OUR METHODOLOGY / VERSION 1.0.0

A number means more
when you know what’s behind it.

SOSH connects genuine social influence to economic value. Every score has components. Every estimate has assumptions. Every deduction has a reason.

Six dimensions. One canonical score.

SOSH Score = 1000 ×
(0.25A + 0.20T + 0.20C
 + 0.15R + 0.10G + 0.10S)
A · Audience quality25.0%
T · Trust & influence20.0%
C · Content performance20.0%
R · Consistency & reliability15.0%
G · Growth quality10.0%
S · Commercial scarcity10.0%

Value that earns your trust

Audience quality measures real audiences, authentic engagement, geographic and demographic relevance, purchase intent, and stability. Canonical calculations use the creator’s declared market; campaign-specific fit is a separate context.

AQS = 0.25R + 0.20E + 0.15G
    + 0.15D + 0.15I + 0.10S
Audience quality, expected outcomes, and composite social value are derived from fictional observations and configurable prototype recipes. No production ML accuracy is claimed.

Translate actions into economic value.

Value per view begins with the probability of a click, the probability of a conversion, and customer value. Media-equivalent value is added, then explicit audience, platform, content, and brand-fit multipliers are applied.

Base VPV = P(click | view) × P(conversion | click) × Customer Value
Adjusted VPV = Base VPV + CPM / 1000
Creator VPV = Adjusted VPV × AQS × PF × CF × BF

Base VPC = P(conversion | click) × Contribution Margin
Adjusted VPC = Base VPC × AQS × BF × LPQ

Engagement Value = Base Engagement Value × Intent Weight × AQS × BF
Post Value = Views × VPV + Σ(Engagements × VPE) + Clicks × VPC
           + Conversions × Contribution Margin + Asset Value + Exclusivity

These additive terms form a composite social-value measure. They do not represent actual sales revenue, profit, or ROAS. Unit values shown after risk adjustment are never fed into another final risk deduction.

Follow the authenticity evidence

Risk Discount = 1 -
(0.40F + 0.20Ea + 0.15Va
 + 0.15Ca + 0.10Ga)

Trusted Value = Gross Value
              × (1 - Risk Discount)

Each authenticity input ranges from zero to one, where one is fully authentic. Seven detectors inspect follower spikes, comments, watch time, click follow-through, geography changes, related bot clusters, and duplicated traffic. Overlapping evidence uses a bounded maximum penalty per dimension.

Understand the intent of an action

like0.10
reaction0.12
comment0.40
share0.75
save0.85
profile Visit0.65
direct Message1.00
Link clicks and purchases are valued directly. They are not counted again in weighted engagement.

Keep financial outcomes precise.

Revenue = attributed order revenue
Contribution profit = order contribution - all-in brand spend
ROAS = attributed revenue / all-in brand spend
ROI = contribution profit / all-in brand spend
CPM = all-in spend / impressions × 1,000
CPC = all-in spend / clicks
CPA = all-in spend / attributed conversions

Attribution uses a valid creator coupon first, then the last eligible tracked click within seven days. Each conversion is credited once. Unmatched conversions remain unattributed. Incremental lift requires test/control evidence; otherwise it is unavailable. Zero denominators produce an unavailable state.

Different follower-value bases, explicitly separated.

Annual VPF = Annual Views × VPV × Retention
Operational VPF = ARF × ER × pre-AQS Adjusted VPV × AQS × LR
Lifetime VPF = Σ[t=1..T] Annual Views[t] × VPV[t] × Retention[t] / (1+d)^t

Follower lifetime value is a separate valuation basis. It is not silently added to annual activity value. Annual value is marked as a forecast when observed history is insufficient for a trailing 365-day calculation.

Confidence without false precision

The deterministic prototype model combines observed history with a platform/category/format prior using n / (n + 10). Recent evidence has greater weight, with a configurable 90-day half-life. Sparse histories and missing data widen the simulated ranges. These intervals demonstrate uncertainty; they are not calibrated production accuracy claims.

Commercially useful. Commercially independent.

Subscriptions control access to features, not the score. Sponsorship buys explicitly labeled placement, not a different valuation. Verification is earned after seven checks and has a recorded expiry. Custom enterprise models are private perspectives and cannot replace the public canonical score.