Cobalt under pressure: Scenario Planning and Simulation of supply chain shocks in the DR Congo and China
- Steffen Konrath
- Sep 6
- 4 min read
For decision-makers in government and industry, the cobalt supply chain is not an abstract topic. It is one of the lifebloods of the electric vehicle industry. Over 70% of the world's cobalt comes from the Democratic Republic of Congo (DRC), while Chinese companies control much of the refining and exports. What happens if there are disruptions in the supply chain at one or both of these points?
In this article, we demonstrate how the combination of Semantic Analysis (to map cause-effect chains from real-world reports) and scenario engineering based on System Dynamics (to simulate flows, stocks, and shocks) helps decision-makers stress-test scenarios. This is scenario planning in its truest form, using numbers, not just narratives.
Scenario planning, simulation: How fragile is the cobalt supply chain from the DR Congo to China?
TL;DR: A single disruption in the DRC mining industry or in China's refining plant has global repercussions within a few weeks. Simulation allows us to test the output after changes to key parameters as part of scenario planning. #fragile
Evidence:
The Democratic Republic of Congo supplies >70% of global cobalt production ; China controls around 50% of it.
China dominates battery refining and cell manufacturing, creating a dependency on a single player.
Historical cases (Evergrande debt crisis, COVID-19 lockdowns) show how quickly national shocks trigger global downturns.
How can Semantic Analysis translate unstructured data into supply chain models?
TL;DR: Semantic Analysis extracts cause-and-effect chains from texts and makes them machine-readable.
Evidence:
Example: “In the DRC, China controls 50% of cobalt production” → dependency chain [DRC mining] → [Chinese operators].
Cause-and-effect chain:
[DRC mining] → [Chinese refining capacity] → [EV* battery exports] → [Global EV* industry]Semantic extraction ensures that scenarios are based on real-world political and market data, rather than assumptions.
This can be monitored in real time – with feeds from media, trade data, and company announcements (monitoring).
*EV = electric vehicle (electric car)
How does System Dynamics translate narratives into simulations?
TL;DR: System Dynamics transforms text into stocks, flows, and feedback loops for scenario testing. #systemdynamics
Evidence:
Example variable:
refining_throughput_cn = MIN(china_cobalt_concentrate_inventory / time_to_refine_batch_cn, refining_capacity_cn_base * refining_disruption_multiplier)Stocks: drc_cobalt_ore_inventory , china_refined_cobalt_inventory.
Flows: shipment_rate_drc_to_cn , export_rate_cn_refined_cobalt .
Auxiliary variables: china_war_severity_index , tariff_rate_eu_on_cn_evs .
This structure enables testing “what-if ” scenarios, such as strikes, tariffs, or conflicts.
Which scenarios are most important for decision makers?
TL;DR: The two most critical shocks in our scenario planning are (1) disruptions in the DRC's mining industry and (2) disruptions in Chinese refining/export. #scenarios
Evidence:
DRC mining shock: lowers mining_rate_drc → export bottleneck within 2–3 weeks.
China refining shock: War or sanctions reduce refining_throughput_cn → Stocks are exhausted in ~60 days.
Tariff shock: EU/US tariffs depress the EV* demand index, but also distort trade routes.
Infrastructure shock: BRI disruptions extend transport times and increase costs.
Ranking by severity: China refining > DRC mining > tariffs > logistics.
*EV = electric vehicle (electric car)
What criteria should managers use to evaluate resilience options?
TL;DR: In our model, five resilience levers determine whether the supply chain system holds up or breaks. #criteria
Evidence:
Diversification – share of cobalt refined outside China.
Inventories – days of cathode and cell inventory available.
Substitution – rate of introduction of LFP*/LMFP**.
Logistical redundancy – availability of non-BRI*** transport routes.
Political coordination – tariffs, sanctions and industrial policy.
*LFP = Lithium Iron Phosphate
** LMFP = Lithium Manganese Iron Phosphate
*** BRI = Belt and Road Initiative
Should you build your own models or use external simulations?
TL;DR: Most companies or organizations benefit from external models; in-house development is only worthwhile for large OEMs or governments. #buildvsbuy
Evidence:
Build: Full control over assumptions; high cost and data load; requires modeling expertise.
Buy/Partner: Access to calibrated models; faster insights; less flexibility.
Benchmark: In-house development ~6–12 months; external partner ~4–6 weeks.
CEOs/ministries often need decision-ready results, not code.
What does a practical process for scenario planning with cobalt look like?
TL;DR: A five-phase sprint takes leaders from narrative to scenario dashboard. #rollout
Evidence:
Phase 1: Semantic Analysis from relevant content sources (covering politics, trade, media).
Phase 2: Mapping dependencies in cause-effect chains.
Phase 3: Translation into System Dynamics variables (state/flow).
Phase 4: Simulation of base and shock scenarios (DRK strike, China blockade).
Phase 5: Presentation of results and resilience options for decision-makers.
A pilot is feasible with one analyst + one modeler.
FAQ
Why did we examine the cobalt supply chain, not the lithium one? Because cobalt has fewer alternatives, and the dependence on China/DRC is higher. We also wanted to outline an example of a simulation used in scenario planning.
How quickly do shocks take effect? Stocks usually only last 1–2 months.
Can substitution solve the problem? Only partially; LFP also depends on Chinese inputs.
Isn't that speculation? No, the dependencies are based on verifiable sources and real-time data.
Who should use it? CEOs, government ministers, and investors should use Scenario Planning with Semantic Analysis.
Does this also apply to other raw materials? Yes – nickel, lithium, and rare earths follow similar patterns. We've provided only one example here to explain the method as concretely as possible and to make it tangible.
Methods & Data
Method: A combination of Semantic Analysis of relevant data sources (online, offline, TV/radio, ...) and translation into system dynamics variables.
Next step: Mini-survey of OEMs and refineries (≤10 questions) to calibrate substitution and diversification.
Table schema: date, region, mining_rate, refining_capacity, tariffs, inventories, cobalt_price.
Learn how Scenario Planning, Semantic Analysis, and System Dynamics mitigate risks.
