Traffic channels
Traffic channels group visitor sources by marketing activity type. Channels combine similar sources into logical buckets so analysis and budget decisions stay manageable.
Classification hierarchy
Web analytics uses three levels.
Source
The specific platform or site. Examples: google, facebook, newsletter, example.com.
Medium
The traffic type or delivery method. Examples: organic, cpc, email, referral, social.
Campaign
The marketing initiative. Examples: summer-sale-2025, product-launch, black-friday.
Full classification example
URL: example.com/?utm_source=facebook&utm_medium=paid&utm_campaign=q1-promotion
- Source: facebook (where)
- Medium: paid (how)
- Campaign: q1-promotion (why)
- Channel: Paid Social (auto-grouped)
Statable's channels
Nineteen of them, the same set Google Analytics 4 uses. Every visit gets exactly one, decided at ingest from the referring domain and your own UTM tags. The set is fixed: there are no rules to write and nothing to configure.
Rules are tested in the order below and the first match wins, so the order is part of the answer.
| Channel | What lands here |
|---|---|
| Cross-network | Campaign name contains cross-network. Ads spread across several Google networks |
| Display | Banner and programmatic ads: medium display, banner, interstitial, cpm, or a dclid click id |
| Paid Shopping | A paid medium together with a shopping source or campaign |
| Paid Search | A search engine with a paid medium, or a gclid, msclkid, gbraid, wbraid or yclid click id |
| Paid Social | A social network with a paid medium, or an fbclid, igshid, ttclid or li_fat_id click id |
| Paid Video | A video platform with a paid medium |
| Paid Other | A paid medium, or a known paid source, that matched none of the above |
| Organic Shopping | Unpaid arrivals from a store or product listing |
| Organic Social | Unpaid arrivals from a social network |
| Organic Video | Unpaid arrivals from a video platform such as YouTube or Vimeo |
| Organic Search | Unpaid arrivals from a search engine |
A mail source, or newsletter or email in the tags | |
| Affiliates | Affiliate networks: Awin, CJ, Impact, Rakuten, ShareASale |
| Audio | Audio advertising: medium audio or audio-ad |
| SMS | Text messages, tagged sms as source or medium |
| Mobile Push Notifications | Medium ending in push, or containing mobile or notification |
| AI Assistant | Assistants: ChatGPT, Claude, Gemini, Perplexity, Copilot, DeepSeek, Doubao and about a hundred more |
| Referral | A named source that matched none of the rules above |
| Direct | No referrer, and no campaign tag to name a source |
Visits from Google's assistants, Gemini, AI Studio and NotebookLM, are filed as AI Assistant from mid-September 2026. Before that they were counted as Google under Organic Search, and that history is not reclassified, so a period that spans the change shows Gemini appearing and Google search dipping by the same amount.
A return from a sign-in page (Google, Microsoft, Apple, Yahoo) does not count as a referral: the visit keeps its earlier source or shows as Direct. Earlier sign-in returns have been corrected in historical data as well. Logins through GitHub, Facebook or X still count as referrals, because the browser passes only the site's address and such a login looks like an ordinary link.
What decides
Two inputs, and most channels accept either.
The referring domain. About 2500 hosts are mapped to one of eight categories, so reddit.com resolves to Organic Social and perplexity.ai to AI Assistant without you tagging anything. A second list marks sources as paid.
Your UTM tags. utm_medium, utm_source and utm_campaign are read for every channel, and for some they are the only way in.
When a visit arrives with no referrer at all but does carry utm_source, that tag becomes the source. This is what keeps a visit from an assistant's app, a mail client or a browser with a strict referrer policy out of Direct: the address says where it came from even though the browser sent nothing. The shorter spellings source and ref are read the same way.
One exception, and it is worth knowing if you run redirects. A tag arriving with utm_medium set to an HTTP status code, 301 or 302 and the rest, is a site's own redirect describing itself rather than a source, so it is ignored. No campaign runs on medium 301.
Order matters
Paid is tested before organic. A visit from Facebook carrying utm_medium=cpc is Paid Social, never Organic Social.
Referral sits second to last. It is not "traffic from websites" in the narrow sense: it collects every visit that has a referrer and matched none of the rules above.
