47.66° N 117.43° W Summit · Profitable Growth
First-party data · 33 accounts · Q3 2026
Google Ads benchmarks 2026: what 33 live accounts show.
Edition: Q3 2026 · Published · Next refresh January 2027 · Download the CSV · How to cite
Across 33 Google Ads accounts I manage, the median cost per click in Q3 2026 (1 July to 30 September) was $8.53 (N = 33). Among the 23 lead-gen accounts with tracking I trust, the median cost per lead in Q3 2026 was $119 (N = 23). In the 11 accounts with a complete Q3 2026 pull, the search terms report showed a median 52% of account spend (N = 11). Of the search-term spend Google did show, a median 30.7% went to terms with 2 or more clicks and zero conversions in Q3 2026 (N = 25).
The wasted-spend number people still quote is ten years old. Disruptive Advertising had audited 2,167 accounts as of 1 March 2016 and put the median account’s wasted spend at 75.8% (Disruptive Advertising, 2016). That audit read waste from the search terms report too. Today Google leaves low-volume searches out of that report, so any waste figure covers only the spend it shows. I measured the visible part, on live accounts, for one fixed quarter. Every one of my figures was read through the Google Ads API, read-only, and every one carries its N.
- $8.53 Median cost per click, all accounts (N = 33)
- 52% Median share of spend shown in the search terms report (N = 11)
- 30.7% Median share of visible search-term spend on terms with 2+ clicks and no conversions (N = 25)
- 23 of 31 Search accounts losing more to budget than to Ad Rank (N = 31)
The N changes from figure to figure because not every account can answer every question. Performance Max accounts have no search terms or Quality Score, and accounts whose lead count I can’t trust stay out of cost-per-lead and waste figures. The method sets out who is in each number, and the per-account CSV has every row I can publish.
01 Cost per click by vertical · N = 33
Average CPC by industry,
from live accounts.
Healthcare clicks were the cheapest of the lead-gen groups, with the middle half of accounts paying $2.05 to $6.98 (N = 8). Home services had a $13.27 median (N = 11). The single dearest click in the sample was a roofing account at $34.14 (N = 1).
| Vertical group | N | Median CPC | Range | Middle 50% (Q1 to Q3) |
|---|---|---|---|---|
| Healthcare | 8 | Not shown | $1.31 to $7.88 | $2.05 to $6.98 |
| Home services | 11 | $13.27 | $7.73 to $34.14 | $11.93 to $19.48 |
| Ecommerce, purchase-tracked | 4 | $1.78 | Not shown | Not shown |
| Other lead gen: local, retail, legal and B2B | 10 | $10.47 | $1.17 to $20.12 | $8.66 to $14.80 |
| All accounts | 33 | $8.53 | $0.82 to $34.14 | $3.36 to $13.27 |
Three of the four stores put 92% or more of spend into Performance Max. The other lead-gen group has the widest spread in the table: a used-vehicle dealer at $1.17 sits in the same group as a law firm at $20.12. So use your own vertical row, never the 33-account median, when you judge your CPC. Quartiles are shown only where N is 4 or more, using the inclusive method. Where a cell says not shown, the figure would give back the number of an account I don’t show as its own row.
02 Cost per lead · clean tracking only · N = 23
What a lead cost,
where the count was honest.
The median cost per lead across 23 lead-gen accounts was $119 in Q3 2026 (N = 23, middle 50% $75 to $210). I left 6 of the 29 lead-gen accounts out: five because their conversion count was broken, padded or full of non-lead actions, one because it logged too few leads to price. Those six are still counted in the tracking findings in section 08.
| Vertical group | N | Median cost per lead | Range | Middle 50% (Q1 to Q3) |
|---|---|---|---|---|
| Healthcare | 6 | $40 | $6 to $119 | $12 to $67 |
| Home services | 9 | $163 | Not shown | $112 to $196 |
| Other lead gen: local, retail, legal and B2B | 8 | $194 | $24 to $765 | $114 to $287 |
| All clean lead gen | 23 | $119 | $6 to $765 | $75 to $210 |
A lead here is the primary conversion as the account counts it: a tracked call, a form or a Local Services lead. It isn’t a booked job, and the gap between the two is the subject of the lead quality library. If yours is climbing, split CPC from conversion rate first.
