Bayesian Entrepreneurship: 🗺️map for startup compass🧭 #159
Replies: 29 comments 15 replies
-
|
2025-05-08 These aren’t just colorful metaphors – they echo real situations observed across diverse innovation ecosystems. In discussions with > 50 ecosystem stakeholders 1) entrepreneurs with different vision and capability, 2) venture capitalists with varying degrees of specializations (generalists, software/tough tech focused) and funding sources (creation/foundries, corporate, government), 3) educating program designers from universities across the globe (SNU (Korea), MIT (U.S.), NUS (Singapore), HPI (Germany)) and educators with diverse teaching styles (case-based, theory-based, learning by doing), 4) opinionated mentors on success factors (execution vs idea, flexibility vs efficiency, ownership control vs growth speed, listen to doubters vs one's gut) and strategy (intellectual property, value chain, disruptor, architectural). I found case after case of guidance failing when it ignored context. For instance, MIT’s Technology Licensing Office often urges patenting as a default strategy – a great north star in biotech, but a false lead for climate-tech ventures that thrive on open collaboration. Similarly, the now-canonical customer-centered playbook works brilliantly for consumer apps, yet I saw it misfire in domains like advanced cancer diagnostics where demand is assured and technical validation is the real hurdle. I heard of founders from wealthy families who skipped the frantic early fundraising that standard advice assumes – with personal capital or patient backers, their priorities lay in product and partnership, not pitching every quarter. Even team-building norms varied wildly: Silicon Valley’s “hire fast, fire fast” mantra falls flat in East Asian contexts that prize loyalty and gradual growth. And within the U.S., East Coast startup hubs often follow a tight-knit “village” model of mentorship, unlike the West Coast “valley” ethos of blitzscaling. Finally, when the macroeconomic climate shifts into a downturn, yesterday’s winning playbook can become today’s wrong map – underscoring that any guidance must adapt to new terrain or risk leading entrepreneurs off course. |
Beta Was this translation helpful? Give feedback.
-
|
in the process of translating propositional knowledge (in black) to prescriptive format (in orange) as a decision tree. this will be embedded to LLM for i1,2 described in p.2 of moon23_nss_gps_ai_pd_demo.pdf 
|
Beta Was this translation helpful? Give feedback.
-
VandenSteen16_formalstrat.pdf |
Beta Was this translation helpful? Give feedback.
-
64, 2, 1 >> 2, 1Regarding implementing the entrepreneurial compass idea, I think this is connected to our justification of automatization and/or starting from the joint space then observe emergence. I wish to suggest auto_test 64, manual_test 2, choose 1 with computation & chatbot’s help in auto_test, outperforms choose 2 test 1 (c2t1). 64 is from 4 scenario ([compete, collab] x [control, execute] at each NSS). |
Beta Was this translation helpful? Give feedback.
-
|
09.12. Jinhua, Angie
generated from https://chat.openai.com/share/336d4f49-40a4-47bf-9683-39d6d462556b
|
Beta Was this translation helpful? Give feedback.
-
|
09.15 Scott, Angie 
3steps:
1. elicit2. localize and pivot
3. decide and branch
|
Beta Was this translation helpful? Give feedback.
-
|
09.15 Charlie Angie
|
Beta Was this translation helpful? Give feedback.
-
combination of strategic design and operation of startup along its lifecycle
request to send the excited evaluation and construction design interesting (curious and learning something - tease) evaluate (information to help you improve to assess performance, rewarded by customers, ) as you professionalize, you reacculturate, question, 
|
Beta Was this translation helpful? Give feedback.
-
|
GPT experiments
|
Beta Was this translation helpful? Give feedback.
-
|
First Final draft based on https://chat.openai.com/share/a67aef02-7043-482e-99f7-286064507cc4 Bayesian Workflow and Startup Operations1. Analogy between Bayesian workflow and Startup operationsThe Bayesian workflow is a principled approach to statistical modeling and decision-making, consisting of three main steps: model building, inference, and model checking/improvement. The Bayesian community has evolved from focusing on Bayesian inference to embracing a more comprehensive Bayesian workflow (Gelman et al, 2022). This shift parallels the transition in entrepreneurship from focusing on entrepreneurial compass, which suggests directions, to a more operational approach that provides the tools to move towards certain direction. This paper aims to draw an analogy between Bayesian workflows and startup operations, exploring how quick, easy, reliable model growth enabled by Bayesian workflow can be applied for entrepreneurial growth. Key analogy is, just as computational representation of mathematical model evolves to pair itself with observed data in Bayesian workflow, organizational representation of business model adapt itself with data from demand. This pair is also known as market-product or need-solution pair. Co-evolution of pair has been raised in Von Hippel and Von Krogh (2016). 2: Bayesian Workflow Components and ModelsThe Bayesian workflow is organized around three main components: P (joint probability distribution), A (posterior approximator), and D (observed data). P represents the joint distribution over observable quantities 3: Startup Operations ComponentsIn the context of startups, P, A, and D components can be mapped to business model, scaling tool, and observed market data, respectively. P represents business model, starting from long term planning on orientation and investment to short term control on customer, technology, competition, product, organization. A represents scaling tools that can implement business models. D would be real-world market data, such as customer behavior and sales figures. These components can also be combined into different models in startup operations: P, PD, PA, and PAD. P Models could represent different business strategies. PD Models would combine the business model with real-world market data. PA Models would combine the business model with analytical tools. PAD Models would integrate the business model, scaling tools, and real-world market data. PAD is the ideal destination for a scaling startup but different evolutionary paths exist, so strategies could be dynamically adjusted along one's lifecycle.
