Value Stories
How the demand signal algorithm creates value. Each story is an intent flow: a scenario triggers an intention, actions produce artifacts, outcomes prove value.
Does the signal get scored?
Raw platform signals become typed demand with strength rating.
Content published 24 hours ago. Platform APIs return 47 saves, 12 shares, 3 DMs, 200 clicks, 0 conversions.
Score the aggregate response into a typed demand signal with strength rating.
Score computed in <5ms. demandType = 'active' (DMs present). signalStrength = 'moderate'. Same inputs always produce same output.
Algorithm returns a score but ignores timeDelta. A save at hour 1 and hour 72 weigh the same. The score looks precise but the signal is noise.
Three content pieces published last week. Piece A got saves only. Piece B got DMs and replies. Piece C got 2 conversions.
Classify each piece's demand type correctly along the progression.
A = latent (saves only). B = active (DMs present). C = validated (conversions > 0). Classification is deterministic.
Everything scores as 'active' because the algorithm weights all signals equally. The progression from latent to validated is lost.
Does the signal stay honest?
Signals decay. Stale data presented as fresh is worse than no data.
Content published 7 days ago. Got 100 saves in hour 1 (viral spike), then 2 saves per day for 6 days.
Time-weighted score that reflects decay — early spike doesn't inflate the long-term signal.
Score at hour 1 is higher than score at day 7 for the same raw save count. Decay curve is monotonically decreasing.
Raw totals used without decay. A piece that got 100 saves on day 1 and zero since still scores high a month later.
Does the loop close?
SPCL predicted. Reality responded. The creator gets smarter.
Creator's content scored SPCL 7.2 (high tier). Demand signal comes back as 'validated' with strong strength.
Patch the creator's PowerIndicators so the next SPCL run reflects proven demand.
Next SPCL run produces higher Power score. The delta between predicted tier and actual demand tier shrinks over time.
powerFeedback only increments, never decrements. Creator with one viral hit and 50 misses still shows high Power. The loop flatters instead of calibrates.
Product manager reviews demand signals across 10 content pieces to score a PRD's Demand dimension.
Demand signals provide evidence for the 5P Demand score — scored input, not replacement.
PM cites 'Active demand: 8 pieces with DM responses, avg score 6.2' as Demand=3 evidence. Scored, not guessed.
Demand signal scores auto-populate 5P scores without human review. The algorithm replaces judgment instead of informing it.
Kill Signal
If after 30 days no content piece has been scored through the full SPCL → publish → demand-signal → feedback loop, the algorithm is solving a theoretical problem. Kill it.