Skip to content
2026 report · For GURU Conference attendees

Email brings the visitors you can most often name.

What's working, what's changing and what small growth teams can prove. Email, paid media, AI assistants and who is really visiting your site. Complimentary for GURU Conference attendees.

November 202615-minute read

Questions about the data? Talk to us

Executive summary

Eight things SMB marketers should know going into 2027

SMB teams have more channels than ever and less certainty about which ones deliver returns. We paired iCustomer's own website visitor data with the most current public benchmarks we could verify, plus published findings from Outcome Media. The headline: email brings the visitors you can most often name.

Email first

Visitor identity. Email brings the visitors you can most often name: 14.0% of new visitors, ahead of AI answers (13.6%) and organic search (13.2%). The identity section has every source.

iCustomer Signal, 2H 2026
33.0%

Source mix. of new visitors arrive with no source at all, the largest single bucket. On consumer sites, more than half of them land on a product page.

iCustomer Signal, 2H 2026
Personal AI agents

AI agents. such as Meta's Muse and Instinct visit your site on a person's behalf. We treat each visit as that person's, so they're included in every source and identification figure in this report.

iCustomer Website Visitor Taxonomy, category 3
30%

Email. of email revenue came from automated emails, just 2% of sends. Outcome Media finds 5+ sends a month lift opens 17% to 20%. Pick winners on clicks, not opens.

Omnisend 2026; Jay Schwedelson, via CMI, Apr 2026
-5%

Paid search. Cost per lead on US search fell 5% in a year while cost per click rose 3%. Conversion rate improved from 7.52% to 8.18%.

WordStream/LocaliQ, 2026 edition
100,000+

ChatGPT Ads. advertisers ran ChatGPT Ads in the first eight months, now in 89 countries. There are no public performance benchmarks yet, so cap the budget and judge it on confirmed leads.

OpenAI, Oct 21, 2026
+90%

AI search and AEO. How much better AI-referred visits converted than other traffic on US retail sites in September 2026. Volume is still small.

Adobe Digital Insights, Sep 2026
66%

Measurement. of B2B advertisers testing their ads didn't get enough form fills to tell if the ads worked. Smaller programs can test on website visits instead.

Primer, 2026

One card per topic. The first three are iCustomer's own data. The rest are public benchmarks and partner findings.

01 · The channel problem

More channels, less certainty

Marketers aren't short on tools or tactics. They're short on a reliable answer to one question: what is working?

68.9%

of brands actively use five or more marketing channels.

HubSpot 2026
33%

of marketing leaders say proving ROI is their top challenge, more than any other.

HubSpot 2026
71%

struggle to keep up with how buyers move across platforms.

HubSpot 2026
25%

strongly agree they have the data they need to reach their audience.

HubSpot 2026

AI is everywhere in marketing. Mostly in content.

HubSpot finds 86.4% of marketing teams use AI in at least some areas, up from 41% in 2024. But the use is lopsided. Supermetrics found marketers lean on AI for writing more than for the reporting and analysis that tells them where to spend.

Content and copywriting50%
Reporting and analytics35%

Share of marketers using AI for each task. Source: Supermetrics, 2026 Marketing Data Report.

What it means for SMBs: HubSpot's survey covers companies of all sizes. Among B2B software teams, mostly $1M to $10M in revenue, the average is 5.5 core channels (Growth Unhinged, 2026). Consumer brands are no simpler: on iCustomer consumer sites, new visitors arrived from nine kinds of source in 2H 2026. For a small team, that many channels produce more data than anyone can read. The bottleneck isn't more content. It's knowing which visitors and which channels are worth the next dollar.

B2B takeaway

Pick one number sales trusts, like pipeline by source. Report every channel against it before you add another one.

B2C takeaway

Pick one number finance trusts, like orders or revenue by source. Report every channel against it before you add another one.

02 · Email

Email: count clicks, automate first

Owned audiences, no auction prices. But the way to measure email has changed.

