independent checking are missing from the page. There is no dollar CPM for any category, only percentage changes and unlabelled charts. The geography of the CPM data is not stated, although every economic indicator in the report is American. The page describes its base as "more than 20,000 online businesses" in the FAQ, but its introduction refers to "tens of thousands of online businesses." Nothing explains how campaigns were weighted, whether the mix of advertisers changed between 2025 and 2026, or which inventory sources sit behind "display."

AdRoll is also explicit that its economic data and its CPM data do not prove each other. The report says the combined datasets "do not establish causation."

The economic backdrop

AdRoll places the CPM figures alongside three sets of US indicators.

On prices, the report states that the US inflation rate "held at 3.4% in August 2026," unchanged from July, with core inflation easing to 2.4% and energy prices up 16.3% year over year. The accompanying chart is labelled U.S. CPI-U, but the page does not name the agency behind the numbers.

On confidence, the University of Michigan Index of Consumer Sentiment fell to 47.8 in September, down 7.5% from August and 13.2% from a year earlier, according to the report. Year-ahead inflation expectations rose from 4.0% to 4.6%.

On spending, the report cites Bank of America's Consumer Checkpoint: card spending per household rose 0.9% month over month and 4.5% year over year in August 2026. General merchandise and big-box retailers outperformed "more traditional retailers" during back-to-school shopping. The report sums it up in one line: "Inflation held steady and consumer sentiment weakened, but Bank of America card data showed spending per household remained solid in August as shoppers continued to favor value-oriented retailers."

From those numbers, AdRoll draws the conclusion that "holiday advertisers have cheaper access to audiences that are still spending." It adds a qualification: "Lower prices make reach easier to buy, but cautious consumers raise the bar," and shoppers "may respond more selectively to price and value." That echoes the previous season, when Google said 61% of shoppers were more selective with spending because of concerns about economic conditions.

AdRoll's holiday guidance

The second half of the report turns from data to recommendations. According to AdRoll, "CPMs typically rise through early Q4 and peak near Black Friday and Cyber Monday," so the current price level represents a window before seasonal competition sets in.

Kapoor framed it in the report this way: "With CPMs down across prospecting and retargeting, marketers have more room to test before holiday competition intensifies." He added: "Compare the cost of reaching each audience with the engagement, conversions, and revenue that follow. Use that full picture to reallocate budget toward the channels producing the strongest business results."

For ABM, the report suggests evaluating results after the season "looking holistically at engagement, pipeline, and revenue." For measurement in general, it says to "judge each campaign by the role it plays," using "consistent tracking for reach, engagement, conversions, pipeline, revenue, and incremental return."

Off-site discovery

A section on off-site awareness argues that shoppers now research across AI assistants, search, social, CTV and marketplaces before reaching a brand's own site. It cites a Google and Tracksuit finding, reported by Inside Radio, that "growth in awareness produces measurable changes in search behavior and, ultimately, market share." AdRoll proposes branded search volume and direct site visits as the signals that show whether awareness spending is reaching the site. AdRoll itself has been extending into new surfaces: in August it began piloting ChatGPT ads for a selected group of AdRoll and AdRoll ABM customers.

What retailers say about retargeting

The retargeting section draws on a separate eTail and AdRoll study of 100 senior retail and ecommerce leaders. The report does not say when the survey was conducted. According to the findings, 99% of respondents said they were at least somewhat effective at re-engaging visitors, yet only 13% brought back more than 20% of cart abandoners. Coordinating messaging and timing across channels ranked as the top challenge, and respondents repeatedly identified the first 24 to 48 hours after abandonment as a "key window."

Asked which signals matter most for retargeting, 67% named CRM or loyalty data, 56% on-site browsing behaviour and 55% cart abandonment. The report also makes the case for stopping conversion ads once a purchase happens: "Once the shopper buys, suppress further conversion ads."

A different picture on Meta

Pennock's post, written by founder Nikki Lindgren and dated September 2, 2026, is an agency blog post rather than a market study. The agency describes itself as working "only with beauty, skincare and lifestyle DTC brands." Its figures are blended across the beauty and skincare brands it manages, sourced from Triple Whale, and cover July 1 to August 31 in 2025 and 2026. The post does not say how many brands are included, and "no individual brand figures are disclosed." It ends with an offer of a 20-minute review of a brand's numbers.

The headline numbers move in the opposite direction from AdRoll's. "Meta CPMs across our book are up 35% year over year," the post states. The table puts Meta CPM at $17.94 against $13.28 a year earlier, a 35.1% increase. New customer CPA rose 77.2%, from $17.79 to $31.52. Blended ROAS fell 28.6%, from 7.06 to 5.04, and ad spend as a share of sales rose from 14.2% to 19.9%.

Pennock argues the creative was not the problem. Meta click-through rate improved from 1.15% to 1.49%, and the post concludes: "The cost increase is auction pressure, not a creative failure." Google ROAS in the same book rose 17.2%, from 7.11 to 8.33.

Where the orders went

Site orders fell 24.1%, from 60,765 to 46,107, while average order value rose 7.9% to $71.28. By PPC Land's calculation, multiplying orders by average order value implies site revenue roughly 18% lower than in 2025. Yet blended sales across the brands' own sites and Amazon rose 20.7%, from $5.27 million to $6.35 million.

