Siftsy Knowledge Base
Working with Siftsy Score

Intro to Siftsy Score V3

Siftsy Score is a 1–10 temperature check for a comment section: it tells you how the comments are reacting to the content. Version 3 is the current scoring model, and it adds a Content vs. Topic split, built-in guardrails, and a three-band readout.

Loading custom HTML embed

The three signals

Siftsy Score blends three independent 1–10 signals, each read by context-aware AI:

  • Sentiment — how commenters feel about the post and the topics it discusses

  • Relevance — how on-topic the conversation actually is

  • Consensus — how commenters are reacting to each other

Sentiment carries the most weight (66%), with Relevance and Consensus at 16.5% each.

What Version 3 adds: Content vs. Topic

Traditional sentiment analysis collapses "I don't agree with this video" and "I hate this subject" into one scale. Version 3 splits Sentiment into two separate readings for every comment:

  • Content Score — how commenters feel about this post and its creator

  • Topic Score — how commenters feel about the subjects being discussed

Overall Sentiment is the average of the two, but the split is where the insight lives. A high Content score with a low Topic score means the audience loves the execution but not the subject — revisit the message, not the creative. The reverse means the subject resonates and the post is what's missing. One number becomes a diagnosis.

Built-in guardrails

Version 3 also hardens the score against the ways comment sections mislead:

  • Negativity cap. If a meaningful share of comments are outright hostile (scoring below 3), the score's ceiling drops — for example, a section that is 12% very negative can score at most 8.5. A vocal positive majority can't paper over a real hostile contingent.

  • Engagement weighting. Comments with more likes and replies count more, on a logarithmic curve: every comment keeps at least 1× weight and the most-engaged comment counts up to 5×. The multiplier applies only to the sentiment readings — Relevance and Consensus are never scaled by likes.

  • Reply hierarchy. A 40-message argument nested under one comment no longer outweighs the rest of the thread. Top-level comments carry full weight; replies share a capped pool.

Reading the score

Version 3 uses a simple three-band gauge:

Score

Read

7.0+

Immaculate — positive and supportive

4.5 – 6.9

Mixed — the section is split

< 4.5

Off — substantially negative and critical

Use the number to rank posts and spot the ones that deserve a closer look. Then use the breakouts — Sentiment for mood, Relevance for focus, Consensus for infighting, and Content vs. Topic for what the audience is actually reacting to — to understand why.

Next steps

Now that you know what the score measures, read it in the product and dig into the comments behind it:

Was this helpful?
Was this helpful?