Proprietary methodology
LSI Market Growth Forecast
Proprietary indicator of market growth propensity for molecules and medicines
Model lsi-explainable-v1 · Engine engine-1.0.0 · Document rev. 1
The question this indicator answers
How likely is this molecule to gain commercial relevance over the next 12 months?
The indicator expresses observed propensity, not calibrated statistical probability. It is an analytical prioritisation instrument, not a revenue projection.
Propensity scale 0–100
0–19
Strong decline
Consistent loss of commercial relevance
20–39
Decline
Falling signals across most dimensions
40–54
Stable
No dominant direction in the period
55–69
Moderate growth
Positive signals, not yet widespread
70–84
Likely growth
Positive convergence across dimensions
85–100
Strong growth
Consistent, corroborated expansion
Illustrative example
Six weighted dimensions
Presence, channel breadth and supply continuity
Informational demand for brand and molecule
Studies, registrations, rulings and normative acts
Capacity, funding, launches and plant expansion
Entrants, generics, similars and portfolio moves
Price trajectory, dispersion and stock continuity
Versioned, auditable weights, configurable per account.
Score composition
Score = Σ ( dimension_score × dimension_weight ) ÷ Σ valid weights
Dimensions without sufficient evidence are excluded and weights renormalised — never filled with zero.
Volume alone does not set the score
Every piece of evidence is weighted by recency, source quality and a per-source contribution cap. A hundred mentions from one source weigh less than ten independent, corroborated ones.
Recency decay
| Up to 7 days | 1.00 |
| 8 to 30 days | 0.85 |
| 31 to 90 days | 0.60 |
| 91 to 180 days | 0.35 |
| Over 180 days | 0.15 |
Collection and sources
Authorised connectors run recurring structured collection and map public pages and documents, always from open sources.
- —Taxonomy of 16 source categories, each with its own reliability weight
- —Regulatory and scientific sources carry the highest weight; aggregators the lowest
- —Every piece of evidence stores origin, date, canonical URL and content hash
Forecast Confidence
An independent 0–100 score measuring evidence strength: volume, source diversity, freshness, time coverage, presence of official sources and dimensional completeness.
| 0–39 | Insufficient — score is not published |
| 40–59 | Limited — read with caution |
| 60–74 | Moderate |
| 75–89 | High |
| 90–100 | Very high |
Integrity rules
- —Deduplication by canonical URL and content hash
- —5% tail winsorisation against outliers
- —35% cap on any single source per dimension
- —Minimum of 12 pieces of evidence and 3 distinct sources to publish
- —Immutable history; recalculation creates a new model version
- —Each score stores an input hash for reproducibility
Normalisation
Each dimension combines level (z-score clipped at ±2.5 deviations), growth over a 90-day rolling window against the previous 90, and percentile position in its own 7-day moving average series.
Projection
The projection applies half of the observed rate, damped by confidence and flattened logarithmically over time. The interval widens with horizon and uncertainty.
Declared limitations
- —Based solely on observable public signals; no private sales data
- —Not a revenue forecast, price prediction or investment recommendation
- —Abrupt regulatory or supply changes may precede the signals
- —Molecules with low public coverage tend to score lower confidence