The Vanishing Traffic Audit: Finding Dark Social in a Real Analytics Export

Marketing Academy · Field Work●Social Media Marketing

All projects·From: Dark Social Strategy: Measuring and Driving Invisible Shares

CoreAudit· 45 minutes

StoneCo

Objective: Given a week of session data by source and day for a new pricing guide, separate a genuine dark-social wave from an ordinary paid-social spike, then recommend which two pages deserve tracking investment first.

You're the growth analyst at StoneCo, the Brazilian merchant-payments fintech, reviewing last week's traffic to a new small-business pricing guide after 'direct' sessions spiked.

Read the daily session export, tell dark social apart from a paid-social spike using the shape of the curve, then rank three landing pages by how urgently each needs share tracking.

Before you start

What you'll need

Free path (everything below is enough to finish)

Google Sheets

FreePivot and chart the analytics export by day and by page

Free, and pivot tables are enough to reveal both the wave pattern and the page ranking without a specialized tool

The process

2 steps

Step 01 of 02

Spotting dark social from the shape of a direct-traffic spike

The lesson says dark social arrives in waves, fastest in the first few hours then tapering as it circulates slower networks, while a paid-social spike peaks fast and drops sharply once the campaign budget or feed placement ends.

Direct sessions to /guia-de-precos-pme: 420 on day 1, then 310, 240, 190, 150 over the next four days. Paid-social sessions to the same page: 380 on day 1, then 90, 40, 20, 15. Which curve is the dark-social pattern, and which is the paid campaign winding down?

Google Sheets— The exported session-by-day-by-source table, pivoted into a line chart per source.

Procedure

Import the export and build a pivot table of sessions by day, split by source
Chart the direct-traffic line and the paid-social line side by side
Compare decay rates: a slow multi-day taper points to dark social, a sharp one- or two-day drop points to a paid or referral spike ending
Confirm by checking whether the paid campaign's scheduled end date lines up with the sharp drop

Sample output

Day 1  2  3  4  5
Direct:  420 310 240 190 150  (slow taper, ~25%/day)
Paid:    380  90  40  20  15  (sharp drop, campaign ended day 1 evening)

Healthy

Correctly reads the 420->150 direct-traffic taper as dark social continuing to circulate, and the 380->15 paid line as a campaign that simply ended.

Unhealthy

Treats both curves as 'the campaign working,' or dismisses the direct-traffic taper as unexplained noise instead of a trackable pattern.

What this means

A slow taper in direct traffic is not noise, it is dark social sharing decaying naturally as the guide moves through progressively smaller private networks.

So what do I do about it?

Symptom	Action	Effort
Direct traffic spikes and slowly decays after every content launch, unexplained	Log the decay shape each time and treat any multi-day taper as a dark-social signal worth instrumenting	30 min

YouYou can do this yourself, no engineering access required.

Step 02 of 02

Prioritizing which pages get share tracking first

The lesson's toolkit section frames branded links, share buttons, and copyable stat callouts as an investment, best spent on the content already generating the most untracked traffic.

Three pages have untracked-direct-session totals this month: the pricing guide (1,310), the careers page (90), and the product changelog (40). With budget for one round of share-tracking instrumentation, where do you start?

Google Sheets— The same session export, now grouped by landing page instead of by day.

Procedure

Group untracked direct sessions by landing page for the full month
Rank pages by untracked-session volume
Recommend branded short links and a share survey for the top page only, not all three
Note the other two pages as 'monitor, revisit next quarter' rather than instrumenting them now

Sample output

Untracked direct sessions by page (30 days)
1. /guia-de-precos-pme: 1,310
2. /carreiras: 90
3. /changelog: 40

Recommendation: instrument the pricing guide first; the other two don't yet justify the setup cost.

Healthy

Instrumentation budget goes to the pricing guide, the page carrying 93% of the untracked volume.

Unhealthy

Spreads tracking evenly across all three pages, spending equal effort on a page with 40 untracked sessions and one with 1,310.

What this means

Share-tracking setup has a fixed cost per page, so it only pays off on pages where the untracked volume is large enough to justify it.

So what do I do about it?

Symptom	Action	Effort
Tracking effort spread thin across many low-traffic pages	Rank pages by untracked-session volume and instrument only the top one or two	5 min

YouYou can do this yourself, no engineering access required.

Final deliverable

A one-page attribution memo identifying which spike is dark social and which two pages should get share tracking first, ranked by untracked-session volume.

See a reference example

Sample output

Nubank, Traffic Attribution Memo (excerpt)

Finding: the /guia-taxas direct-traffic spike (610 -> 480 -> 350 -> 260 over 4 days) is a dark-social wave, not noise; paid social on the same page ended day 1 and dropped to near-zero by day 2.

Recommendation: instrument /guia-taxas with a branded share link and a one-question intake survey before the next content push.

Success criteria

You're done when you can:

Correctly separates the dark-social taper from the paid-social drop using decay shape
Ranks landing pages by untracked-session volume and recommends instrumenting only the top page