Tagging is required
Four channels have almost nothing to recognise on their own. SMS, Audio, Cross-network and Display are reached through your tags or an advertising click id, and nothing else. A text message carries no referrer, so an untagged link in one is counted as Direct.
The same holds wherever the medium decides the answer. A creator's link in a video description resolves to youtube.com and lands in Organic Video; tagging it utm_medium=cpc moves it to Paid Other instead.
Default channels
Analytics tools group sources into channels by rules. Here are the main ones, as the industry generally defines them.
Organic Search
Unpaid transitions from search engines. Triggered when source is a known search engine and medium is "organic" or empty.
Conditions:
- Source matches search engine list (Google, Bing, DuckDuckGo)
- Medium = organic or missing
- No paid ad parameters
Paid Search
Paid traffic from search engines. Needs both: source is a search engine, medium signals paid.
Conditions:
- Source = search engine
- Medium contains: cpc, ppc, paidsearch, paid
Direct
Visitors with no detectable source. Includes bookmarks, direct URL entry, and clicks from mobile apps without UTMs.
Dark traffic problem
Up to 60% of mobile organic traffic gets misclassified as Direct due to browser and app limits. Transitions from messengers, PDFs, and email clients often lose referrer data.
Referral
Transitions from other sites via plain links. Catches all external traffic that doesn't match other channel rules.
Social
Traffic from social networks. Modern analytics platforms support large lists, including niche and regional networks.
Recognized platforms:
- Main: Facebook, Instagram, LinkedIn, Twitter/X
- Video: YouTube, TikTok, Vimeo
- Professional: GitHub, Stack Overflow
- Messengers with social features
Email needs UTM tagging. Without it, traffic falls into Direct, hiding email marketing performance.
Display
Traffic from display ads, banners, native, programmatic. Use medium=display or banner.
Affiliates
Partner traffic. Tag with medium=affiliate.
Custom channels
Default channels don't always fit. Custom channels adapt grouping to your business.
Statable has no custom channels: the grouping is fixed and cannot be redefined per site. What follows describes the general practice on platforms that do offer it.
When to add them
Problem: Specific acquisition channels
Solution: Separate channels for:
- Podcasts and audio ads
- Influencer marketing
- Offline QR codes
- Internal communications
Problem: Default channels are too broad
Solution: Split into subcategories:
- Social → Paid Social / Organic Social
- Search → Brand Search / Non-Brand Search
- Email → Newsletter / Transactional / Automation
Problem: Need grouping by business criteria
Solution: Channels by:
- Funnel stage (Awareness / Consideration / Decision)
- Geography (Local / National / International)
- Product line
Setup rules
Setup principles
- Order matters
Rules run top-down. Put specific rules above general ones.
Use RegEx
Test on history
Check rules against historical data before deploying.
- Document the logic
Maintain a reference describing each channel and its conditions.
Unclassified traffic
The "Unassigned" or "(other)" channel signals tagging issues.
| Cause | Fix |
|---|---|
| Missing UTM parameters | Mandatory tagging on all campaigns |
| Typos | Use auto link generators |
| New sources | Update channel rules regularly |
| Technical issues | Verify parameter transmission |
Attribution and channels
Single-touch attribution is giving way to multi-touch.
Single-touch
First-Touch
100% to the first interaction. Good for evaluating new acquisition channels and brand campaigns.
Last-Touch
Industry default. All credit to the last click before conversion. Ignores prior interactions.
Multi-touch
Splits conversion value across all touches.
Linear
Equal credit across touches. Simple, but ignores funnel-stage importance.
Linear example
Journey: 1. Organic Search (blog) → 25% 2. Paid Social (retargeting) → 25% 3. Email (newsletter) → 25% 4. Direct (return) → 25%
Total: 100% conversion
Time-Decay
Touches closer to conversion get more weight. Logic: recent matters more.
graph LR
A[First touch<br/>10%] --> B[Middle touch<br/>20%]
B --> C[Second-to-last<br/>30%]
C --> D[Last touch<br/>40%]
D --> E[Conversion]Position-Based (U-shaped)
First touch 40%, last touch 40%, middle touches share 20%. Recognizes acquisition and closing.