The home-services median of $163 comes from nine accounts (N = 9). I kept the HVAC account in with a flag, because 39% of its phone-call conversions were click-to-call taps rather than connected calls (N = 1). I left out the paving account, which converted 0.2% of its clicks, and a general contractor counting scroll and form-start events. Four of the nine include Local Services leads in their account total, and section 09 splits those out. Cost per lead by trade for the same quarter, with the July edition alongside, is in the home services benchmark.
The healthcare low end of $6 is an ENT account that counts every call from an ad as a conversion, which likely includes existing patients (N = 1). The other lead-gen group runs from $24 to $765 because it holds eight unrelated businesses, from a used-vehicle dealer to a law firm, and two of them logged fewer than 10 conversions in the quarter. Read that row as eight data points, not a typical value. The fuller healthcare breakdown is in the healthcare Google Ads benchmark.
03 What the search terms report hides · N = 11
Google showed me about
half the spend.
In the 11 accounts where my Q3 pull returned every search-term row, the terms Google showed covered a median 52% of account spend (N = 11). The range was 20% to 71%, and the middle 50% sat between 48% and 56%. The rest of the money bought clicks on searches the report never listed.
Google’s own help page gives the reason: search terms without enough query activity are left out to meet its data privacy standards (Google Ads Help, About the search terms report, read 3 October 2026). I only counted complete pulls here. The other accounts either hit the 1,000-row cap on my query or had no completeness check, so their visible share is a floor rather than a reading.
Every negative keyword you add is chosen from that visible half. The waste figure in the next section applies only to it.
04 Wasted spend in the search terms · N = 25
Nearly a third of visible spend
bought no conversions.
Of the search-term spend I could see, a median 30.7% went to terms with 2 or more clicks and zero conversions in Q3 2026 (N = 25). The range was 10.0% to 56.5%, and the middle 50% sat between 23.9% and 34.6%.
| Vertical group | N | Median zero-conversion share | Range | Middle 50% (Q1 to Q3) |
|---|---|---|---|---|
| Healthcare | 6 | 26.5% | 10.0% to 41.0% | 17.0% to 31.3% |
| Home services | 9 | 27.2% | Not shown | 23.9% to 34.0% |
| Ecommerce, Search campaigns only | 2 | Not shown | Not shown | N under 4 |
| Other lead gen: local, retail, legal and B2B | 8 | 31.9% | 23.9% to 56.5% | 28.0% to 34.7% |
| All clean accounts with term data | 25 | 30.7% | 10.0% to 56.5% | 23.9% to 34.6% |
Two caveats before you quote it. A quarter is short, so a term with 2 clicks and no conversion in July to September might convert in October. And waste is only as real as the tracking behind it. An account that counts calls and ignores forms will show its form-driving terms as waste. Ten of the 25 accounts in this number count calls only, so read their share as a ceiling. The six excluded accounts are out of it entirely. This is also a different measure from the 2016 audit: I counted terms with 2 or more clicks, inside one quarter, in the visible half of spend only.
The six patterns that kept showing up
- Competitor business names. Someone searching for a named rival, sometimes with a city attached. The most common pattern in the healthcare accounts and the showroom account.
- Manufacturer and product-brand names. A model or brand query from a shopper comparing one exact product, often one the advertiser doesn’t stock.
- Cost research. Questions in the shape of “how much does X cost per square foot”. Early research with no contact intent in the quarter.
- Out-of-area searches. Terms naming places outside the service area, which location targeting let through.
- “Free” and rate shopping. Free-service searches, plus rate and financing research.
- Navigational searches for another provider. Someone trying to reach a different clinic, hospital scheduling line or showroom, who clicked the ad on the way.
If you want a dollar figure for your own account, the wasted spend calculator takes your spend and your red flags, and the wasted ad spend library covers each leak in turn.
05 Budget versus Ad Rank · N = 31
Most accounts hit their
budget before their rank.
In 23 of 31 accounts running Search, the impression share lost to budget was larger than the share lost to Ad Rank in Q3 2026 (N = 31). The median account held 33.9% of eligible impressions, lost 37.9% to budget and lost 26.2% to rank.