Email click rate by industry

All industries 2.09%

Ecommerce1.07%
Software and web apps1.15%
Retail1.27%
Health and fitness1.45%
Business and finance2.37%
Consulting2.41%
Nonprofit2.90%

Source: MailerLite, 3.6M campaigns, 2026 benchmark report.

Why clicks, not opens

Apple Mail Privacy Protection loads images for the reader, so many "opens" are machine events. MailerLite's 2025 open rate was 43.46%, but it calls click rate the most accurate engagement signal today. Click-to-open rate was 6.81%. Ecommerce and software sit near half the all-industry click rate.

Triggered beats scheduled

30%

of email revenue came from automated emails, just 2% of sends.

16x
More revenue per automated email

than per campaign email: $2.87 vs $0.18.

$8.46

per back-in-stock email, the highest-earning automation.

Source: Omnisend 2026.

Three tests that moved open rates

+20%

opens from 5+ sends a month for B2B. +17% for B2C.

Jay Schwedelson, via CMI, Apr 2026
+14%

opens from all-lowercase subject lines, over 90 days.

Worldata Research, Jan 2026
+28%

opens when Apple Mail marks it priority. Confirmation-style subject lines help.

Jay Schwedelson, via HubSpot

These lifts were measured on opens, which privacy features inflate. Test each one and pick the winner on clicks.

iCustomer data. Email in our visitor data. Email brought 5.6% of first-time visitors in 2H 2026, and 6.5% on consumer sites. Those visitors were named as a person 14.0% of the time, the highest of any source. Email mostly reaches people who already know you, so a first-time-visitor share understates it.

B2B takeaway

Count clicks to high-intent pages (pricing, demo) per 1,000 delivered. Retire open rate as a goal.

B2C takeaway

Build flows before campaigns: welcome, abandoned cart, back-in-stock. They earn far more per send.

02 · Email and direct traffic

A third of new visitors arrive with no source

On consumer sites, more than half of them land on a product page. Untagged email links are a likely cause, and tagging them takes five minutes a campaign.

33%

of all new visitors arrive with no source, the largest single bucket.

47%

of direct visits to consumer sites are on a phone. On B2B sites it is 15%.

Where direct visits land
First page seenConsumer sitesB2B sites
Home page30.9%46.7%
Product page52.5%Under 1%
Article or content page2.9%19.4%
Other inside page13.7%33.8%

Share of direct visits by the first page seen. iCustomer Signal, 2H 2026, first-time visitors, pooled across sites. The full source mix is in the sources section.

What the data shows

  • On consumer sites, 52.5% of direct visits land on a product page, a URL people rarely type.
  • Many arrive on a phone, which fits links tapped in an inbox or a messaging app.
  • B2B looks different: 85% of direct visits are on desktop, which fits bookmarks and work chat.

What it doesn't show

  • How much of it is email. Untagged links, forwarded messages, in-app browsers and some mail apps all drop the source.
  • Email is one likely contributor. Messaging apps, AI tools and bots we didn't catch are others.

The fix is cheap either way.

UTM cheat sheet for email

ParameterWhat goes in itExample
utm_mediumAlways the word emailemail
utm_sourceThe sender, list or platformnewsletter, klaviyo, partner-name
utm_campaignThe send, dated2026-11-black-friday
utm_contentThe link or block clickedhero-button, footer-link

Lowercase only. No spaces. Tag every link, including images and the footer. Use the same names every time. Then send yourself a test and check that the click lands under Email.

Takeaway. Tag every link in your next four sends, then compare. If direct falls and email rises, you've found traffic that was yours all along.

See what is inside your own direct traffic. Try it free at icustomer.ai/ireveal.

03 · Paid media

Clicks cost more. Leads cost less.

On US search, cost per click rose 3% in a year and cost per lead fell 5%, because conversion rate improved from 7.52% to 8.18%.

