hyunjimoon
commented
3 months ago business case studies + hypothesis that can requires inference, that has similar inference based sequential decision making structure.
This is code that has decision and inference
-
IT (AI-NoAI X B2B-B2C): An IT startup is considering whether to develop an AI-powered product or a non-AI product, and whether to target the B2B or B2C market. The company runs experiments by developing prototypes and testing them in different market segments. They analyze data on user adoption, engagement, and revenue to determine if they should scale the current product-market combination, pivot to a different product (AI to non-AI or vice versa), pivot to a different market (B2B to B2C or vice versa), or fail if the results are consistently poor.
-
MedTech Operation (Outsource-Inhouse, Global-Local): A MedTech startup is developing a medical imaging device and must decide on their manufacturing strategy. They can choose between outsourcing or in-house production, and between global or local manufacturing. Each option has different costs, production speeds, and quality outcomes. The company runs prototypes through each combination, measuring product performance (e.g., image fidelity and processing speed) against FDA requirements. They must balance the trade-offs between cost, time, and quality to achieve FDA approval before running out of funds.
-
Gene Therapeutics (Old vs New Base Editor X Dravet Syndrome vs SETBP1): A biotech startup is developing gene therapies using base editing technology. They must choose between an older, proven base editor and a newer, potentially more effective but less tested version. They're targeting two genetic conditions: Dravet syndrome and SETBP1 disorder. The company conducts clinical trials, analyzing efficacy data (e.g., percentage of cells successfully edited) and safety profiles. They must decide whether to continue with the current editor-disease combination, pivot to a different editor or disease target, or scale up based on promising results, all while considering manufacturing costs and market size for each condition.
(4. Quantum Computing for Drug Discovery)
three hypothesis: H1: Signal Strength Hypothesis As the signal strength (clarity and reliability of information) increases in one dimension of innovation (e.g., product, technology, or market), the ratio of pivots in that dimension relative to other dimensions will increase. This is due to innovators having more confidence in making changes based on clearer signals. H2: Uncertainty Reduction Hypothesis As the overall uncertainty (σ) in the innovation process decreases, the ratio of pivots will initially increase across all dimensions as more reliable information becomes available. However, this ratio will eventually stabilize or decrease as the innovation landscape becomes more predictable. H3: Resource Accessibility Hypothesis The ratio of pivots in different dimensions (e.g., product vs. market) will be influenced by an organization's access to relevant resources. Organizations with greater access to resources in a particular dimension will show a higher pivot ratio in that dimension compared to organizations with limited resources.
| Aspect | IT | MedTech | Gene Therapeutics |
|---|---|---|---|
| Upstream (Product) Choice | AI vs Non-AI | Outsource vs In-house | Old vs New Base Editor |
| Downstream (Market) Choice | B2B vs B2C | Global vs Local | Dravet Syndrome vs SETBP1 |
| Combinations | AI-B2B, AI-B2C, Non-AI-B2B, Non-AI-B2C | Outsource-Global, Outsource-Local, In-house-Global, In-house-Local | Old-Dravet, Old-SETBP1, New-Dravet, New-SETBP1 |
| Predict | Estimate user adoption, engagement, and revenue for each combination | Project product performance, cost, and time-to-market for each manufacturing strategy | Forecast efficacy, safety, and market potential for each editor-disease pair |
| Decide | Choose to scale, pivot product, pivot market, or fail based on data analysis | Select manufacturing strategy balancing FDA approval likelihood and fund availability | Opt to continue current path, switch editor/disease, or scale based on trial results |
| Infer | Update beliefs about product-market fit and adjust strategy for next iteration | Refine understanding of manufacturing trade-offs and adjust approach | Revise estimates of editor efficacy and disease treatability for future decisions |
| H1: Signal Strength Hypothesis | As AI model performance improves, product pivots (AI architecture changes) increase relative to market pivots | As manufacturing quality metrics improve, production strategy pivots increase relative to market pivots | As base editing efficiency increases, editor type pivots increase relative to disease target pivots |
| H2: Uncertainty Reduction Hypothesis | As AI development tools mature, pivot ratio increases initially but stabilizes as AI becomes more predictable | As manufacturing processes standardize, pivot ratio increases initially but stabilizes as outcomes become more predictable | As gene editing outcomes become more predictable, pivot ratio increases initially but stabilizes as the field matures |
| H3: Resource Accessibility Hypothesis | Companies with better AI talent and computing resources show higher product-to-market pivot ratios | Organizations with advanced manufacturing capabilities show higher production-to-market pivot ratios | Companies with better gene editing tools and expertise show higher editor-to-disease pivot ratios |
summarize https://github.com/Data4DM/BayesSD/discussions/159 for introduction part of the paper. this paper will describe the last row and explain observations using simulation models