What Has Data Delve Been Up to in 2022?

With consulting work, it is easy to just keep your head down and work on client projects. This post is a lifting of my head for a 2022 review. At top level, I help people build models and gain data insights to grow their business. This means defining problems in a rich and actionable way and then turning them into coded models and visualizations to uncover hidden gems for decisions and growth. Sometimes these solutions are prototypes that lead to more permanent solutions. Sometimes they are themselves the answer.

Here is a bullet point listing of what we were up to in 2022. A paragraph on each below gives more details:

  • Business and manufacturing simulation for a chemical industry client
  • Python-coded data solutions for an e-Commerce appliance product that is online-connected through AWS
  • Troubleshooting the interplay of a client’s raw materials and manufacturing process to resolve quality problems
  • A foray into online training with launch of our Digital Transformation with Excel course on mastering Microsoft Excel in organizations
  • Building a “control panel” front-end user interface for a complex, process and formula simulation model for a CPG product
  • Notes on the the Global SciPy conference I attended this Summer in Austin
  • Overview of our software toolbox

A current project is helping a client’s $US 150+ MM sales B2B chemical business. I work with their business and operations leadership to build a business forecasting/simulation model. They know there are nuggets there that can help but want to translate their intuition into simulations guiding what to emphasize in the business and operation. The business simulator allows them to vary product mix and manufacturing capacity and look at how to grow sales and profit. That links to a capacity and cost model for their two plants. The latter predicts production volume and cost for their batch-making operation.

Work with a long-term client uses Python libraries I wrote and continue developing to connect their AWS, eCommerce and fulfillment vendor data from an internet-connected appliance they sell online. The data foundation is then a basis to explore and answer questions with creative visualizations and models. This helps the team understand the in-market dynamics and consumer experiences with the product. From “wish for” learning questions they provide, we define requirements and then develop code for both exploring those questions and keeping ongoing track of consumer usage of the fleet of in-market devices.

For another current client, I am working with their engineering and manufacturing team to troubleshoot their combined raw material supply and chemically-based manufacturing process. We seek to eliminate intermittent off-quality product that causes expensive customer issues. We use my 5-step, complex troubleshooting methodology as a roadmap to collaboratively define the problem, engage a broad range of expertise, brainstorm a learning plan and execute experiments to create robust solutions.

On the online training front, I’m partnered with Toronto-based Predictum Consulting LLC and its CEO, Wayne Levin. In October 2022, we launched an online course called Digital Transformation with Excel or DTE. It brings together training and mentoring insights from P&G days and recent consulting projects. DTE recognizes the pervasiveness and unabashed usefulness of Microsoft Excel in companies. It teaches mastery in context of how to structure and curate data and models. This includes how to build great user interfaces and how to make data portable and extensible for collaboration.

A recently-completed, two-year project involved building an extensive Excel/VBA “control panel” to feed a simulation model for a CPG brand’s batch-made product. This application lets users import or enter the product formulation and process setup as inputs. It then generates JSON input decks to feed chemical process simulation software. The VBA work involved envisioning a user interface and behind-the-scenes tracking/wayfinding for this work process –making it possible for users to experiment with varying the formula and process in ways that honor plant and ingredient constraints. Because of the complexity of the model, I wrote and utilized an open-source VBA validation suite to continuously run dozens of validation cases. This helps ensure that VBA code changes do not break previous work.

While not a project, I attended the Global SciPy conference this Summer in Austin. I learned a lot and made numerous contacts for the future –despite coming home with a mild case of Covid. This community ranges from astronomers to nuclear physics experts, so it is an extreme intellectual challenge for this chemical engineer. I took tutorials to build my expertise in working with geospatial and map data in Python. I networked with a lot of interesting folks and formed some lasting collaborations. In addition to the scientific and upstream R&D industry worlds, SciPy has significant attendance by the financial analysis community. However, it is almost absent any downstream industrial viewpoints such as mine. This is perhaps understandable because this is not SciPy’s focus, but this also speaks to the state of Python tools for experimental design and quality control. In my opinion, this also causes SciPy to have a siloed focus on Python tools wielded by upstream corporate people and by academics –even when Python is not the optimal tool for the job.

Finally, my software toolbox continues to expand and is pragmatic. It uses a “right tool for the job” mentality. To start, I develop coded solutions in Python and its Pandas, Numpy and visualization libraries. This Github repository (see tutorial.md there) is a code sample and discusses the under-publicized synergy between two common Python practices, TDD and OOP (aka Test Driven code Design and Object Oriented Programming or “Tuhdoop” as I call the combo). I use the Python Scikit-learn library for advanced modeling. In addition to Python, I use JMP Statistical Discovery software (by SAS Institute) for exploratory data analysis, modeling, multivariate analysis, quality analysis, stats and visualization. I use Microsoft Excel/VBA to build simulation models that can be used by non-coding clients. As discussed above, Excel is also useful for building user interfaces for models that link to other applications. I use PowerBI to create and publish dashboards for clients.

Here’s to an interesting 2023 of helping clients and continuing to grow in bringing solutions!