Average cost per click

2016$2.32
2025$5.26
2026$5.42

Average cost per lead

2016$59.18
2025$70.11
2026$66.69

US search campaigns on Google Ads and Microsoft Ads combined. 2026 edition covers 13,474 campaigns, Apr 2025 to Mar 2026; average CTR 6.64%, conversion rate 8.18%. Against 2016, both costs are higher; the 2016 figures are Google Ads only. Source: WordStream/LocaliQ.

iCustomer data. Paid media in our data. Paid search and paid social together brought 38% of new visitors across iCustomer sites in 2H 2026: 42% on consumer sites and 27% on B2B sites. Google Ads sent 96% of paid search visitors and Meta 96% of paid social. Paid social visitors were among the hardest to name: 6.7% were identified as a person, against 13.2% for organic search.

Takeaway. Rising click prices are survivable when conversion rises faster. Judge search on cost per lead or order, then on what those leads become.

B2B budgets are small, and a new ad channel opened. Two more things to know beyond search prices.

B2B paid

$5.4K
Median monthly ad spend

at $1M to $10M companies. $14K at $10M to $50M, and $63K above $50M.

Primer 2026
59%

of $1M to $10M companies advertised on only one platform. Above $50M, 22% did.

Primer 2026
5.3x
LinkedIn's CPM against Meta's

for the same advertiser.

Primer 2026
+27%

rise in Google's platform-reported cost per conversion, against an 11% rise in cost per click.

Primer 2026

Primer, State of B2B Paid Advertising 2026: 379 B2B advertisers on Google, LinkedIn and Meta, year to August 2026. These are Primer customers, mostly B2B tech, and Primer sells B2B ad targeting and measurement. That mix is likely why Google's cost per conversion rises here while the all-industry average above falls.

A new paid channel: ChatGPT Ads

Feb 9, 2026

US ad testing begins

100,000+

advertisers had run campaigns by Oct 21

89

countries with ChatGPT Ads

1.2B

people using ChatGPT each week

Advertiser count is cumulative since launch. OpenAI has not published spend or performance benchmarks. Source: OpenAI, Oct 21, 2026.

If you test it. There are no public performance benchmarks yet. Cap the budget, tag every link and judge it on confirmed leads or orders, the same way as any other channel.

Takeaway. In Primer's data, most small B2B advertisers ran one platform. Before adding a second, check what the first one's leads become in your CRM.

04 · Where visitors come from

Direct and paid bring seven in ten new visitors

Where first-time visitors came from on iCustomer client sites in 2H 2026, largest source first.

Share of new visitors
SourceAll sitesB2BB2C
Direct and dark traffic (no source)33.0%39.5%30.7%
Paid social19.6%21.8%19.5%
Paid search and shopping18.5%5.4%22.8%
Organic search11.8%14.0%10.9%
Referral, affiliate and partners6.1%7.4%5.8%
Email and newsletters5.6%3.7%6.5%
Organic social and communities2.4%6.7%0.9%
Display, video and other paid1.9%0.7%1.8%
AI search and answer referrals0.8%0.3%1.0%
Unknown and unclassified0.2%0.5%0.1%
All new visitors100%100%100%

Pooled across sites, so larger sites weigh more. B2C is concentrated in a few large sites, so read it as directional. Personal AI agents are counted as the person they act for. AI crawlers are in the AI agents section.

The Website Visitor Taxonomy: twelve categories

#CategoryCounted as
1Organic searchPeople
2AI search and answer referralsPeople
3Personal AI agentsThe person they act for
4AI crawlers and discovery botsCounted apart
5Paid search and shoppingPeople
6Paid socialPeople
7Organic social and communitiesPeople
8Email and newslettersPeople
9Referral, affiliate and partnersPeople
10Display, video and other paidPeople
11Direct and dark trafficPeople
12Unknown and unclassifiedPeople

People are shown in the table above. Personal AI agents are counted as the person they act for. AI crawlers are counted apart, in the AI agents section.

What stands out

4x
Paid search

is about four times as large a share on consumer sites (22.8%) as on B2B sites (5.4%).

iCustomer Signal, 2H 2026
39.5%

of new B2B visitors arrive with no source, against 30.7% on consumer sites.

iCustomer Signal, 2H 2026
7x
Organic social

is 7x as large a share of new visitors on B2B sites (6.7%) as on consumer sites (0.9%).

iCustomer Signal, 2H 2026

Takeaway. Sort your own sources the same way, largest first. Then start with the biggest bucket you can't explain. On most sites that is direct.

