Google Analytics Direct Traffic: Investigate a Spike With Your Own Data

Use session scope for a traffic spike

Begin with sessionSource, sessionMedium and sessionDefaultChannelGroup for the property and period you are investigating. First-user source describes how a user was first acquired, which is a different question. A person acquired through organic search can return in a later session with different attribution.

Google’s direct-traffic explanation describes missing referral information and common collection scenarios. Keep the returned source, medium and channel labels together so another analyst can reproduce the segment.

GA4 attribution also has source-precedence rules. The absence of a referrer on one visit does not necessarily create a new Direct-labelled session in every circumstance. Google’s campaign and traffic-source guidance explains session-source processing. Avoid claiming that every direct visit represents a person who typed your address.

In SourcetoMCP, use supported dimension filters to isolate the intended session segment before the row limit. Check whether the returned labels and totals match the question before adding more breakdowns.

Quantify both direct-session volume and share

Suppose a hypothetical previous 28-day period has 10,000 total sessions and 1,000 Direct sessions. The current comparable period has 15,000 total sessions and 3,000 Direct sessions. Direct volume has increased by 2,000 sessions, or 200%, while total sessions have increased by 50%.

Direct share moves from 10% to 20%, a ten-percentage-point rise. Showing only the 200% increase hides the growth in overall visits. Showing only the share hides the absolute size of the investigation.

The demo page accounts for 1,400 ÷ 2,000 = 70% of the additional Direct sessions in this example. That makes its distribution and redirects useful next checks. It does not prove that 1,400 email clicks lost their tags.

Use consistent hosts, landing-page matching rules and period boundaries. Keep unavailable rows separate from zeros, and retain any row-limit or incomplete-data warning.

Compare the segments that changed

Retrieve a daily Direct-session trend to locate the start of the change. Then inspect landingPage, deviceCategory, country or browser in separate focused reports. Adding every dimension at once creates small groups and can obscure the larger pattern.

A spike concentrated on the demo page after an email send supports checking the email links. A spike concentrated in one browser after a site release supports a browser and collection test. Neither pattern establishes the cause without inspecting the relevant links, release or test session.

The connector can return pageReferrer where available, but it does not reconstruct referrers that were never collected. Missing or aggregate referrer values cannot identify the original person, app or message behind every session.

Compare engagement and the property’s key-event measure alongside sessions. Use the bounce-rate guide to interpret fractions and scope. Sparse outcomes or unusual geography can justify further investigation without proving bot traffic from these reports alone.

Test external campaign links and their redirects

For recognised AI assistant sources, follow the AI traffic analytics workflow to inspect GA4’s AI Assistant channel. Keep unidentifiable Direct sessions outside that measured total.

Inspect the actual links distributed by your campaign, including shortened URLs and intermediate redirects. Test the complete route to the landing page. Record whether the intended parameters survive and whether the correct GA4 property receives the resulting activity.

Google’s campaign URL guidance explains parameters such as utm_source, utm_medium and utm_campaign. For an illustrative external newsletter link, a suffix of utm_source=newsletter&utm_medium=email&utm_campaign=october_demo names the source consistently.

Use that convention on the external campaign link it describes. Do not add campaign UTMs to ordinary internal navigation to manufacture a source label. Keep personal information out of URLs and preserve existing click identifiers or valid tracking parameters when repairing a link.

For Google Ads, review auto-tagging and the effective tracking settings with the tracking-template audit. A manual UTM example does not replace validating the advertising account’s actual measurement configuration.

Ask AI to rank investigations using observed evidence

Investigate a Direct-traffic change in GA4 property [ID] between [two complete periods]. Make no changes. Start with session source, medium and default channel group, then filter the observed Direct segment. Quantify total sessions, Direct sessions, share and percentage-point changes. Retrieve daily, landing-page and device reports with supported fields and filters. Rank three hypotheses using observed evidence, list the manual link or browser test for each, and state what cannot be recovered. Do not relabel historical sessions or claim Direct means typed URLs only.

Add your campaign-send calendar and site-release dates as separately labelled evidence. SourcetoMCP’s traffic report does not know when your team sent a newsletter or changed its redirect configuration unless you supply that information.

Use the GA4 missing-values guide for (not set) dimensions. Direct attribution and a missing dimension value need separate diagnosis.

Verify future collection after a fix

Once a specific link or collection problem is confirmed, approve the correction in the system that owns it. Repeat the full click-through test and inspect GA4’s relevant debugging or realtime views. A successful configuration edit needs collection evidence too.

Record the fix date and compare future complete periods with the same session-source definitions. A change in reported Direct share can also reflect a different campaign mix, so retain total visits and the campaign calendar in the review.

Do not redistribute historical Direct sessions among channels using an invented percentage. Keep the original observations and the investigation notes, then use the corrected collection process for future decisions.