ENGINEERING CASE 05 / Analytics

Supply chain graph: identify disruption risks across orders

Link a component to an order date. See the addiction. Test alternative before replacement.

Modelled engineering case. Developed by ITTECHE to analyze architecture and solutions. Company, educational data and calculations; the confirmed result of client implementation is not declared.

TASK

Build a traceable chain from delivery to order and recalculate deadlines when input data changes. An alternative is considered only if compatibility is confirmed and quantity is available.

Knowledge GraphGraph AnalyticsERP / MRP
Go to demo
The two suppliers are linked through components to assembly lines and shipping; the break in one delivery is marked with a brass marker.
Conceptual illustration of the solution.
01 / CONTEXT

From the business task
to a verifiable solution.

The model manufacturer assembles two types of modules. Controller A is included in both products, drive B is included only in the first. Delivery times, product composition and promises to customers are kept separately. The delay of the common controller affects two orders at once, although each purchasing document appears to be a local problem.

customer orders in example
2
required components
2
common dependency of two assemblies
1
single point of time reference
D0
  1. 01

    Reconcile IDs

    Vendor, position, BOM version and order receive persistent keys. Similar names are not automatically combined.

  2. 02

    Build dependencies

    The edges of the graph store the type of connection, quantity, duration and source. The common component becomes visible on the path to each dependent order.

  3. 03

    Check whether the deadline is achievable

    For mandatory components, the latest date of availability is taken. Then the build duration is added. The model uses a single day scale from D0.

  4. 04

    Check action option

    Replacement requires technologist approval, quantity and confirmed delivery. The system explains the change in deadline; the decision is left to the planner.

02 / INTERACTIVE DISCUSSION

Which orders will be affected by the delay?

The training graph relates two components, two assemblies, and two orders. D0—conditional settlement date; all deadlines are specified in calendar days from it. Change the delay and alternative delivery terms to see which orders are affected.

The main supplier of controller A is used.

Component, Assembly, and Order DependenciesController A is needed for both assemblies. Drive B is only needed by module M. The timing and influence paths are also given in the text table.Controller AD11Drive BD3Module M 2 daysD13Module N 3 daysD14K-101 · term D8D13K-102 · term D9D14
An arrow indicates a required dependency. Crossing lines do not form a node. Colour highlights delays; exact statuses are available in the table.
Result of recalculation of the same graph
OrderReadinessPromised dateDeviation
K-101D13D8+5 days
K-102D14D9+5 days
  • Controller A + drive B → module M (2 days) → K-101
  • Controller A → module N (3 days) → K-102
Initial data and calculation rule

Main controller A: 2 parts to D4; when delayed - to D11. Drive B: 1 part for D3. Alternative A: 2 parts for D6; if there is insufficient quantity - 1. Each order needs one A, and K-101 additionally needs a B. Substitution is only applicable if accepted, sufficient quantity and earlier date; if the main date is unknown, a confirmed alternative can provide an estimate.

Readiness M = max(A, B) + 2; readiness N = A + 3. K-101 is ready after M, K-102 - after N. If the mandatory date is unknown, the result is also unknown. A cycle and a missing link are considered a graph error. Machine queues, transportation and work calendars are not modeled.

03 / ARCHITECTURE

From measurement -
to an explicable action.

  1. 01

    ERP / MRP

    Product compositions, balances, deliveries and customer deadlines with download versions.

  2. 02

    Connection graph

    Entity keys, edge types, composition versions, and origins of each fact.

  3. 03

    Scenario calculation

    Propagate dates through directed dependencies and check alternatives.

  4. 04

    Scheduler

    Path of influence, time frame before and after change, unknown data and employee decision.

Infrastructure, storage and integration periods are clarified during the survey. The event's identifier, time, source data, model version, and decision are saved.

04 / ENGINEERING SOLUTIONS

The difficulty lies in the conditions
errors and verification.

01

All required personnel required

The shortest path from the supplier to the order does not indicate the readiness of the product: the assembly may require several parts at the same time.

Solution
For mandatory components, use the maximum availability dates and operation duration. Replaceable branches should be modeled separately from the required ones.
How to check
The common component is delaying two orders; the absence of a drive only affects the assembly that uses it. The result contains a chain of dependencies.
02

Alternative with conditions

An early deadline from another supplier does not mean that the part is authorized and available in the required quantity.

Solution
Check compatibility, tolerance, quantity and confirmation date. A rejected alternative remains visible with a reason for rejection.
How to check
Denial of admission, insufficient quantity and late confirmation. In these cases, the promised period is not improved.
03

Unknown Date - Single State

It’s easy to turn an empty term into zero and mistakenly receive an order without risk. The cycle in the composition is capable of stopping the calculation.

Solution
Do not substitute a fictitious date; show missing ratings and affected orders. Check cycles and composition versions before publishing the result.
How to check
Removal of delivery date, loop in test composition and outdated snapshot. Completeness control must operate before a recommendation is issued.
05 / CHECK AND PILOT

What needs to be confirmed
on real data.

Reconciliation of connections

The planner checks the compounds and influence paths on the test set. The share of connections with the source and version, as well as errors in combining items, are measured.

History without knowledge of the future

Scenarios are reconstructed from snapshots available on the calculation date. Subsequent actual deliveries are not fed into the original forecast.

Result for the employee

The detected affected orders, the correctness of the calculation of deadlines and the time for analyzing the cause are checked. The economic effect is consistent with the pilot data rather than derived from the training graph.

Path from survey to transmission

  1. 01

    Data and semantics

    Identifiers, composition versions, quantity units and definition of the promised period.

  2. 02

    First Count

    Limited nomenclature, origin of connections and reconciliation with the planner.

  3. 03

    Scenarios and exceptions

    Common Component Delay, Rejected Replacement, Unknown Date, and Cycle Check.

  4. 04

    Working pilot

    Versioning of images, monitoring completeness and transmitting explanations to employees.

This is a sequence of work, not a promise of a deadline or budget. The assessment comes after checking the source data and integrations.

06 / BORDERS

Where a solution is required
a different approach.

  • The training calculation does not take into account equipment queues, work calendar, transport and partial shipments.
  • The same item name does not prove interchangeability. The solution requires confirmed characteristics and approval.
  • The highlighted chain shows the dependence of the deadlines, and not the probability of failure. An unknown period does not translate into a 0% risk estimate.
FROM SCRIPT TO YOUR OBJECT

Let's start with the data
and acceptance criteria.

For the first discussion the following are useful:

  • Version compositions of products and rules for alternatives.
  • Remains and confirmed deliveries with quantity and date.
  • Customer orders, promised terms and duration of operations.
  • Responsible planner and control scenarios for reconciliation.
Discuss a similar project View the direction of work ↗All cases ↗