P - Business Model
A - Scaling Tools
D - Observed Market Data
4: Growth Path to PAD model in Startup OperationsTo scale well, three components (business model P, scaling tools A, and real-world data D) should be harmoniously orchestrated. Starting from one nailed component (among P, A, D), different growth paths exists to reach PAD. Principled guideline on when to focus on which component could enable quick, easy, reliable growth toward PAD. 
Figure 2: Growth Path in Startup Operations
5: Future Work and Research QuestionsEx-Ante
Ex-Post
By exploring these research questions, we can further validate and refine the analogy between Bayesian workflows and startup operations, providing a robust framework for both statistical modeling and entrepreneurial decision-making. References
AppendixFormal definition and examples of component and model in Bayesian Workflow for further analogy. Refer to sections from Burkner et al (2022) for detail. P component -
|
Beta Was this translation helpful? Give feedback.
-
|
From: Bayesian Workflow and Startup Operations: Paragraph 2 says "P represents the joint distribution over observable quantities y and latent parameters θ. Examples include Gaussian Mixture Models and Bayesian linear regression." Paragraph 3 says: "P represents business model, starting from long term planning on orientation and investment to short term control on customer, technology, competition, product, organization." It is not clear to me how to connect or map the second sentence onto the first , or vice versa. Could you provide a concrete example? You do expand on the second sentence in the text that follows the first diagram, and you expand on the probability model in the appendix, but I don't see how to map them together. It might be useful to share the simplest version of the Gelman model with an a=illustration that Gelman might use, and then show the parallels and mapping as to how ot do this for the PAD model you propose for entrepreneurship. As one might ask ChatGPT: Explain it to me as if I were a high school student :-) |
Beta Was this translation helpful? Give feedback.
-
|
Example for updating P, A, D and its typified path in https://discourse.mc-stan.org/t/bayesian-workflow-gps-to-navigation/32864 (@tomfid might also be interested). Evidence based testing i.e. adjusting the standard for Mammography (recommended to age > 45 to age > 35 woman, give caution sign to > 2micro gram of ? to >1 micro gram (numbers may be inaccurate) @okaranfil explained in #156 (reply in thread) is example of adjusted "good enough diagnostics" that affects further action. Different standard for diagnostic are simulated and tested beforehand, among which one chosen standard is presented to society as a form of a prior policy. I'm lacking knowledge on what the exact analogue for prior predictive, simulation-based calibration, posterior predictive check diagnostics would be in the above evidence-based testing context, but it would be useful to lay out this mapping. |
Beta Was this translation helpful? Give feedback.
-
|
@hyunjimoon's GPT’s analysis on ministry of supply: “few thoughts on how I might use the PAD framework and Bayesian workflow navigation ideas to make decisions after receiving bad diagnostics for the Ministry of Supply case: P (business model feature + strategy) - The business model and strategy for Ministry of Supply seems quite strong overall. Their focus on technology-enabled high-performance professional apparel is a compelling value proposition. However, some aspects of their business model may need refinement, such as their approach to managing inventory and supply chain. A (scaling tools) - Ministry of Supply relies heavily on analytics tools and partners for inventory forecasting, supply chain management, and fulfillment. These tools and partnerships are critical for scaling the business. However, there may be opportunities to improve how they are leveraged. For example, enhancing information sharing and coordination with partners. D (observed data) - The observed bad diagnostics, such as high inventory variability, stockouts, and cash flow issues, provide critical signals about problems in the system. Digging deeper into the drivers behind this variability could uncover useful insights. After receiving bad diagnostics, I would first try to update the analytics tools and processes (A) to improve forecasting, inventory management, and coordination with partners. If issues persist, I would look at refining the business model (P) by reconsidering sourcing strategies, supplier relationships, and inventory policies. I would also gather more fine-grained market data (D) to better understand demand patterns and variability. An iterative process of updating A, P, and D could help navigate towards a model that yields improved diagnostics. Frequent checkpoint assessments of key metrics would guide next steps. Collaborating across functions and partners would be critical throughout this process.” Prompt was: PAD_componentsP, A, and D components can be mapped to business model, scaling tool, and observed market data, respectively. P represents business model, starting from long term planning on orientation and investment to short term control on customer, technology, competition, product, organization. A represents scaling tools that can implement business models. D would be real-world market data, such as customer behavior and sales figures. These components can also be combined into different models in startup operations: P, PD, PA, and PAD. P Models could represent different business strategies. PD Models would combine the business model with real-world market data. PA Models would combine the business model with analytical tools. PAD Models would integrate the business model, scaling tools, and real-world market data. PAD is the ideal destination for a scaling startup but different evolutionary paths