06 · AI agents

The visitors you can't tell from people

Personal AI agents now research, compare and buy on the user's behalf. They show up on your site looking like people.

  • Sep 8, 2026Meta launches Muse, a personal agent in WhatsApp and a standalone app. Each agent runs its own browser and can check out with a single-use card.
  • Sep 20, 2026Amazon blocks Muse from shopping on Amazon.com. Shopify merchants stay open to it.
  • Oct 2, 2026Profound finds Muse filled 4 of 6 shopping picks from websites it opened live, not from Meta's catalog.
  • NowInstinct, a second personal agent, searches Shopify's catalog and browses live. Private beta.

Why your analytics will miss them

Profound reports that Muse and Instinct don't identify themselves by name when they browse, and don't name the sources they used in a response. Meta says Muse's browsing "will appear as your activity," meaning the user's. To your analytics, an agent researching your pricing page on someone's behalf looks like an ordinary visit from a person.

What SMBs can do now

  1. Make pages machine readable. Agents read structured content and the accessibility tree. Prices, stock and specs locked in images or scripts get skipped.
  2. Watch behavior, not labels. Visits that read many pages in seconds or fill a form instantly are better clues than the browser's name.
  3. Decide your policy. Amazon chose to block. Shopify chose to open checkout to agents. Either way, decide on purpose.

iCustomer data. Agents and crawlers in our data. When a personal agent shops for someone, we treat the visit as that person's, because it is. So agents sit inside the source mix and the identification rates in the identity section. AI crawlers are counted separately: 2.4% of first visits in 2H 2026, 5.8% on B2B sites and 0.1% on consumer sites. Of those, 23% were verified, 12% declared themselves and 65% were detected from behavior, such as headless browsers and scripts. Tag-based counts are a floor. Edge installs capture crawlers that never run the tag.

B2B takeaway

Agents research before a buyer books a demo. Keep pricing and integration details in plain text on the page.

B2C takeaway

Decide whether agents may check out, then test that one can read your price and stock.

07 · Measurement

What the platform says, what happened, what you caused

Ad platforms, your CRM and a lift test answer different questions. Put them side by side, but never treat them as the same thing.

LayerWhat it countsThe question it answers
Platform-reportedConversions an ad platform credits to itself under its own attribution rulesWhat does the platform say it drove?
ConfirmedDeduplicated outcomes in your CRM, store or billing systemWhat actually happened?
IncrementalExtra conversions in an exposed group versus a held-out control groupWhat happened because of the marketing?

Google's help docs draw the same line. Standard conversions follow your attribution settings. Conversion Lift counts only the difference between people who saw ads and a held-out group.

Two ratios to track

Reconciliation ratio = confirmed conversions matched to a channel / conversions that channel reports.

Incremental ROAS = (revenue in the exposed group minus revenue in the control group) / spend.

7x
LinkedIn

touched about 7 times more confirmed form fills than last-touch reports credit it with. Meta 1.6x, Google 1.1x.

Primer, 379 B2B advertisers
66%

of advertisers with lift data (92 of 140) had too few conversions to measure form-fill lift.

Primer 2026
92%
Agreement between website-visit lift and form-fill lift

where both were measured.

Primer 2026

Run the check yourself

  1. Pull last month's conversions as each ad platform reports them.
  2. Match them to CRM or store records by click ID, email or order ID.
  3. Divide confirmed by reported for each channel. Track that ratio every month.

Takeaway. Start with email: hold back a random 10% of a list for one campaign and compare orders or pipeline. For ads, small budgets rarely produce enough form fills to prove lift, so use website-visit lift as the early signal.

08 · Visitor identity

Who you can identify

How many visitors can be matched to a person or a company on iCustomer client sites, and what changes the answer.

