Google Ads Benchmarks for DTC Ecommerce

The headline numbers

Across 28 direct-to-consumer brands and $4,361,045 of managed Google Ads spend in the twelve months to August 2026, the median account bought impressions at $16.76 CPM, clicks at $1.11, clicked at 1.79% and returned 4.28× on ad spend.

Every figure is measured the way the rest of this series is measured: the rate is calculated per account first, then equal-weight percentiles are taken across accounts. No single large spender sets the benchmark. That matters more on Google than anywhere else in this cluster, because the top account in this panel spends 50× the bottom one.

These are managed accounts, and that is the most important caveat on the page. Read this as what an actively managed DTC Google Ads account looks like, not as an industry average.

The full distribution

Delivery metrics — all 28 accounts. These need no conversion tracking to be comparable, so every account in the panel contributes.

Purchase metrics — the 25 accounts whose tracking is purchase-based. Three accounts are excluded here and only here: in each one, the largest conversion category by value is not a purchase, so a purchase-only ROAS would describe something the account is not trying to do. They remain in the delivery table above.

On CPM, CPC and cost per purchase, lower is better — so the 75th and 90th percentiles there represent worse performance, not better. On CTR, ROAS and click-to-purchase rate, higher is better. Mixing those up is the most common way a benchmark table gets misread.

The $151.57 CPM at the top of the range is a single low-volume search account buying a few thousand expensive impressions a month. It is real and it is disclosed rather than trimmed, but the median and the interquartile range are the numbers to use.

The same panel, split by campaign type

A blended account figure is the weighted average of three very different auctions. Split this panel by campaign type and the spread is wider than anything else on this page — and it changes what the headline numbers mean.

Performance Max is 70% of the money. That single fact is why the blended medians at the top of this page track the Performance Max column almost exactly. A blended Google Ads benchmark — ours or anyone else’s — is in practice a Performance Max benchmark with a search garnish.

Delivery, by campaign type. Cells with at least $1,000 of spend in the window; n is the number of accounts contributing to each column.

Search CPM is sixteen times Shopping’s. That is not a defect and it is not a problem to fix: search impressions are scarce and bought on intent, while Shopping and Performance Max buy broad feed and display inventory by the million. What turns an expensive search impression into a cheap search click is the click rate — 14.64% against Shopping’s 1.01%.

This is the single most useful thing on the page for reading anyone else’s benchmark: a CPM is meaningless without knowing which inventory it bought, and the same is true of a CTR. Only CPC survives the comparison, and only roughly.

Purchase metrics, by campaign type. Restricted to accounts in the purchase panel, with at least 25 purchases and $1,000 of spend in the cell.

Search wins every efficiency measure and buys the least volume. Performance Max is the most expensive way to buy a purchase in this panel ($31.04 against Search’s $24.04) and carries the highest implied order value — consistent with it being credited for a broader mix of demand, including customers who would have arrived anyway.

Demand Gen, Video and Display are not published here. Demand Gen reaches only 7 accounts above the spend floor and Video and Display reach one each — too thin for a median. Their combined share of panel spend is about 1%.

How this compares to the published benchmarks

A 28-account panel is small. The useful thing to do with it is not to publish it alone, but to set it beside the large third-party panels and say plainly where it agrees and where it does not. Every source below discloses a panel size, a window and a segment; sources that disclose none of those were excluded, however confident they looked.

Four metrics compare directly against Triple Whale’s panel, which is both the largest and the closest in window. Two favour this panel and two do not — and the pattern is more informative than any single number.

A higher ROAS alongside a worse cost per purchase looks contradictory until you multiply them. ROAS × cost per purchase = average order value, because revenue ÷ spend × spend ÷ purchases leaves revenue ÷ purchases.

• This panel: 4.28 × $32.47 = $139 implied average order value
• Triple Whale panel: 3.27 × $28.14 = $92 implied average order value, against the $87.65 they publish directly — the identity checks out

The brands here sell at roughly 1.5× the order value of the broad ecommerce advertiser base. That explains the worse cost per purchase — a $139 order can carry a $32 acquisition cost that a $92 order carries less comfortably — and it inflates the ROAS comparison too, because ROAS rises with order value independently of how well the media is bought.

So the honest read is narrow. Strip the order-value effect out and what is left is the pair of metrics that do not depend on it: CTR 8% higher and CPM 9% higher. This panel earns better click rates while paying more per impression. That is a real finding, and it is a smaller claim than the ROAS row suggests.