"Demand did not disappear. It moved," the post says. "What changed is where the order lands and what it costs to buy." The growth on Amazon came with its own bill: according to Pennock, Amazon total advertising cost of sale went from 2.25% to 8.94%, which the post describes as growth that "was purchased, not free."

Owned channels shrank as well. Email went from 34.5% of sales to 22.1%, and the post notes that the owned channel "is carrying a third less of the load heading into the most email dependent quarter of the year."

What peak week looked like in 2025

Pennock's second table compares the eight days from November 24 to December 1, 2025 with October 2025. Meta CPM rose 76.0%, from $14.59 to $25.68, and Meta CPC rose 87.7%. Even so, Meta CPA fell 23.8%, from $25.32 to $19.30, and new customer CPA fell 20.5%. Revenue per day was 4.2 times the October level, and Meta ROAS rose from 3.19 to 4.68. The post attributes the gap to conversion rate, though conversion rate is not a row in the table.

Discounts accounted for 57.5% of sales during that peak week, against 25.6% in October. "Black Friday still pays for itself, but only for accounts that were built before October ended," Lindgren writes. A brand that only starts in the third week of November, according to the post, "pays the 76% and captures none of the 24%."

The calendar Pennock works back from

Black Friday falls on Friday, November 27, 2026, and Cyber Monday on November 30. Counting back, Pennock names September 18 as the last date that gives an account a full 10-week ramp and October 2 as "the 8 week floor." After that, according to the post, brands pay peak CPMs while their tracking pixel is still gathering the conversion data it needs. Its checklist covers costing offers against contribution margin, clean pixel and conversion data, a completed creative test cycle, warm audiences, email and SMS, and inventory. On tracking, it says: "Everything downstream depends on this, and it is the one item that genuinely cannot be fixed in November."

Why the two data sets diverge

The AdRoll and Pennock figures are not measuring the same thing. AdRoll reports display CPMs across a large base of online businesses, split by campaign type. Pennock reports Meta CPMs for an undisclosed number of beauty and skincare brands. The windows overlap but differ, ending on September 8 and August 31 respectively.

Other data points sit closer to Pennock's direction. Meta itself reported that its average price per ad rose 12% year over year in the second quarter of 2026, while ad impressions grew 14% and advertising revenue reached $59.36 billion. On the open web, DataBeat's network data showed US programmatic display CPMs up 26.9% year over year in June 2026, after a 24.2% rise in April. Those figures come from a different network, an earlier month and a supply base weighted toward publishers, so they do not contradict AdRoll's numbers directly. They do show that a 45% annual fall in display prices was not visible in every data set covering the first half of 2026.

AdRoll does not say whether the decline reflects fewer advertisers bidding, a change in its own customer mix, more available inventory, or shifts in how its platform buys. Pennock does not say whether its 35% Meta increase holds outside beauty and skincare. Neither source tests whether cheaper impressions produced more sales, and none of the figures is a measure of incrementality.

Why this matters for marketers

For advertisers planning the final weeks before Black Friday, the AdRoll data describes a price environment for display that is markedly cheaper than in 2025, at least for the July to early September window it covers. The report's own guidance is that this window closes as CPMs rise through early Q4. The retargeting reversal is the more unusual shift. Through most of the past year, PPC Land's coverage of AdRoll's data tracked retargeting costs climbing while prospecting costs fell, a pattern the company described in March as structural. A 29.1% annual fall in retargeting CPMs, and a 40.2% drop from the prior quarter, breaks that pattern, and AdRoll ties at least part of it to audience availability rather than demand.

Pennock's numbers show what that cheaper display picture leaves out. For brands that depend on Meta, the agency's data points to higher auction costs, a heavier load on paid media as email's share shrinks, and orders shifting to Amazon at a higher advertising cost. Both sources agree on one point: prices rise into peak week. Pennock's 2025 data put that rise at 76% for Meta CPMs. AdRoll's report says display CPMs typically climb through early Q4 and peak around Black Friday and Cyber Monday, without stating by how much.

What neither document establishes is how the two pictures fit together for a single advertiser buying both. Is cheaper display inventory a substitute for pricier Meta reach, or a different audience altogether? The sources leave that open.

Timeline

Summary

Who: AdRoll, the multichannel advertising platform, and its chief executive Vibhor Kapoor; Pennock, an agency working with beauty, skincare and lifestyle DTC brands, and its founder Nikki Lindgren.

What: AdRoll's Q3 2026 State of Digital Advertising Report found display prospecting CPMs down 45.0% and display retargeting CPMs down 29.1% year over year, while ABM CPMs rose 4.4%. Pennock's blended client data showed Meta CPMs up 35% and new customer CPA up 77% over July and August.

When: AdRoll's data covers July 1 to September 8, 2026, and the company announced the report on September 22, 2026. Pennock's post is dated September 2, 2026, and covers July and August 2026 against the same months of 2025.

Where: AdRoll's figures come from its platform data on more than 20,000 online businesses, paired with US economic indicators; the CPM data's geography is not stated. Pennock's figures cover its own book of brands, measured through Triple Whale.

Why: Both sources are aimed at holiday planning ahead of Black Friday on November 27, 2026. AdRoll's data shows display reach costing far less than in 2025 before the usual seasonal rise, while Pennock's shows Meta auction costs rising and orders moving to Amazon. The sources measure different channels and samples, and neither explains the gap between them.