Scoring & Signals
Every opportunity card shows three headline metrics: Fit, Relevance, and Survival. Together they answer 'is this worth engaging with, and can I actually post here safely?' in a glance.
Using the Scores Together
A quick mental model:
- Fit tells you whether to engage at all — high Fit = reply is probably worth your time.
- Relevance tells you why it's in your feed — is it genuinely about your space, or adjacent?
- Survival tells you whether you can safely engage — HIGH means go ahead, LOW means consider skipping even if Fit is strong.
High Fit + LOW Survival is the trickiest combo: the LLM thinks this is a great match, but the subreddit has a history of removing similar posts. Lead with helpful value, skip the product link, and be prepared for the comment to get pulled — the reply is still worth trying if your account handles takedowns without penalty.
The Three Headline Metrics
Each card shows these side-by-side so you can make a call in a few seconds. Tap the ⓘ icon next to Survival to see the full breakdown, including internal pipeline signals.
Fit (0–100)
"All things considered, how good is this post for my product?"
Fit is our daily curation LLM's overall judgment on a post. Unlike Relevance (which only measures topical similarity), Fit blends several things at once:
- how closely the post matches your product profile,
- how strong the buying intent is,
- how safe the subreddit is to engage in,
- how fresh the post is and whether the thread is already saturated.
A post can have high Relevance but low Fit (e.g. topically on-point but in a hostile subreddit, or buried under 80 comments), or low Relevance but high Fit (e.g. asking for your exact use case in a less obvious way). Fit is the number to trust when you can only look at one.
Fit only appears on posts that have been through daily curation — not every post gets a Fit score. Curation picks the top ~50 candidates per day and has the LLM rank them. If a card has no Fit pill, fall back to Relevance and Survival.
Relevance (0–100%)
"Is this post even about my space?"
Relevance is a cosine-similarity score between the post and your Product Profile, computed from text embeddings. It's a pure topical signal — it doesn't know anything about subreddit risk, intent, or freshness. A high Relevance % means the post mentions problems, use cases, or language that matches your profile.
Use Relevance when you want to filter: in the Notifications page you can set a minimum Relevance % so your daily digest only includes posts above that bar. We default to 60%.
If you're seeing high Relevance but low Fit (or low Relevance but high Fit), that's the signal working correctly. The two measure different things — disagreement usually means the subreddit risk, intent strength, or saturation is pulling Fit away from the pure topic match.
Survival (LOW / MEDIUM / HIGH + probability)
"If I post here, will my comment stay up?"
Survival is a per-post probability that the thread won't be removed by moderators. It's computed from that subreddit's history — how often posts like this have been removed, how strict moderation is, whether similar content has been flagged before.
Internal Pipeline Signals
Open the ⓘ icon on any card to see these. They feed into Fit, Survival, and your feed ranking, but we keep them off the headline row because the three scores above are more actionable at a glance.
Opportunity (HIGH / MEDIUM / LOW)
The pipeline's composite tier built from intent + freshness + saturation + relevance. Before we introduced Fit this was the headline signal; today it's retained as a filter and kept accessible in the info popover.
Intent Direction
What the person is trying to do:
Intent Strength
How urgently the person needs a solution:
Freshness
How old the post is — earlier replies get more visibility:
Saturation
How crowded the comments section is:
Stealth Ad Detection
Some posts look like genuine questions but are actually disguised promotions ("I found this amazing tool...", "Not affiliated, but..."). We flag these so you don't waste time replying to someone who's just advertising.
Risk Similarity Flag
If posts with similar content in this subreddit have been removed before, Survival is automatically bumped into the LOW tier:
How the Scores Are Calculated
Fit
Produced by the daily curation LLM once per day for the top ~50 candidates per organization. The model reads the post, the subreddit's risk profile, the intent signals, and your product profile, then returns a 0–100 score with a short reasoning that you can see in the daily email and the notifications page's digest history.
Relevance
Recomputed whenever your Product Profile changes or when a new post is ingested. We generate embeddings for the post (title + body + subreddit) and for your profile (description, use cases, differentiators), then store the cosine similarity as a 0–100% score per (post, organization). This is why better Product Profiles produce sharper Relevance numbers.
Survival
Built from each subreddit's observed removal patterns: what fraction of promotion-adjacent posts get removed, how mod response time correlates with similar content, and whether this post resembles past takedowns. When we don't have enough observations yet (fewer than 5 samples in that subreddit), we fall back to a cautious default.
Opportunity (internal)
Computed deterministically from the pipeline signals above:
- Automatically LOW if Relevance is very low or the post is too old
- HIGH needs strong Relevance + high intent + fresh post + few comments
- MEDIUM covers everything in between
- Posts flagged as SELLING or STEALTH AD get downgraded by one tier