It is also the opposite shape to what the same brands show on Meta, where our Meta Ads benchmark sits 10% below the third-party CPM. Cheaper impressions on feed, more expensive impressions on search, better click rates on both.

Brand versus non-brand search

Splitting Search by intent turns out to matter more than splitting by campaign type. We pulled every keyword in every Search campaign in the panel — 4,700 keywords covering 95.6% of Search spend — and classified each one as brand or non-brand against that account’s own brand names.

The median account spends more on its own name than on everything else combined. The range is enormous — from 0.4% to 99.8% of Search budget — so this is not a panel-wide habit so much as two different strategies sitting in one dataset. The interquartile range, 37.2% to 89.1%, is the honest band.

That single split explains most of what looked strange about our Search numbers. A brand click costs $1.02 and converts at a 27.26% click rate, because the person typed the brand’s name and the brand’s own ad is the most relevant result on the page. A non-brand click costs $2.14 and clicks at 8.15%, because it is bought in an open auction against everyone else selling the category.

Why this matters more than any other number on the page: a blended Search CPC is a weighted average of two prices that differ by more than 2×, mixed in a ratio that varies from 0.4% to 99.8% between accounts. Two accounts with identical media skill will report Search CPCs that differ several-fold purely on how much of their budget is defensive brand bidding. A blended Search CPC is not a comparable metric. Non-brand CPC is.

The same logic applies to your own reporting. If your Search CPC improved this quarter, check whether your non-brand CPC improved or whether brand simply took a larger share of the budget. They look identical on a blended chart and they mean opposite things.

Why our CPC is a quarter of the published figure — two hypotheses, one survived

This section was supposed to write itself. We report a blended CPC of $1.11. WordStream’s study of 13,474 US campaigns reports $4.44 for apparel and fashion and $4.14 for shopping, collectibles and gifts. WordStream measures Search campaigns only; this panel measures whole accounts including Shopping and Performance Max. Obvious explanation.

Then we segmented the panel, and it did not survive.

Search-only CPC in this panel is $1.31. That is barely above the blended $1.11, and still roughly three times below the published search figure. Campaign mix moves our number by about 18%. The gap to WordStream is about 300%. Mix is not the explanation.

The CTR row pointed at what might be, so we tested it. Search click-through rate in this panel is 14.64% against WordStream’s 6.64–8.28% for the same industries — and a click rate that high on search is the signature of brand-term bidding. If that were the explanation, then stripping brand terms out should pull our CTR down into their band.

It does, almost exactly. Non-brand Search CTR in this panel is 8.15%, sitting inside WordStream’s published 6.64–8.28% range. That was a prediction made before the split was run, and it held. The CTR anomaly is brand bidding, and nothing else needs to explain it.

On CPC the same test only gets us halfway. Non-brand CPC is $2.14 against a blended Search $1.31 — so brand mix accounts for a 1.6× step. But $2.14 is still about half of WordStream’s $4.14–$4.62. Brand share closes roughly half the gap and something else closes the rest.

What that something else is, we cannot settle from this panel, and we are not going to pretend otherwise. The two most plausible candidates:

• Category composition. WordStream’s buckets — “Apparel/Fashion & Jewelry”, “Shopping, Collectibles & Gifts” — are broader than DTC ecommerce and include higher-CPC businesses. Their all-industry median is $5.42; several of their verticals run far above our categories.
• Quality Score and account management. Every account here is actively managed, and relevance advantages compound into lower click prices. Plausible, unmeasured, and exactly the claim an agency would want to be true — which is reason to hold it loosely.

So the honest final position: campaign mix explained almost nothing (18%), brand share explained the CTR gap completely and about half the CPC gap, and the remaining ~2× on CPC is unresolved. We would rather publish it that way than round it into a story about how well we buy media.

The practical advice, which survived all three versions of this section:

• Compare non-brand CPC, never blended. Blended Search CPC mixes two prices that differ by more than 2×, in a ratio that runs from 0.4% to 99.8% across this panel alone.
• If your blended search CPC is near $1 and your non-brand is near $4, you have not found an efficiency. You have found out how much of your budget is buying your own name.
• Check whether a CPC improvement was real. A rising brand share lowers blended CPC while non-brand performance is flat or worse. The two are indistinguishable on a blended chart.
• Compare CTR and click-to-purchase rate across accounts too — but split those by brand as well, for the same reason. Our blended 14.64% and non-brand 8.15% tell completely different stories.