Three accounts reported impression share at Google’s floor, which it shows as 10% when the real figure is below that, so their share is an upper bound. The eight accounts where rank loss was larger included the cancer treatment center, the HVAC account and one of the stores. If rank is your bigger loss, budget isn’t what’s holding you back, and the raise-bids math shows whether a higher bid pays.
Budget-capped accounts spent their mornings harder
Twenty lead-gen accounts reported spend by hour block (N = 20). The 12 that lost 35% or more of impression share to budget put a median 60% of their quarter’s spend into hours 7 through 11, account time (N = 12). The 8 that lost less put a median 43.5% there (N = 8).
| Accounts that | N | Median spend share, hours 7 to 11 | Range |
|---|---|---|---|
| Lost 35% or more of impression share to budget | 12 | 60% | 26% to 67% |
| Lost under 35% of impression share to budget | 8 | 43.5% | 29% to 48% |
Three accounts break the split. A showroom that only runs 8am to 6pm landed at 48%. A small account on manual CPC bidding lost 66% to budget and put only 26% of spend into the morning, with 39% in midnight to 6am instead. The other nine budget-capped accounts landed between 56% and 67%. My read is that a capped daily budget gets spent on the first searches of the day, which leaves less for the afternoon. Check your own hour report before you decide where your money goes.
06 Quality Score · N = 31
Over a third of keyword spend
had no score.
A median 38% of keyword spend ran on keywords with no Quality Score reported in Q3 2026 (N = 31). A median 4% ran on keywords scored 7 to 10 (N = 31), and 9 of the 31 accounts had no keyword spend at all in that band. The middle 50% of accounts had between 19.5% and 70% of keyword spend unscored.
At the low end, the median account had 5% of keyword spend on QS 1 to 3 (N = 31), but 4 of 31 had 49% or more there. The two Performance Max-only stores report no keyword Quality Score and aren’t in this N.
When more than a third of the spend has no score, Quality Score can’t tell you much about an account. I use it to find the worst keywords, not to grade the whole thing.
07 Device · N = 23 lead gen, N = 4 ecommerce
Desktop only won
in ecommerce.
Mobile converted at a higher rate than desktop in 21 of 23 lead-gen accounts with clean tracking (N = 23), though in 3 of those 21 the gap was 0.2 points or less. On all four purchase-tracked stores, desktop clicks cost more than mobile clicks and desktop converted at a higher rate, though one store had too few purchases to read (N = 4).
| Finding | Segment | N | Result |
|---|---|---|---|
| Mobile converted at a higher rate than desktop | Lead gen, clean tracking | 23 | 21 of 23 |
| Mobile clicks cost more than desktop clicks | Lead gen, both CPCs reported | 21 | 16 of 21 |
| Median mobile share of clicks | Lead gen, clean tracking | 23 | 83.8% |
| Desktop clicks cost more than mobile clicks | Ecommerce, purchase-tracked | 4 | All 4 |
| Desktop converted at a higher rate than mobile | Ecommerce, purchase-tracked | 4 | All 4 (one on too few purchases to read) |
The two lead-gen exceptions were a senior living community, where desktop converted at 10.1% against 5.2% on mobile (N = 1), and a fencing account where the two sat 0.2 points apart (N = 1). Part of the mobile lead is built into the tracking: 10 of these 23 accounts count phone calls only, and calls come from phones.
08 Conversion hygiene · N = 29 lead gen, N = 4 ecommerce
One in five counted
the wrong thing.
6 of 29 lead-gen accounts counted a non-lead action or a click proxy as a primary conversion in Q3 2026 (N = 29). Where an account bids on conversions, Google optimizes toward whatever sits in that column, so those actions shape who the ads go looking for. The three that counted a non-lead action are out of the cost-per-lead figures. The three that counted a click as a phone lead stay in with a flag, because the click sat alongside real calls and forms.