45% to 75%

of consenting visitors were matched to a person or a company. Site setup decides where you land. Edge tracking gets sites to the high end.

About 79% of new visitors accepted the consent banner.

For context: on high-traffic consumer sites, forms and logins alone, without a visitor ID tool, identify 7% to 9% of new visitors. And 42% of B2B GTM teams already act on website visits as a buying signal (Growth Unhinged, 2026).

By source, for first-time visitors

All sources: about 1 in 10

Email14%
AI answer referrals13.6%
Organic search13.2%
Paid search12.4%
Direct and dark10.2%
Paid social6.7%
Referral and partners5.5%
Organic social2.5%

New visitors named as a person, 2H 2026. The dashed line is the rate across all sources. The 45% to 75% above also counts returning visitors and company matches. Method in the methodology section.

Takeaway. Start with the visitors your emails bring: they are the ones you can most often name. Send a named visitor who views pricing or a product page to sales, or into a flow, the same day.

Your scorecard

Benchmark your own program in 20 minutes

Fill in the right-hand columns from your own analytics, ESP and ad accounts. Compare like with like, and note the source of every number.

MetricBenchmarkSourceYour numberGap or next step
Email click rate2.09% overall; 1.07% ecommerceMailerLite
Email click-to-open rate6.81%MailerLite
Share of email revenue from automations30% (from 2% of sends)Omnisend, B2C
Paid search cost per click$5.42WordStream, US
Paid search cost per lead$66.69WordStream, US
Monthly paid spend, B2B $1M to $10M revenue$5,400 medianPrimer, B2B
Ad platforms in use2+ on 45% of sites; 1 for 59% of small B2BiCustomer; Primer
Direct share of traffic30% of sessions; 33.0% of new visitorsWisepops; iCustomer
Email share of traffic4% of sessions; 5.6% of new visitorsWisepops; iCustomer
AI referral share of traffic0.1% of sessions; 0.8% of new visitorsWisepops; iCustomer
Visitors you can identify45% to 75%, by site setupiCustomer. Needs a visitor ID tool
Untagged links in last month's sends0 is the goalBest practice
Platform vs confirmed conversion ratioNo public standardTrack monthly
Holdout tests run this year1 per quarterBest practice
What to do next

Five questions to benchmark your growth program

  1. Do we know which visitors matter? Track the share of visitors you can name, by source, and how many of them reach pricing or demo pages.
  2. Does email create confirmed outcomes? Report clicks and conversions per 1,000 delivered. Retire open rate as a goal.
  3. Do ad platform claims match our CRM? Compute a reconciliation ratio for each paid channel every month, and run one holdout test a quarter. Email is the easiest place to start.
  4. Are AI assistants sending us demand? Give AI referrals their own source group, and track which pages AI crawlers read.
  5. Do we know what is inside our direct traffic? Tag every link in every send, then watch whether direct falls and email rises.

About iCustomer

iCustomer is a marketing decision intelligence company. It solves three problems for growth teams: audience intelligence (who is visiting and whether they fit), decision activation (what to do next, across channels) and causal measurement (what worked). An MIT spin-out founded in 2024, iCustomer runs on top of the CRM and data a business already has. icustomer.ai

About Outcome Media

Outcome Media, a division of Worldata Group, runs more than 40,000 marketing campaigns a year for brands. Its founder and CEO, Jay Schwedelson, hosts the "Do This, Not That" marketing podcast and founded GURU Media Hub, home of the GURU Conference.

See who is really visiting your site. Add your website to iCustomer and see where your visitors came from, which of them you can name, and which AI crawlers read your pages. Free to start: 100 revealed visitors a month. Try it now at icustomer.ai/ireveal.

Methodology

How we built this report

Three kinds of evidence: public benchmarks we verified at the source, published findings from Outcome Media, and aggregated, anonymized data from iCustomer.

