The Siftsy Content Selection Framework

Background

Comment sections are responses to content. To get usable insights into how audiences are reacting, you need a handle on the right content to evaluate — high-signal posts that actually carry the conversation worth analyzing.

You might be sourcing content links yourself, fielding requests from a client to analyze specific posts, or pulling from a social listening platform for trend analysis. In every case, you'll have more candidate content than budget. This framework decides what to enrich first.

It ranks your candidate pool from best to worst in four steps, each one tightening the selection:

• Volume — is there enough conversation in the comments to analyze?

• Relevance — does the post fit the brief and come from an account that matters?

• Coverage — does the selection mix reflect what the brief is actually asking?

• Exclusions — are there hard reasons to drop the post regardless of score?

Analyze through them in this order.

1. Volume — is the post big enough?

Small comment sections produce unreliable analysis. Social posts land roughly 2 comments per 1,000 views, so as a quick rule, drop anything under 25K views or 25 comments before scoring.

2. Relevance — score what's left

Give each remaining post a 0-3 on two things, then multiply.

Target — does it match the brief?

• 3 — Direct. On-brief content: the exact brand, campaign, product, or topic the analysis is built around.

• 2 — Adjacent. Competitor content, category-level conversation, or related campaigns that inform the brief by comparison.

• 1 — Tangential. Same audience or vertical, but not directly answering the brief's question.

• Drop — Off-brief. Doesn't connect to what the client is trying to learn.

Importance — does the account matter to the brief's goals?

• 3 — Owned or Tier-1. Accounts the brief is built around: the brand's own handle, official campaign partners, or target influencers that shape category narrative.

• 2 — Mid-tier. Established creators with credible audiences in the space or viral UGC that people are talking about.

• 1 — Long-tail. Smaller accounts whose content shows high quality reaction to relevant topics.

• Drop — Suspect. Spam, bought engagement, aggregators, parody accounts. Comment sections won't be authentic.

Sort posts descending by your Post Relevance (T × I).

3. Coverage — match the mix to the brief

A ranked list alone usually over-weights one kind of content. Coverage makes sure the selection reflects the brief's intended composition. Set a target mix up front, then walk the ranked list and promote/demote posts to hit it.

Common mixes by brief type:

Layer in these baseline rules unless the brief says otherwise:

• At least 2 posts per in-scope platform for every 10 selected

• No more than 2 posts from the same account in any 10

• Spread across the campaign timeline, not clustered on launch

If a content type is underrepresented in the pool ("as available"), document the actual mix you achieved so stakeholders see the constraint.

4. Exclusions — drop regardless of score

• Locked or no comments

• Off-language at high concentration

• Outside the brief's date window

• Incidental brand presence (background only, not the topic of the content)

Three examples

Social listening, 12,000 links. Heavy exclusions first (duplicates, dead links, off-language, locked comments, out-of-window). Volume floor cuts the pool significantly. T × I scoring promotes direct brand mentions. Coverage target: proportional to where the actual conversation is happening — if 60% of qualifying posts are on TikTok, the selection skews TikTok. Client analyzes from the top down until budget runs out.

Crisis management, viral content. Volume floor barely applies — everything's viral. Scoring emphasis shifts to Target and reach-weighted Importance. Coverage target: weighted toward narrative drivers, not balanced across platforms or accounts. If three creators are driving 80% of the conversation, three creators get analyzed. A small number of top-ranked posts is enough to draft a response.

Brand reputation across creators. Scoring stays balanced — long-tail authentic creators can outscore paid mega-creators. Coverage target: 1/3 owned brand, 1/3 influencer, 1/3 UGC. Cap any single creator at 1-2 posts, force platform balance, spread across time. Reputation is a portrait, not a highlight reel.

What to tell stakeholders

"From [X] candidate posts, [Y] cleared our engagement floor. We ranked them by relevance and account importance, then balanced the selection to [target mix — e.g., 1/3 owned, 1/3 influencer, 1/3 UGC]. Analysis reflects the top [N] posts in that ranked queue."

The ranking turns a budget constraint into a methodological credential — the selection logic holds whether you analyze 20 posts or 20,000.