| Finding | N | Result |
|---|---|---|
| Counted a non-lead action or a click proxy as a primary conversion | 29 | 6 of 29 |
| of which: map-direction taps, store visits, scroll depth or form starts | 29 | 3 of 29 |
| of which: click-to-call taps, text-button clicks or phone-number clicks counted as phone leads | 29 | 3 of 29 |
| Optimized on phone calls only, with no form or website lead action as primary | 29 | 12 of 29 |
| All conversions divided by primary conversions, lead gen, median | 22 | 2.3x (range 1.0x to 504x) |
| Lead-gen accounts where all conversions ran above 10x primary | 22 | 5 of 22 |
| All conversions divided by primary conversions, purchase-tracked ecommerce | 4 | Median 373x |
| Purchase accounts counting cart, page-view or engagement actions as primary | 4 | 0 of 4 |
The clearest cases, each N = 1. In a self-storage account, 72% of primary conversions were map-direction taps and store visits. A general contractor logged more conversions than clicks, a 114% conversion rate, every one of them a 50% scroll event or someone starting to type in a form. In an HVAC account, 39% of phone leads were click-to-call taps. A paving account counted only calls from ads and converted 0.2% of its clicks. A roofing account had a form-submit action log 4 leads as secondary, where bidding never saw them.
The all-conversions column is no help either. In lead gen it ran a median 2.3x primary conversions (N = 22), and above 10x in 5 of 22. In purchase-tracked ecommerce it ran a median 373x (N = 4), because analytics engagement events pile into it. None of the four stores counted cart or page-view events as primary, a contrast with the ecommerce tracking accuracy benchmark, where 2 of 3 stores counted cart or checkout events in an earlier window. If any of this sounds like your account, the conversion tracking diagnostics walk through how to confirm it.
09 Local Services Ads versus Search · N = 4, 3 rows shown
LSA leads cost a fraction
of Search leads.
In the 4 home-services accounts running both, Local Services Ads cost $31.65 to $45.59 per lead against $137.68 to $562.28 for Search, in the same account and quarter (N = 4). Search leads cost 3.8x to 14.3x as much in every one of them, and three of the four still gave LSA 6% of spend or less. One of the four is an account I don’t show as a row, so the table has three rows and no median.
| Trade | N | LSA cost per lead | Search cost per lead | Search as a multiple of LSA | LSA share of spend |
|---|---|---|---|---|---|
| Painting | 1 | $31.65 | $137.68 | 4.4x | 6% |
| Pest control | 1 | $45.59 | $174.14 | 3.8x | 1% |
| Roofing | 1 | $39.27 | $562.28 | 14.3x | 1% |
Don’t read this as LSA beating Search like for like. An LSA lead is a call or message Google charges for, while a Search conversion is whatever the account tracks, so the two counts use different definitions. And the sample under the LSA column is thin: in the pest control and roofing accounts, LSA ran at 1% of spend, too small a sample to lean on. What it does show is that in all four accounts the cheaper lead source got the smaller budget, which is worth a look if you run both. The trade-by-trade version, with LSA and Search side by side, is in the home services benchmark.
10 Quotable findings
Eight lines you can
lift as written.
Each one stands on its own with its N and window. Link this page when you use one.
- The median Google Ads cost per click across 33 live accounts was $8.53 in Q3 2026 (N = 33, 1 July to 30 September 2026).
- The median Google Ads cost per lead across 23 lead-gen accounts with clean tracking was $119 in Q3 2026 (N = 23, 1 July to 30 September 2026).
- Only a median 52% of account spend appeared in the Google Ads search terms report in Q3 2026 (N = 11 accounts with a complete pull).
- A median 30.7% of visible search-term spend went to terms with 2 or more clicks and zero conversions in Q3 2026 (N = 25).
- In 23 of 31 Google Search accounts, impression share lost to budget was larger than impression share lost to Ad Rank in Q3 2026 (N = 31).
- A median 38% of keyword spend ran on keywords with no Quality Score reported in Q3 2026 (N = 31).
- Mobile converted at a higher rate than desktop in 21 of 23 lead-gen Google Ads accounts with clean tracking in Q3 2026 (N = 23).
- 6 of 29 lead-gen Google Ads accounts counted a non-lead action or a click proxy as a primary conversion in Q3 2026 (N = 29).
11 Method
Who is in each number,
and why.
Source and window
Every figure comes from read-only Google Ads API queries on accounts I manage. Nothing was written to any account. The window is fixed: to , in each account’s own time zone. Accounts that were closed or cancelled returned no data and aren’t in the sample.
Inclusion and exclusion
- Minimum spend. An account needed at least $300 of spend in the window to be included. All 33 cleared it.