Public and partner findings

  • MailerLite: 3.6M email campaigns, 2025
  • Omnisend: 27B emails, 150,000 ecommerce brands, 2025
  • WordStream/LocaliQ: 13,474 US search campaigns on Google and Microsoft Ads, Apr 2025 to Mar 2026
  • HubSpot: 1,505 marketers surveyed for the 2026 State of Marketing
  • Supermetrics: 2026 Marketing Data Report
  • Similarweb: 2026 Generative AI Landscape, Jun 2025 to May 2026
  • Adobe Digital Insights: 1 trillion+ visits to US retail sites
  • Demandbase: B2B web properties, Jun 2025 to Jun 2026
  • Primer: 379 B2B advertisers, year to Aug 2026
  • Wisepops: about 704M sessions, Apr to Jun 2026
  • Growth Unhinged: 300 B2B software GTM leaders, Sep 2026
  • OpenAI, Meta, Google, Profound: product announcements and documentation
  • Outcome Media and Worldata Research: published findings on subject lines and send frequency, as reported by CMI and HubSpot

Proprietary data

  • iCustomer Signal: aggregated, de-identified first-party pixel data from B2B and B2C client websites. Source-mix and engagement figures are pooled across sites, so larger sites weigh more
  • Identification: 2H 2026. The match rate is the share of consenting visitors, new and returning, matched to a person or a company. It ranged from 45% to 75% with website infrastructure. By-source rates are narrower: new visitors only, named as a person, capped days included.
  • Read as indicative: Signal ran on 500+ websites in 2H 2026; a website is one domain running the tag. Source-mix tables use the sites with a complete October. One period, so no trend is implied. No customer is named and no single customer's figures are shown. B2C volume is concentrated in a small number of large sites
  • Rules: aggregates only, no personal data. Each visitor is counted once, by the source of their first page visit: click IDs first, then UTM tags, then referrer, then direct. Figures are pooled unless a median is stated. Ad platforms in use is the share of websites with paid visits from two or more ad platforms.

Read with care. These studies cover different populations, regions and methods. We report each number with its source and date and don't blend them into a single "SMB average." Public figures are context, not estimates of our own customers. Where a study is US-only or ecommerce-only, we say so. Most public studies here cover companies of all sizes; only Primer's $1M to $10M band is specific to small companies.

How Signal captures and identifies visitors

Identification. Signal uses first-party signals collected after consent. Details are available on request.

Bots. Googlebot and Bingbot are checked by reverse DNS, signed agents are verified and generic bots are dropped. This covers only bots that run the tag.

Request-level capture. Edge installs (Cloudflare Worker, Vercel, Netlify, nginx or a CDN configuration) capture every request, including visits that never run the tag. The high end of the range in the identity section depends on it.

Glossary and sources

Terms we use and where the numbers came from

Terms we use

  • Session. One visit to a website.
  • Named, matched. Named: linked to a known person. Matched: linked to a person or a company.
  • Daily cap. A limit a site sets on how many visitors it reveals each day.
  • Confirmed conversion. An outcome recorded in your CRM, store or billing system.
  • Platform-reported conversion. A conversion an ad platform credits to itself.
  • Incrementality. Extra conversions caused by marketing, measured against a held-out group.
  • Holdout test. Keeping a random group from seeing ads, then comparing results.
  • Incremental ROAS. Extra revenue caused by ads, divided by spend.
  • Reconciliation ratio. Confirmed conversions divided by what a platform reports.
  • Click-to-open rate. Clicks divided by opens.
  • Direct traffic. Visits with no referrer and no tags. Often mislabeled.
  • Personal AI agent. Software that browses and buys on a person's behalf.
  • New visitor. A new browser. The same person on a second device counts again, which lowers every rate.
  • Edge tracking. The tag is installed at the site's edge, such as a Cloudflare Worker, so every request is captured.
  • AEO. Answer engine optimization: making your pages easy for AI assistants to read, cite and recommend.

Sources and notes

Notes. Percent changes are as reported by each source. Adobe figures are US retail only and compare AI-referred visits with all non-AI traffic. WordStream averages blend Google and Microsoft Ads. Proprietary figures are aggregated and anonymized; no individual customer, subscriber or visitor data appears in this report.