Detect statistical drops in shipment frequency on any sourcing lane. A 2-sigma drop leads value disruption by four to six weeks. Time to pre-buy.
The early-warning index flags telmisartan IN to KE at -27.6 sigma: shipment baseline 189 per year, 2024 actual 169 on a tight prior-year standard deviation. Twelve other lanes cross the -1 sigma early-warning threshold; four of those are above -2. Each is a 4-to-6 week lead time on a likely value disruption. Where the underlying value has not yet dropped, the recommended action is a pre-buy or an alternate-origin qualification before the lane confirms cold.
Upload or define your sourced lanes. Each lane gets a two-year shipment-count baseline (mean + standard deviation) computed from 2022 and 2023.
Latest-year shipment count, normalised against the baseline. Z-scores below -1 are tracked; below -2 fire as alerts; below -3 escalate to high severity.
Citation-anchored brief with the baseline, the latest count, the z-score, and a recommended action (pre-buy, alternate origin, escalate to dual-source).
Shipment count drops 4-6 weeks before value does. Catching the frequency drop gives time to pre-buy at trend pricing rather than disruption pricing.
A -2 sigma drop is the default alert threshold; the data is normalised against the lane's own variance. Tune to -1 for an aggressive early-warning posture; to -3 for emergency-only.
Lanes that recovered after a prior z-score alert get a 'reverted' flag. The brief tracks the alert history so a buyer doesn't fire twice on the same incident.
A z-score alert at -3 sigma combined with a 30% YoY value drop is a near-certain disruption. Both signals fire on the same dashboard.
A real chat thread in PharmaDB. Type a question, the AI agent runs the tools, the answer lands as a saveable note.
pharmagraph_query Twelve plants exporting on lanes currently flagged by the early-warning index · plant inspection class and recent observations shown.
The disruption early-warning index is the leading-indicator feed for sourcing risk. PharmaDB normalises shipment-count drops against each lane’s own two-year baseline and surfaces statistically significant outliers as alerts. The view is molecule_country_flow with z-score computation; the join is to lane context for severity scoring.
Shipment count is the leading indicator; value is the lagging one. A 20% drop in shipments often shows up four to six weeks before the corresponding drop in declared USD value. Catching the count drop gives the lead time.
Mean and standard deviation of annual shipment count on the lane, 2022 and 2023. Lanes with fewer than 50 baseline shipments are excluded; the z-score becomes unreliable on thin lanes.
The baseline standard deviation was tight (low year-on-year variance) so the actual deviation of 20 shipments translates into a very large z-score. The shape says the prior pattern was steady and the latest year departed from it materially.
Yes. When subsequent refreshes pull the z-score back above -1, the alert is marked 'reverted' rather than 'active'. The brief retains the history so the buyer can track the recovery.
A z-score alert at -3 sigma on a high-HHI lane escalates the dual-source qualification from routine to urgent. The composite signal carries more weight than either alert in isolation.
Trade aggregates refresh weekly. Z-scores recompute on every refresh. The catalog row carries the last-computed timestamp surfaced on every result page.
Bring the molecule, the lane, or the supplier you're sourcing this week. The AI agent runs it on PharmaDB in 30 minutes. You keep the brief.