Google versus Meta in the same brands

Because this cluster measures both platforms across an overlapping set of clients, the two benchmarks can be read side by side — which is rare, and more useful than comparing either against an industry average.

The direction is consistent and it is what you would expect: search harvests demand, feed creates it. Google clicks cost about twice as much and convert about two and a half times as often, which nets out to a materially lower cost per purchase and a higher ROAS.

The CPM row deserves the campaign-type caveat, though. Meta’s $13.52 sits above Google Shopping’s $10.56, just below Performance Max’s $16.00, and nowhere near Search’s $168.73. Comparing a feed CPM to a blended Google CPM is comparing it mostly to Performance Max — which is a reasonable comparison, but worth knowing you are making it.

That is not an argument for moving budget. A channel that only harvests existing demand runs out of it, and the ROAS gap partly reflects Google taking credit for purchases that earlier Meta impressions created. Our returning vs new customer benchmark shows why any channel weighted toward people who already know the brand flatters itself: returning visitors convert 2.60× better regardless of what brought them back.

The tracking problem nobody discloses

Most published Google Ads benchmarks report “conversions” as Google reports them. That number counts every conversion action an account has marked as primary — which in real accounts includes add-to-carts, begin-checkouts, phone calls, form fills, store visits and get-directions clicks.

In this panel, using the unfiltered figure would have been badly misleading. One account records 507,795 conversions in the window, of which 995 are purchases — the rest are store visits and get-directions clicks. Its unfiltered ROAS reads 7.2; its purchase ROAS is 0.03. It is a retail chain, not a DTC store, and it is excluded from this panel entirely.

So every purchase metric here is restricted to conversion actions in Google’s PURCHASE category. Three further accounts sit in the delivery table but not the purchase table, because in each the largest conversion category by value is something other than a purchase. Whenever you compare your own ROAS to a published benchmark, check which definition of conversion the benchmark used. Most do not say.

What to do with this

• Diagnose in order: CPM, then CTR, then click-to-purchase rate. CPC is a symptom. It is the product of the first two, and the fix is different depending on which is driving it.
• Segment before you benchmark, then segment again. Pull Search, Shopping and Performance Max separately — then split Search into brand and non-brand. The first split tells you which auction you are in; the second matters more, because brand and non-brand clicks differ by more than 2× in price and by 3× in click rate.
• Report non-brand CPC as your headline search number. Blended Search CPC moves with brand share, which is a budget decision rather than a performance one. Compare $2.14 non-brand, not $1.31 blended.
• Check what share of your spend is Performance Max. If it is near this panel’s 70%, your blended account figures are effectively Performance Max figures and should be read against that column, not the blended one.
• Compare CPM against the $13.71–$27.11 band, not against a single number. Above it usually means a narrow keyword set or thin Performance Max asset coverage.
• Judge ROAS against your own margin first, and only then against the 3.32–5.42 range here.
• Restrict your conversions to purchases before comparing anything. If your account counts add-to-carts as conversions, your ROAS is not the same metric as the one on this page.
• Check your implied order value. ROAS × cost per purchase tells you it in one step. If it is far from this panel’s $121 median, the cost-per-purchase row here does not apply to you.

Methodology and limits

Figures come from the Google Ads API at account level, queried per client account for the trailing twelve months from 1 September 2025 to 31 August 2026. CPM, CPC and CTR are derived from cost, impressions and clicks. Purchase ROAS, cost per purchase and click-to-purchase rate use only conversion actions in Google’s PURCHASE category, segmented explicitly rather than taken from the account “Conversions” total. Each account’s rates are computed first, then equal-weight percentiles are taken across accounts.

Panel: we examined all 81 enabled client accounts under our Google Ads manager account. 28 met the rules, all of which were set before any result was seen: a direct-to-consumer ecommerce brand, billing in USD, with at least $10,000 of spend in the window. The 53 exclusions break down as 23 below the spend floor, 14 not DTC ecommerce by account type (software, professional services, publishers, our own agency account), 11 with no spend in the window, 4 not billed in USD, and 1 optimised for store visits rather than online purchases. Two further accounts were visible but could not be queried. No account was added or removed on the basis of its results.