exist, so strategies could be dynamically adjusted along one's lifecycle. P - Business ModelP in startups could represent the core business model, including elements like product features, market strategy, and organizational structure. For example, a startup in the renewable energy sector might focus on solar panel efficiency, government subsidies, and partnerships with construction firms. A health tech startup's P model may include variables like patient engagement, telehealth features, and partnerships with healthcare providers. SaaS startup might have a P model that includes variables like subscription pricing, feature sets, and target customer demographics. A - Scaling ToolsA could represent tools that help refine or approximate the optimal business model. This includes data analytics platforms, customer relationship management (CRM) systems, and even decision-making frameworks like SWOT or PESTLE analysis. D - Observed DataD would be the real-world market data that the startup collects. This could include customer behavior, sales figures, and market trends. For instance, a fintech startup might look at user engagement metrics, transaction volumes, and regulatory changes. GPS2NavigationGPS provides the raw data about a location, navigation systems use this data to guide users to their destinations. Diagnostics on Bayesian workflow allows construction of GPS in model space. How can the typifying paths idea (below) be improved to build good navigation system that provides each modeler localized decision support? I’m curious about relevant literature (e.g. Bayesian workflow casestudies with real data & specific application) that can help me flesh out navigation idea to provide documentation complementing SBC package vignettes e.g. Small model implementation workflow • SBC 1. Looking forward to learn from any inconsistencies between our internal world models! As a modeler explores model space with Bayesian workflow map, which among joint probability p(y,θ) (P), approximator pA(y|θ) (A), data ~y (D) can he/she decide to update in which sequence based on what signal? Definition and example of P, A, D are in @paul.buerkner, @scholz, Stefan Radev’s paper 5 (which I highly recommend!) and for simplicity I divided this journey into decision node and paths. Basic idea is a modeler’s journey is a sequential decision problem where signal at each time is high dimension diagnostics and actions are updating P or A or D. This naturally led to the question: can we typify paths to reach the state of good enough diagnostics? For instance, building on Birthday problem, a time series model for numbers of birth per day, let’s say a modeler start with a simple slow trend model (P) with optimization approximation algorithm (A) with birthday per day in USA from 1969-1988. After some modeling, he/she retrieved good enough community-recommended diagnostics (prior predictive check, simulation based calibration, posterior predictive check) using a time series model with long term, seasonal, weekly, day of year, and special floating day variation (P’), Markov chain approximation algorithm (A’), birthdays per year in USA from 1969-2018. However, there exists at least 20 paths (excluding give-up at each decision) that this modeler could have gone though. 3 types of decisionQ1. prior predictive check diagnostics? bad: update P good enough: go to step 2 Q2. simulation-based calibration diagnostics? bad: update P or A good enough: go to step 3 Q3. posterior predictive check diagnostics? bad: update P or A or D good enough: finish good enough prior predictive check diagnostics (ppc1) := range of simulated observed data based on prior is not extreme: irrejectable. good enough simulation based calibration (sbc) diagnostics:= ecdf-based plots is within confidence band with alpha for every test quantities (t)) good enough posterior predictive check diagnostics (ppc2) := range of simulated based on posterior is acceptable for every quantities of interest (observed data, utility) Subjectivity in “good enough” diagnostics makes this a decision problem. Bars would be high in pharmaceutical and defense industry where human’s life is at stake. And even in the same industry, company with enough resource (computation, time, human capital) would have higher bars, and even in the same company, bar can be dragged down as project deadline approaches. >20 types of pathbad ppc1 1b_P (1): update P to retrieve good enough ppc1 diagnostics good enough ppc1, bad sbc 1g2b_P, 1g2b_A (2): update P or A to retrieve good enough sbc 1g2b_PA, 1g2b_AP (2): update either P or A gives bad sbc, update P-A or A-P to retrieve good enough sbc good enough ppc1 and sbc, bad ppc2 1g2g3b_P, 1g2g3b_A, 1g2g3g_D (3): update P or A or D to retrieve good enough ppc2 1g2g3b_PA, 1g2g3b_AP, 1g2g3b_PD, 1g2g3b_DP, 1g2g3b_AD, 1g2g3b_DA (6): bad ppc2 after updating one, retrieve good enough ppc2 after updating two 1g2g3b_PAD, 1g2g3b_PDA, 1g2g3b_ADP, 1g2g3b_APD, 1g2g3b_DAP, 1g2g3b_DPA (6): bad ppc2 even after updating two, retrieve good enough ppc2 after updating all three |
Beta Was this translation helpful? Give feedback.
-
|
Scott's gpt analysis(?) casestudy on Ministry of Supply Background: MoS's journey so far includes:
Assignment:
Interestingly, Kit initiated this analysis using ChatGPT 4.0, and the memo is available on the subsequent page. Your team's integrated analysis should build upon the ChatGPT memo, offering clear recommendations. Entrepreneurial Strategy Group HW Assignments, Fall 2023 Introduction: Key Reasons for Rapid Establishment in Men’s Professional Fashion Space:
Benefits and Costs of Brick-and-Mortar vs. Online:
Costs:
Changing Competitive Landscape:
Recommendations:
In conclusion, a judiciously scaled hybrid model can empower MoS to harness the advantages of both online and offline channels while minimizing risks. |
Beta Was this translation helpful? Give feedback.
