Table of Contents
In the previous section, we covered the basics of setting up substitution reasons in SAP for inventory planning. But given today‘s technology capabilities, there is potential to take this to the next level using artificial intelligence.
Let‘s re-explore our 25-year old substitution functionality with a modern lens to see how AI and advanced analytics can transform it.
Limitations of Legacy Manual Methods
The traditional approach has been manual definition of substitution reasons at the plant or material group level based on planner inputs. This has some inherent drawbacks:
- Does not easily adapt to dynamic market changes
- No quantitative way to evaluate substitution rules
- Prone to planner biases or oversight on potential options
- Time consuming to analyze all materials for replacements
Besides these factors, identifying the right fit between original and substitute material depends on many dimensions:
- Technical specifications like size, performance etc.
- Supplier capabilities and quality history
- Cost impact of replacement
- Past consumption trends
- Interactions with other materials (BOMs etc.)
Evaluating these interdependent variables simultaneously for the thousands of materials even in mid-size organizations becomes incredibly tricky through manual approaches. This is where AI opens up new possibilities.
AI to Discover Optimal Substitution Pairs
Instead of relying solely on human intuition, we can augment planners with AI capabilities in a few impactful ways:
1. Mine Historical Data to Identify Candidates
By analyzing past material consumption data, purchase orders and inventory movements using techniques like association rule mining, we can detect interesting correlations and substitution potential:
- Materials frequently purchased together can be potential substitutes
- Sudden drop in consumption for one, matched with uptick for another indicates possibilities
- Materials with high co-occurrence in BOMs may be viable replacements
This allows surfacing of many non-intuitive material substitution pairs that humans may miss, but algorithms can uncover by number crunching vast volumes of data.
2. Optimization Models for Substitution Selection
Once candidate materials are known, more advanced analytics can help select the optimal choice among multiple options by balancing different objectives:
- Maximize inventory availability
- Minimize procurement costs from alternative materials
- Maintain production output quality
- Limit number of production/design changes required
Multi-objective optimization techniques evaluate tradeoffs between the above goals far more systematically than manual hierarchical rules ever can. The result is substitutions tailored to business priorities.
3. Continuously Learn from each Substitution Instance
The real power of AI lies in continuously incorporating learnings back to improve future decisions. We can train predictive models to become smarter at material replacements based on outcomes of each instance:
- Gather feedback on acceptance rate for each recommendation
- Assess production quality impact when substituted
- Analyze resultant inventory costs post replacement
- Identify correlated effects on other materials connected in BOM/routings
This allows our substitution recommendations to constantly get more contextual, relevant and optimized.
Industry Use Cases Demonstrating Value
While the techniques described above represent the future, there are already some real precedents of leveraging AI for material substitutions and inventory optimization creating significant value:
Electronics Manufacturer
- AI based application provided dynamic substitutions for 20k component types
- Increased availability by 19% and reduced costs by 12% over 8 months through data driven recommendations
Automotive Supplier
- ML model identified substitute polymers with properties aligned to design needs
- Rapid material replacement helped account for market supply volatility
Industrial Distributor
- Product recommendation algorithms identified ~18% more substitute product groups
- Conversational interfaces like chatbots explained substitutions better to customers
The key takeaway here is AI and analytics can make a dated functionality like substitution reasons smarter, faster and simpler – delivering manifold efficiency gains in the modern context.
Let‘s wrap up with some final thoughts…
Concluding Thoughts
Legacy capabilities like substitution reasons have largely remained static whereas technology innovations in AI/ML have continued rapidly advancing in parallel. This creates massive potential to reimagine such solutions with present day tech capabilities.
As the examples demonstrate, combinations of powerful algorithms and availability of extensive data now allows us to completely transform the scale, intelligence and automation around substitutions. We can displace intuitive guesses with statistically backed confidence measures for each material replacement recommended. We can run complex simulations to evaluate business tradeoffs rather than using simple rules alone.
Hence it is an exciting time for supply chain practitioners to relook at parts of existing processes with a fresh perspective to drive step function improvements. The possibilities to push the envelope further by combining institutional knowledge with cutting edge analytics are endless.
What other established capabilities do you think are ripe for disruption with modern AI? Share your ideas or feedback on other such opportunities you see. Together we can craft the future.