Known limits, stated plainly:

• Managed accounts, not an industry average. Every account here is actively managed by one agency. That is the single biggest reason these figures may sit apart from an unmanaged baseline.
• n = 28 is a small panel. Real first-party data, but not a census. Percentile estimates at the tails are indicative only.
• Campaign mix is reported, not normalised. The headline figures are blended, because that is what an account owner sees; the by-type section splits the panel so the mix effect is visible rather than buried. Accounts run different blends, and Performance Max is 70% of panel spend.
• Brand and non-brand are split from keyword-level data covering 95.6% of Search spend. The residual 4.4% is Dynamic Search Ads and other keywordless Search traffic, which carries no keyword to classify and is excluded from the brand split (but not from the Search totals).
• Brand classification uses company brand names only. A keyword counts as brand if it contains the account’s brand name or a compressed form of it. Own-product-line names that do not carry the brand — a flagship product name, a founder’s name — stay in non-brand. Those terms behave like brand terms and cost like them, so the true non-brand CPC is somewhat higher than the $2.14 reported here. The bias runs against our own argument, which is the direction we prefer it to run.
• The residual CPC gap is unexplained. Brand share closes about half the distance to the published search benchmarks; category composition and Quality Score are plausible candidates for the rest and neither is measured here.
• By-type cells carry their own floors. Delivery figures use cells with ≥$1,000 of spend; purchase figures additionally require ≥25 purchases in the cell and an account already in the purchase panel. Demand Gen, Video and Display fall below a publishable count and are disclosed rather than reported.
• Google’s attribution. ROAS and purchases use Google’s own data-driven attribution, which credits search more generously than a last-click or post-purchase-survey view would.
• USD only. Accounts billing in AUD, EUR and HKD were excluded because cost metrics are not comparable across currencies without an exchange-rate assumption we would rather not bake in.
• Google only. Meta and TikTok are measured separately in this series and cross-channel conclusions should not be drawn from one panel.
• Category mix. The panel skews to home, pet, apparel, craft, consumer electronics and CPG. Auction dynamics differ sharply by category.
• Two outliers disclosed, not trimmed: a $151.57 CPM on a very low-volume search account, and a $439.17 cost per purchase on a high-order-value brand. Medians are used throughout.

How to benchmark your own account

• In Google Ads, set the date range to the last full 12 months and read cost, impressions, clicks and conversions at account level.
• Before anything else, open Goals → Conversions and check which actions are marked primary. Filter to purchases only, or your ROAS is not comparable to this page.
• Segment by campaign type and compare like with like: Search CPC against $1.31, Shopping against $1.02, Performance Max against $1.12, blended against $1.11.
• Then add a brand / non-brand split on Search — this is the one that changes the answer. Compare your brand CPC to $1.02 and your non-brand CPC to $2.14. If your Search CTR is well above the 8.15% non-brand median, most of that spend is buying your own name.
• Check your brand share of Search spend against this panel’s 59.8% median (IQR 37.2–89.1%). A much higher share is not automatically wrong — defending the brand term is often cheap and sensible — but it means your blended figures describe your defence, not your acquisition.
• Place CPM first. Above $27.11 usually points at a narrow keyword set or weak Performance Max assets before creative is the problem.
• Then CTR against 1.79% whole-account, 14.64% Search-only, 1.48% Performance Max or 1.01% Shopping — the four numbers are not interchangeable.
• Then click-to-purchase rate against 3.13% — that isolates the landing experience from the media.
• Finally, divide revenue by purchases. If your order value is far from $121, ignore the cost-per-purchase row entirely and use click-to-purchase rate instead.

Sources and methodology

Frequently asked questions

• Google Ads API — Reporting
• Triple Whale — Google Ads Benchmarks by Industry (21,000+ brands, Aug 2025 – Jul 2026)
• WordStream — Google Ads Benchmarks (13,474 US search campaigns, Apr 2025 – Mar 2026)
• Two Minute Reports — Google Ads Benchmarks by Industry (Jan – Dec 2025)
• Meta Ads Benchmarks
• TikTok Ads Benchmarks
• Average Order Value Benchmarks
• Ecommerce Conversion Rate Benchmarks
• TGM CPC Benchmarks & Calculator

See also our average order value benchmark for the number that makes cost per purchase comparable, and our Meta Ads benchmark for the same brands on feed.