-
1. Testing Entrepreneurial Hypothesis and Experiment
Entrepreneurs test hypotheses about value creation and capture from their ideas. However, due to high complexity and dynamics of startup operations, crystallizing hypothesis (what) and experiment (how) are challenging. Rent the Runway's founders Jennifer Hyman and Jennifer Fleiss were one of the few who tested key value creation hypotheses, in a systematic, rapid, and cost-effective manner. Critical hypothesis of their business were (1) college students and young professionals would find renting previously worn designer gowns superior to purchasing those dresses new, (2) dresses would, in most cases, be returned by customers in an acceptable condition such that after dry cleaning, they could be rented again, and (3) leading designers would not only be amenable to their dresses being rented but also would partner with a young startup to enable this new avenue. Before fully committing to their idea, the founders sought insights from renowned designers, leading them to pivot from working directly with designers to becoming a new distribution channel for them. They also conducted targeted experiments at universities, with the constructive feedback guiding their strategic direction and validating their idea's worth. With the purpose of supplying testing tools for entrepreneurs, we apply Bayesian workflow (BW) to add structure to entrepreneurs testing. BW introduces how to build statistical model from ground zero, adding inference, and checking/improvement, along with the comparison of different models, not just for the purpose of model choice or model averaging but more importantly to better understand these models. We start by defining optimality measure of experiment (1) starts from a heuristic measure, Criticality X Fidelity / Opportunity Cost, introduced in 15.911, Entrepreneurial Strategy by Scott Stern. Each components are defined as "how much does the ideal version of this test (measured accurately and at low cost) meaningfully reduce uncertainty surrounding the core value creation and capture hypotheses? (Criticality)", "Is there a (reasonable cost) test with a reasonable level of accuracy and precision? (Fidelity)", "What are the resource, time, and strategic costs of an experiment? (How much) is there a tradeoff between faster, better, and cheaper? (Opportunity cost)". Connecting this with verification and validation developed in simulation1, product management2, software development 3, scientific theory4 literature, criticality is validation test asking "are we making the right product?" whereas Fidelity is verification test asking "are we making the product right given that (or sometimes before knowing) we are making the right product?". Optimal testing is characterized with critical hypothesis, high fidelity, low opportunity. Statistical, approximation, optimization bias as three components of inductive bias from machine learning 5 literature is a useful framework to capture the tradeoff between the three, which is why we frame hypothesis criticality and fidelity as bias. Entrepreneurs take action to decrease associated cost. Table 1 introduces the definition of the costs. To test a pair of experiment and hypothesis, we can use conditional probability Table 1: Definition of Costs associated with Entrepreneurial Experiment
Table 1 lists costs associated with entrepreneurial experiment The first component of fidelity or verification is Increasing the N decreases the Table 2 presents actions to lower cost component from optimality measure of experiment Table 2: Actions to Lower Costs with Examples
These examples provide a more in-depth look into the strategic decisions Rent the Runway might have made during its early stages, showcasing the trade-offs and considerations in entrepreneurial testing. 2. Bayesian workflow to Entrepreneurial testing (tbc)SOME MAY BE WRONG
This table integrates the Bayesian workflow with the entrepreneurial testing process, drawing parallels between the two and providing examples from Rent the Runway's business model and strategies. Fig1 shows, on its left, three major diagnostics (prior predictive check, simulation-based calibration8, posterior predictive checks) in red. The right describes how hierarchical model can be used to generate data on quantities of interest. Fig1. Three diagnostics and its application to entrepreneurial testing ReferenceFootnotes
|
Beta Was this translation helpful? Give feedback.
-
|
feedbacks
literature on investment (collab vs compete) and orientation (control vs execute)
literature on cognitive aspects in or
|
Beta Was this translation helpful? Give feedback.
-
vision and T4In the context of the Mens, Manus and Machina project, my interests align most closely with T4 Topic1, 2 (learning and improving dynamic human decisions). My goal is to develop an AI mentor (nss_navigator) that helps guide entrepreneurs to build operations strategies and capabilities that support their chosen business strategies. The intent is to build on the “operations for entrepreneurs” framework of Fine, et al (2022) which proposes ten tools for scaling entrepreneurial ventures: how and when to segment, automate, processify, automate, acculturate, professionalize, collaborate, platformize, capitalize, and evaluate. Furthermore, this framework encourages entrepreneurs to continuously test the business and operations model by providing smart and scalable hypothesis testing in an experimental design framework. ultimately, the goal is to create a “smart assistant” that can emulate a mentor's insights and democratize consulting. Below, I will elaborate how this vision and choice (entrepreneurs and AI mentor) can fit into and be supported by T4. customer: entrepreneursEntrepreneurs have a critical role to play in economic development. (See, e.g., Roberts & Eesley, 2011; Eesley & Miller, 2018). Generative AI has catalyzed the blossoming of many AI startups, and many more will likely be created in the coming years. Additionally, many more entrepreneurs will be affected by AI, as a tool to help them, or sometimes as a tool that puts them out of business, if they are not sufficiently nimble and clever. A large fraction of all startup companies fails, yet there is also evidence that more experienced and mature entrepreneurs probabilistically have greater success. Work at the MIT Sloan School by several entrepreneurship scholars/educators, including Charlie Fine (my advisor) on Nail, Scale, Sail framework on operations for entrepreneurs), Scott Stern on entrepreneurial strategy, Bill Aulet on disciplined entrepreneurship and tactics, seeks to codify some of the knowledge that may help entrepreneurs improve their likelihood of success. Admittedly, the decision space for entrepreneurs is large and complex. However, we believe that there are many case examples of successful and unsuccessful entrepreneurial ventures that can support a database to help us codify the knowledge we seek. technology: AI mentorBased on (1) my practical experience of founding a startup that supplies demand forecasts and inventory optimization support for small businesses, (2) my theoretical preference of pre-asymptotic to asymptotic, instability/disequilibrium to stability/equilibrium, tools to reach normative to analysis from normative, growth diagnostics to growth theory; and (3) my interactions this summer with practitioners and scholars in entrepreneurship, I identified three areas to focus on: (1) pitfall modes describing existing problems (I felt or observed), (2a and 2 b) as-is practice and academic approaches to these problems, and (3) to-be options for direction of improvement. Our plan is to deploy generative AI to open up new avenues of research to speed the development of our codification of the knowledge most needed by entrepreneurs. 