- The 33. 8 healthcare, 11 home services, 4 purchase-tracked ecommerce, and 10 other lead-gen accounts: 6 local and retail, 2 law firms and 2 B2B. Law firms and B2B sit inside the other lead-gen group, because at N = 2 a group figure would give back one account’s number.
- Accounts without a row. Some accounts belong to clients I name elsewhere on this site. They count in every median and total here, but I don’t show them as their own row, because a row would publish a named client’s numbers. Ten of the 33 are held back that way, so the CSV has 23 rows. Where a group figure would let you work one of them back out from the rows I do show, the cell says not shown.
- Excluded accounts. 6 of the 29 lead-gen accounts are excluded from cost-per-lead and waste figures: one flagged as undertracking, one with too few leads in the quarter to price, one flagged as over-counting, one that behaves like undertracking on review, and two counting map-direction taps or store visits as primary conversions. All six are counted in the tracking findings.
- Flagged, not excluded. Three accounts count a click proxy (a click-to-call tap, a text-button click or a phone-number click) among their phone leads. The rule I held: a click counted as a phone lead next to real calls and forms stays in with a flag; a non-lead action counted as a lead takes the account out. Each flagged account is marked in the CSV.
- Search terms visibility. Only accounts where the pull returned every row (under the 1,000-row cap) count toward the 52% figure.
- Performance Max-only accounts. Two stores have no search terms, keyword Quality Score or Search impression share, so they drop out of those Ns.
Definitions
- Lead or conversion. The primary conversions column, meaning the actions the account counts toward bidding. For lead gen that’s tracked calls, forms and Local Services leads. For ecommerce it’s purchases.
- Cost per click and cost per lead. Account cost divided by clicks, and by primary conversions, for the full quarter.
- Zero-conversion term spend. Spend on search terms with 2 or more clicks and zero conversions in the window, as a share of all spend visible in the search terms report.
- Visible share. Spend on the rows the search terms report returned, divided by total account spend.
- Impression share. Search impression share and the shares lost to budget and to rank, spend-weighted across each account’s Search campaigns.
- Quality Score bands. Share of keyword spend on keywords scored 1 to 3, 4 to 6, 7 to 10, or with no score reported.
- Morning share. Share of the quarter’s spend in hour blocks 7 through 11, account time.
How the medians were taken
Every median is a median of per-account values, unweighted, so a large account counts the same as a small one. Quartiles use the inclusive method and appear only where N is 4 or more. Where N is under 4, no group figure is stated. Single-account figures are labelled N = 1 where they appear.
Spend, revenue, conversion values and budgets aren’t published, and no business is named. For ecommerce, no per-account return on ad spend, order value, order count or product mix is published. Descriptors are kept general on purpose, because a precise one can identify an account against the case studies elsewhere on this site.
12 How to cite this
Use it, and link it.
Suggested citation:
Crowe, C. (2026). Google Ads Benchmarks 2026: What 33 Live Accounts Show (Q3 2026). Edition: Q3 2026, published 3 October 2026. https://connercrowe.com/benchmarks/google-ads-account-benchmarks/
If you quote a figure, please link this page rather than a screenshot of it, so readers land on the N and the method. The per-account data is in the Q3 2026 CSV: anonymized rows only, one per account for 23 of the 33, with code, descriptor, N and window.
Data licensed CC BY 4.0, attribution to Conner Crowe with a link. License terms.
13 Questions
What people ask about these numbers.
What is the average cost per click for Google Ads in 2026?
What is a good cost per lead on Google Ads in 2026?
How much Google Ads spend is wasted?
How much of my Google Ads spend shows up in the search terms report?
Is my Google Ads campaign limited by budget or by Ad Rank?
Do mobile clicks convert better than desktop clicks on Google Ads?
More first-party benchmarks
- Healthcare Google Ads benchmarks The healthcare accounts in this sample, broken down further.
- Meta Ads CPM by placement The same quarter on the Meta side.
- Home services cost per lead by trade Trade-by-trade cost per lead and the tracking failures behind it.
- All benchmarks Every report, and the five rules each one is held to.
Want this run on your account?
I will tell you which
of your numbers are real.
Thirty minutes on the phone. I read your account against these numbers before the call: how much of your spend the search terms report shows, what your primary conversions count, and whether budget or rank is your ceiling. You leave knowing which figures to trust and what to fix first.