Data-Driven
ML computes each channel's real contribution from historical data. Most accurate, needs lots of data.
Picking a model
| Business model | Recommended | Why |
|---|---|---|
| Short-cycle e-commerce | Last-Touch or Time-Decay | Focus on conversion channels |
| Long-cycle B2B | Linear or Position-Based | All funnel stages matter |
| Subscription | Data-Driven | Complex paths, many touches |
| Content project | First-Touch | Audience acquisition first |
Performance analysis
Key metrics
Channel metrics
Volume:
- Sessions
- Users
- New Users
Quality:
- Bounce Rate
- Pages/Session
- Avg. Session Duration
Conversion:
- Conversion Rate
- Revenue/Session
- ROAS
Segmentation
Channel analysis without segmentation stays shallow. Slice by:
Mobile and desktop behave differently:
- Mobile: higher bounce, lower conversion
- Desktop: more pages/session, higher AOV
- Tablet: middle ground
Performance varies by region:
- Local: SEO and Direct dominate
- International: Paid Search grows
- Developing markets: Social leads
Behavior shifts by first-visit time:
- New users: respond to Paid
- Regular: prefer Direct and Email
- Returning: respond to Retargeting
Cross-channel analysis
Channels don't work alone. Look for synergies.
Cross-channel example
Discovered patterns:
- Organic Search + Email: 8.2% conversion
- Paid Search + Retargeting: 6.7% conversion
- Social + Email: 2.1%
Action: Shifting budget from Social to SEO content lifted overall conversion by 23%.
Limits of standard tools
Standard platforms cap channel management. GA4 limits custom channels and blocks retroactive grouping logic on historical data.
Statable removes those limits. Unlimited custom channels. Rules apply retrospectively. Planned: dynamic channels that adapt to traffic shifts automatically.
We focus on the dark traffic problem. Statable will use behavior patterns and contextual signals to recover lost attribution.
Automation
Dynamic rules
Manual channel management breaks at scale.
Automation
Rule-based:
- Auto-create rules from patterns
- Detect new sources, suggest classifications
- Validate rules against anomalies
ML-driven:
- Cluster sources by user behavior
- Predict the most likely channel for untagged traffic
- Find hidden source connections
Monitoring and alerts
| Alert type | Trigger | Action |
|---|---|---|
| Unassigned growth | >5% of total | Check new sources |
| Channel anomaly | >30% deviation | Review campaign changes |
| New source | Unknown source/medium | Add classification rule |
| Attribution shift | >20% model shift | Review channel weights |
Privacy impact
iOS App Tracking Transparency
ATT cuts attribution accuracy 15-25%, mostly on mobile. Adapt by:
- Probabilistic attribution
- More first-party data
- SKAdNetwork for iOS
Third-party cookie removal
Cross-site tracking is blocked in browsers, breaking cross-domain attribution. Solutions:
- Server-side tracking
- First-party identifiers
- Privacy Sandbox APIs
New traffic sources
AI platforms create new classification problems.
AI traffic
ChatGPT, Claude, Perplexity recommend sites without standard referrer data. New approaches:
- "AI Referral" channel
- UTM in prompts
- Behavior pattern analysis
Business metric integration
Customer Lifetime Value
CLV reveals long-term channel value:
graph TD
A[Acquisition channel] --> B[First purchase]
B --> C[Repeat purchases]
C --> D[CLV]
D --> E{Analysis}
E -->|High CLV| F[Increase investment]
E -->|Low CLV| G[Optimize or cut]Contribution margin
Real contribution after costs:
| Channel | Revenue | Ad Spend | Operational Cost | Contribution Margin |
|---|---|---|---|---|
| Organic Search | $100K | $0 | $15K | 85% |
| Paid Search | $150K | $60K | $10K | 53% |
| Social Paid | $80K | $45K | $8K | 34% |
| $120K | $5K | $5K | 92% |
Grouping sources into channels and reading them through attribution drives smart marketing decisions. Privacy shifts and new platforms require flexibility and constant strategy updates.
Ready to take control of your web analytics? Try Statable free for 30 days. No credit card required, full feature access, built for GDPR. Start your free trial or view a live demo.