organizationOther than ongoing communication on the above ideas with Charlie, Scott, Jinhua, and Sandy, I submitted nss_navigator to MIT ignite (generative AI entrepreneurship competition) to get feedback. Also, I'm proactively reaching out to Singapore networks to ask for support:
Reference[1] Fine, C. H., Padurean, L., & Naumov, S. (2022). Operations for entrepreneurs: Can Operations Management make a difference in entrepreneurial theory and practice?. Production and Operations Management, 31(12), 4599-4615. https://onlinelibrary.wiley.com/doi/full/10.1111/poms.13851 journey login NSS_GPS (nss_navigator's mothership) and Applying NSS_Navigator for EV architecture decisions final version: |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
|
|
Beta Was this translation helpful? Give feedback.
-
|
ORANGE: Brian, Angie
|
Beta Was this translation helpful? Give feedback.
-
|
RED: Goerge Methaos, Angie
|
Beta Was this translation helpful? Give feedback.
-
|
transcript for Paul Cheek's talk on startup operations (tactics). Highly recommend! https://otter.ai/u/iUm6V-Knazz-_bF7m4wjym9n2FE?utm_source=copy_url |
Beta Was this translation helpful? Give feedback.
-
|
Left applies inductive bias classification (approximation, statistical, optimization) to product development / entrepreneurial operation |
Beta Was this translation helpful? Give feedback.
-
|
Isabella, Angie
|
Beta Was this translation helpful? Give feedback.
-
|
invest in action with max marginal utility per expenditure: comparing ml ops and ofe higher resol: mupe_ops.pdf |
Beta Was this translation helpful? Give feedback.
-
|
I came by this slide of Andrew which lists his examples of computation (implementation) affecting statistical model (concept) i.e. hand affecting mind as we've discussed. More excitingly 😆, past Andrew seemed to support my analogy between sampling algorithms and startup operations (p.59) which Vikash also agreed so I elaborated in #174 (comment). With sampling_algorithm_theory.pdf prior and this gpt likelihood process, posterior:
This table summarizes the analogy between sampling algorithms and startup growth in the biotech and semiconductor industries, drawing parallels between the theoretical constructs of sampling algorithms and the practical steps taken by startups during their growth phases. |
Beta Was this translation helpful? Give feedback.
-
|
HBS has pitchbook data with the following columns before ipo and we could use investments as capital state Q:
Company General InfoCompanies, Company (former, aka, legal) name, Competitors, PitchBook-identifier (pbid), Description, Primary Industry Sector, Primary Industry Group, Primary Industry Code, All Industries, Verticals, Keywords, Emerging Spaces, Company Financing Status, Total Raised, Business Status, Ownership Status, Universe, Website, Employees, Employee History, Exchange, Ticker, Year Founded, Parent Company, Last Updated Date, Daily Updates, Weekly Updates Company InvestorsFinancing Status Note, Active Investors, # Active Investors, Acquirers, Former Investors, Other Investors, Active Investors Websites, Former Investors Websites, Other Investors Websites Company Contact/LocationPrimary Contact PBId, Primary Contact, Primary Contact Title, Contact, Email, Contact Phone, HQ Location, HQ Address Line, HQ City, HQ State/Province, HQ Post Code, HQ Country/Territory, HQ Phone, HQ Fax, HQ Email, HQ Global Region, HQ Global Sub Region First Financing ValuationFirst Financing Valuation, Valuation Status, [Deal Type 1,2,3], Deal Class, Debt, Status First FinancingFirst Financing Date, Size, Size Status, PatentsTotal Patent Documents, Total Patent Families, Active Patents, Pending Patents, Patents expiring in next 12 months, Inactive Patents, Top CPC Codes Last Known Valuation InfoLast Known Valuation, Date, Deal Type Last FinancingLast Financing Date, Size, Size Status, Valuation, Valuation Status, [Deal Type 1,2,3], Deal Class, Last Financing, Debt Date, Debt Size, Debt, Status Company FinancialsRevenue, Revenue Growth %, Gross Profit, Net Income, Enterprise Value, EBITDA EBIT Market Cap Net Debt Fiscal Period Company Growth RateGrowth Rate, Growth Rate Percentile Company Profile Data SourceProfile Data Source Company Size MultipleSize Multiple, Size Multiple Percentile Service ProvidersGeneral Services, Services on a Deal Company Signals Data PointsSimilarWeb Unique Visitors, Majestic Referring Domains ,Twitter Followers FilingsUCC Filings #, Last UCC Filing Date, Last UCC Expiration Date |
Beta Was this translation helpful? Give feedback.
