T&S #2: A Rationale for Uncaught Spam in UGVC

Why Write This Blogpost?

Previously, I focused on spam on three platforms that best deliver delightful experiences: YouTube, Instagram and TikTok. Following that review, one may begin to wonder, “Why are there 'so many' uncaught cases of spam?”. To address this question, this blogpost presents a rationale on why there may be uncaught spam.

Published: 2025 June 25
Disclaimer: The following is an educational analysis based on the author's interpretation of publicly available data using the framework outlined in this post. It constitutes the author's opinion and is not a definitive assertion of fact.

Who Are The UGVC Ecosystem’s Participants?

For a UGVC ecosystem to operate, it requires the following participants:

Core Participants
  • Video platform providers
  • Content producers
  • Content consumers
  • Advertisers
Other Participants
  • Regulators
  • Software/Hardware Service Providers
  • Creative Process Service Providers

To date, spammers have participated in the ecosystem by posing as “Content producers”, “Content consumers” and/or “Advertisers”.

Why is not all spam caught?

Video platform providers work tirelessly to execute computer-program-driven (classifiers) and people-driven (investigations) approaches. Unfortunately, less than 100% of spam activities are stopped. Some reasons are presented below:

Rationale Category Reason
Computer-program-driven Approach Reasons The current models are trained on irrelevant data.
The current models are run on insufficiently frequent schedules.
The current models have computational resource constraints.
The current models made tradeoffs in recall and precision.
The latest models have not yet been deployed.
People-driven Approach Reasons The team has insufficient confidence that something is spam.
The team has deprioritized mitigation for this type of spam.
The team’s tools may need enhanced and/or new capabilities.
The team has not yet executed its mitigation approach.
Spammer-driven Approach Reasons The spammer community and thus activities have grown.
The spammer has developed novel, adversarial approaches.
The spammer has identified grey areas in a platform's policies.
Hypothesis: Mitigation efforts will bring spam asymptotically closer to zero, but never reach it due to the evolving nature of adversarial behavior.
Disclosures:
  1. I worked in the Trust & Safety for 7 years. In 0 of those years, I worked on the 3 UGVC platforms listed.
  2. I interview for open roles occasionally with written-about companies.
  3. I own equity in some companies written about.

Footnotes:
  1. "So many" is a subjective term. To some, one case satisfies the threshold.
  2. This rationale table should not be considered comprehensive.
  3. There is a cynical reason for uncaught spam: occasionally, analysts may be incentivized by performance reviews or promotion candidacy to deprioritize "low impact" spam in favor of "high impact" opportunities. While this behavior is exercised by a minority, it exists.