-
|
Beta Was this translation helpful? Give feedback.
-
hyperparameter(trials can apply to investors, periods can apply to long vs short term vision (exit is more usual in u.s., n-samples is number of customers to interview for startups as theorized above) |
Beta Was this translation helpful? Give feedback.
-
|
charlie <> angie 
this framework is in development thanks to input from Bernard (changi airport) and Linus (green ant) and many others |
Beta Was this translation helpful? Give feedback.
-
|
Q1. what is the highest uncertainty?
|
Beta Was this translation helpful? Give feedback.
-
|
Three chat snippet with Kris Ferreira on operations to nail and scale visionary research empirical TOM seminar under NSS lens (Jan.22.2024)nail stage
scale stage
sail stage
different structure of solo vs collaborator/empirical research (Feb.16.2024)
identifying core capability (Feb.16.2024)
|
Beta Was this translation helpful? Give feedback.
-
|
I love the way you are framing our research progress in the NSS way from a startup development perspective! Even tho we are still at nail stage, I feel we can iterate fast enough (fail fast too!) to lock a research question that is both valuable to us!! |
Beta Was this translation helpful? Give feedback.
-
|
complementing agent-based view of supply chain, is global view below. prob.comp is algorithm builders, angie-abdullah is tool builders, charlie is model builder, scott-jinhua is policy designer, startup practitioners/investor/educators are stakeholders |
Beta Was this translation helpful? Give feedback.
-
|
i thought matt's theory could benefit from Sterman's WIDGETS (left WIDGETS.MDL.zip) WIDGETSW (right WIDGETSW.MDL.zip) models 
in business dynamics textbook, john applied this model to other systems
ask matt
3.1 SD view: sterman-wittenberg-1999-path-dependence-competition-and-succession-in-the-dynamics-of-scientific-revolution.pdf 3.2 Bayesian entrepreneurship view: 4 Persuasion and Choice.zip from #234
i used https://claude.ai/chat/1214ceb3-5a83-41c0-b322-3901eb9cb224 tried to map with below table. very preliminary. matt's feedback on each rows would be helpful!
Key concepts from Figure 1 in the attached document:
Classification and Big Category:
to future angie: start from gpt-10-Year Academic Planning Table for Entrepreneurial Scholar and 276need-sol-bayes-ent |
Beta Was this translation helpful? Give feedback.
-
|
Had a fun chat with doug williams and christine hsieh on using orbit as distribution channel of startup biology (genome, growth, ecosystem) @chasfine and I've been developing. implemented (poc level) of seven modules (M1,2,3 are in orbit / AGENT1 in #199 (taxanomy), AGENT_ENVIRONMENT1 is #184 (casestudy database), AGENT2 in #184 (comment) (inverse reinforcement learning where we infer reward function weight from observed startup actions from AGENT_ENVIRONMENT1), ENVIRONMENT1 in #190 (industry effect), AGENT_ENVIRONMENT2 in #191 (comment) (growth); M is from Martin trust center, A is from Agent, E is from Environment) can be assembled as below. Higher resolution figure in orbitXstartup_biology.pdf 
|
Beta Was this translation helpful? Give feedback.
-
|
I wonder what form it would be if we decide to embed automation evaluation (like the table introduced in https://sloanreview.mit.edu/article/scaling-automation-two-proven-paths-to-success/ article) in orbit platform. Below is an example of the kind of scorecard used to rate opportunities for automation across multiple dimensions on a scale from 1 to 5 (low to high impact or opportunity), or with actual figures where appropriate (hard savings, hours, etc.). Leaders review scorecards to identify tasks that are good candidates for automation. Different organizations will weigh the various factors differently.
|
Beta Was this translation helpful? Give feedback.
-
|
example of industry map output shared by startup ecosystem which captures covariance of actions (AGCT) and effect of discrete events (e.g. FDA approval, GPT version update) in different industries (gene therapeutics vs Intelligence Processing Unit (IPU)) learned from hierarchical latent gaussian process planned here. very drafty |
Beta Was this translation helpful? Give feedback.
-
|
my 💠diamond model (two by two supply and demand matching) is an internal 🧭compass for startups. quantum computing x diamond modelThe key idea is that quantum computing has the potential to significantly accelerate and enhance the drug discovery process by enabling more accurate and efficient simulations of molecular systems. This could impact the pivoting decisions made by pharmaceutical companies and biotech startups. I interpreted my hypotheses R1, R3, and R4 in quantum computing for drug discovery context: R1: With quantum computing, there may be a higher signal in product pivoting (i.e., changing the molecule, strength, or dosage form) compared to market pivoting (i.e., changing the target disease or population). This is because quantum computers are particularly well-suited for simulating chemical systems and could provide more accurate insights into how changes in the molecule or formulation affect its efficacy and safety. This could lead to a higher product-to-market pivot ratio. R3: The noisy intermediate-scale quantum (NISQ) era, which is the current state of quantum computing, is characterized by noisy and error-prone qubits. In this context, σ (the noisiness of the data) is high. As quantum computers become more reliable and move towards fault-tolerant quantum computing (FTQC), σ will decrease. A higher σ in the NISQ era may lead to a higher product-to-market pivot ratio, as companies may focus more on optimizing the molecule and formulation to compensate for the noisy data. R4: Larger pharmaceutical companies with more resources may have a larger k (the cost threshold for pivoting) compared to smaller biotech startups. With the advent of quantum computing, the cost of pivoting the product (i.e., running quantum simulations to optimize the molecule and formulation) may decrease relative to the cost of pivoting the market (i.e., conducting clinical trials in a new disease area or population). This could lead to a higher pivot product-over-market ratio, especially for larger organizations. Additionally, quantum computing could impact the four key questions mentioned:
In summary, quantum computing has the potential to significantly impact the pivoting decisions and strategies in drug discovery by providing more accurate and efficient simulations of molecular systems. This could lead to a higher product-to-market pivot ratio, especially in the NISQ era and for larger pharmaceutical companies. |
Beta Was this translation helpful? Give feedback.
-
|
probabilistic (P), relational (R), social (S) reasoning applied to nine operational choices: 1.facilities 2.capacity 3.vertical integration & supplier management 4.production technologies & processes 5.workforce & management 6.information technologies 7.supply chain & materials 8.organization & incentives 9.business processes |
Beta Was this translation helpful? Give feedback.
-
|
agenda preparation for scott <> angie 1. planning based on desire, verified belief, faith
2. specific questions:q1: setting scope, how to choose audience (customer) & your be plan
q2: feedback on the need-solution
q4: ironing jumbled stopping rule, bayesian, hierarchy, dynamic modeling in sec.3 (Entrepreneurial Decision Making (M2)) - compiling ideas in #102 (comment) q5: word choice: AI vs computer - assisted |
Beta Was this translation helpful? Give feedback.
-
|
erin scott was interested in below, asking me to share saying she'd think more about it and i shared with explaining main motivation is from decomposition of inductive bias.
an machine learning example for using neural network. you'd need to
---
related to this, i typified five uncertainties from chintan's thesis to create uncertainty-opportunity-learning mapping. agentvalue create
value capture
value deliverenvironment
|
Beta Was this translation helpful? Give feedback.
-
|
money in the bank i.e. accumulated good track record that gives me credibility and negotiating power
logic on strategy of Lakatosian defense from Meehl90_appraise_amend_theory.pdf |
Beta Was this translation helpful? Give feedback.
-
|
Considering "map" is title of this thread, probabilistic program with language interface (U.S. map visual) can have good synergy with scott's startup cartography project (U.S. map visual) Enhancing the Startup Cartography Project with Probabilistic Programming cld
Executive SummaryWe propose to integrate advanced probabilistic programming techniques, specifically Generative Population Models (GPMs), into the Startup Cartography Project. This enhancement will significantly improve our ability to analyze, predict, and visualize entrepreneurial ecosystems across diverse regions and time periods. Current Limitations and Proposed Enhancements
Key Benefits
Action Steps for Development
|
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
I'm devoted to combining Scott Stern and Charlie Fine's research on entrepreneurial strategy and operation. Scott's strategic choice among disruptor, architectural, value chain, IP strategy is a compass🧭and Charlie's phase-based learning is a map 🗺️. Combined they can serve personalized navigation and moreover suggest alternate routes.
Furthermore, early stage scholars have the right to be benefited from this tool. Knowing the existence of different strategy and paths in reaching their goal would encourage them to be more visionary.
nss_gps stands for nail-scale-sail global positioning system and it leverages genAI's information retrieval capability for operation and innovation management. Charlie Fine has been developed nss from early 2000 e.g. https://operations4entrepreneurs.com/.
From below, nss_gps is framed as a tool for startups just as an example. As the need for operation and innovation management (e.g. how and when to segment, processify, acculturate, professionalize, automate, collaborate, capitalize, platformize, replicate, evaluate) is shared by both startups (e.g. Lucid, VinFast, Rivian) and established companies (e.g. Hyundai, Tesla), nss_gps would architecturally support wide range of use-case. More in blueprint below.
product description
blueprint
detail:nss_gps_arch.pdf
supporters
nss_gps is a graybox with
code for blackboxes (B1,2,3,4,5,6) exist but needs configuration, f1-4, b1,2 need implementation.
help needed:
q1. how to encode general theory and wisdom from vc and casestudies? btw, Luca Guis (TIES phd working with Scott) recommend looking into crunchbase for startup data.
roadmap
1. solution development process: supply, design, develop, launch.
1.1. supply
Introduce each individual's capability. Angie and Charlie shares experience of pivoting from an non-successful startup organization. In terms of domain, Angie and Charlie has experienced different types of business (B2B, B2C, B2G (government)). For educational organization, Charlie was initial dean of Asian school of business and Angie is manager of Bayesian computation community (StanKorea, 1.03k subscribers). We have not only expertise, but also good network with MIT TIES group (Fiona Murray, Scott Stern, Bill Autlet, Eric von Hippel), Singapore SMART innovation center (Howard Califano), Korea SNU's innovation center (Janggil Kim, Ensuk Seo).
1.2. design
NSS_GPS for entrepreneurs, which combines Charlie's NSS, TIES group's entreprenerial compass, and Angie's actionable workflow.
1.2.1. NSS [1] is a sequential three time stages of startup in its lifecycle. The nail stage involves rapidly iterating to establish a viable value proposition and business model. The scale stage involves growing the company by building capabilities and processes to serve an expanding market. Finally, the sail stage is when growth slows and the focus turns to incremental improvement and sustaining the now-mature business.
1.2.2. entreprenerial compass [2] is four strategical directions to develop ideas of each startup. Intellectual property strategy focuses on gaining control of innovations through patents and trademarks, and collaborating to reduce costs. The value chain strategy aims to be the preferred partner in a slice of an industry's value chain through strong execution and collaboration. The disruption strategy targets underserved segments and uses iteration and learning to expand. Finally, the architectural strategy creates an entirely new value chain by controlling a key resource or interface that coordinates multiple stakeholders to provide new consumer value.
1.2.3. Actionable workflow [3] is an entrepreneurial decision making architecture. Iterative testing of how 1. needs of stakeholders, 2. structural knowledge of experts, 3. machine learning methods come together is the key. It hinges upon Simulation based calibration, widely adapted testing framework in Bayesian community, which can check whether the implementation of statistical and computational model align with our expectation (prior knowledge). Based on its graphical unalignment signal, one can converge to a consistent model, more quickly and reliably.
1.3. develop (todo)
theory
NSS needs simplification or dynamic introduction to prevent cognitive overload
entreprenerial compass needs to dynamic learning component embedded
Actionable workflow needs architectural update for automating the iteration so that its product, explainable policy, can be broadly distributed
the three can be combined under Bayesian decision making [4] framework
tool: system architecture to combine the three is needed to establish database
1.4. launch
theory (mens)
1st launch (Jan.2023): 15.785 digital product management projects
2nd launch (Spring.2024): 15.785 digital product management course
tool (manus)
1st launch (Sep.2023): AI product management demo day (with Charlie Fine)
2nd launch (Oct.2023): System dynamic research seminar (with Tom Fiddaman)
3rd launch (March.2024): Bayesian Entrepreneurship conference (with Scott Stern, Charlie Fine)
2. fulfillment process: supply, produce, distribute, sell, customer.
2.1. supply
theory: scaling up experimental social, behavioral, and economic science (Abdullah Almaatouq [5], Thomas Mallone)
tool: combine LLM on probablistic programming language (PPL) framework (Vikash Mansinghka [6])
2.2. produce
theory: 1st (Aug.2023): this proposal
tool: 1st (Sep.2023): AI product management demo day (with Charlie Fine) - pivot after feedback from 1st production
2.3. distribute
theory: 1st (Fall.2023) project for Jinhua and Scott's class, combining AI and entrepreneurship, upload on arxiv
tool:1st (Sep.2023) make first version of strategy chat service available online
2.4. sell
theory: get positive response on theoretical contribution from TIES group and M3S project members
tool: get positive response on practical contribution startsups, mentors, VCs
2.5 customer
alpha: delta v's eleven cohorts (~200), 15.785 digital product management students (~100), 15S.10 students (~30)
beta: startups using online NSS_GPS service
3. need development process: supply, design, develop.
Before the theory and tools become concretely proposed, we should develop its need by introducing the usefulness of NSS, entrepreneurial compass, actionable workflow to the public.
3.1. supply
theoretical and practical discoveries as we go through solution development process and fulfillment process.
3.2. design (todo)
construct three scenarios of how the proposed tool and/or theory can sovle the problem of potential customer
3.3. develop
Linkedin post like [7] show how quality and quantity of potential customers are updated.
Reference
Technology, innovation, entrepreneurship, strategy
[1] NSS
https://operations4entrepreneurs.com/
Operations for entrepreneurs: Can Operations Management make a difference in entrepreneurial theory and practice?
[2] entrepreneurial compass
https://www.entrepreneurial-strategy.net/academic-research
Power and Prediction: The Disruptive Economics of Artificial Intelligence
[3] actionable workflow
Simulation-Based Calibration Checking for Bayesian Computation: The Choice of Test Quantities Shapes Sensitivity
seminar video at MIT probablistic programming lab
Bridging statistics and dynamic modeling with Vensim, Python, and Stan, ISDC 2023 poster
[4] Theory of lean startup
Computational social, behavioral, and economic science
[5] Almaatouq, A., Becker, J., Bernstein, M. S., Botto, R., Bradlow, E. T., Damer, E., Duckworth, A., Griffiths, T., Hartshorne J. K., Lazer D., Law, E., Liu, M., Matias, J. N., Rand, D., Salganik, M., Satlof-Bedrick, E., Schweitzer, M., Shirado, H., Suchow, J. W.; Suri, S., Tsvetkova, M., Watts, D. J., Whiting, M. E., & Yin., M. (2021). Scaling up experimental social, behavioral, and economic science. Technical Report.
[6] Sequential Monte Carlo Steering of Large Language Models using Probabilistic Program
[7] https://www.linkedin.com/posts/jeffreyrwinter_industry40-jeffwinterinsights-activity-7099416280356544513-vBoU?utm_source=share&utm_medium=member_desktop
Beta Was this translation helpful? Give feedback.
All reactions