Productionalizing Data Analytics & Tech: (1) Dynamic Architecture

This is the first part of a seven-part series on productionization and product management.?I feel it’s a very interesting topic to discuss as many of us are going through #digital and #data transformations efforts. I promise to publish all content with the next couple weeks.

(1) Dynamic Architecture

(2) Core & Common

(3) Scaling

(4) Operating Model

(5) Employee Experience

(6) Costs

(7) Realizing Value

Where do we start??If you didn’t have time to visit the Winchester ‘Mystery’ house yet, spring is a perfect time to do this. ?This house is a well-known landmark near San Jose. It was the personal residence of Sarah Winchester, widow of the gun magnate William Wirt Winchester. The house was under continuous construction for 38 years and incorporated the latest technologies, including steam and forced air heating, modern indoor toilets, plumbing, and much more. Impressive! Sarah had a passion for remodeling. “It was something to keep her busy," people close to her said. "It helped her employ people and share her wealth that way. She just never really stopped building."

I had many conversations recently about “productionalization” of data, analytics, tech, consulting, and business services. ?Many companies are inspired by well knows and successful products from Google, Meta and other digital native companies. In fact, many digital transformation efforts are working on creating “products” and “platforms”.

When we start the #digitaltransformation , we usually conduct evaluation of all data sets, tools, technology, and business value cases.?Not a surprise, we can easily identify business value cases: operational efficiency, new ways of engaging ?with customers and employees, automation of business activities, creation of new data & tech driven business models, etc.?It’s also easy to identify all tools or tech the company uses. ?It’s also possible to identify all data sets even across big globally distributed organizations. But what is difficult is identify how/if all tools work together, how they enable individual local/ global and functional business processes. Through my personal experience and conversations with many CDOs, CIOs, and functional business executives, I’ve found the tools matter less than how they are architected and deployed across business processes. You can claim the success of the ?digital transformation efforts just by connecting business outcomes and processes to data & tech and let business owners decide how to act on findings. Or you can do a lot more by “productionalizing” data analytics and technology.

If something is in “production” for #CPG or other #manufacturing company, that will result a creation of real “goods” and we all will be able to buy goods and enjoy them.

In #digitaltransformation, if something is in production it’s on the path to connecting data & tech to where the value gets created and realized.

Products start from dynamic architecture

In my early days working with IT teams, I remember we have been always talking about “getting this project, data set, or report to be moved to production”. ?In this way, we were assuming all reports in production are “productionalized”.?And we were feeling great about having as many “reports” in production as possible. If the key objective for sales team to get CRM report delivered by 8 am and print a physical copy before they visit customers, yes maybe we can claim a creation of a “product”. ?

In a context of #digitaltransformation , productionization is an architectural objective, NOT series of implementation tasks. ?The Winchester Mystery house example is perfect opposite to digital transformation objectives. Mrs. Winchester usually covered up her mistakes by just continuing to build around them. Because she had no master plan for the house, her architectural ideas didn't always work out. Since she had no deadline for completion, she'd either tear down the mistake or cover it up with something else. There were 147 builders at work on the property, there were no architects involved. This approach resulted in the construction of a state-of-the-art abode which has 65 doors to blank walls, 13 abandoned staircases and 24 skylights in floors.

Many transformation efforts are struggling with #productionalization and are experiencing like the Winchester house challenges. We all use great tech, we follow best technical architecture and constantly improve it with growing AI, ML approaches. At the best and in many cases, we don’t experience significant transformational changes yet and at the worst we observe growing expenses and pain to do the business. The Winchester house had everything, all possible and advanced components and it was technically a house. It had doors, bedrooms, bathrooms, windows, but it was very uncomfortable to live in it. ?Most people are generally happier with a very simple design. The #digitaltransformation is usually more efficient and creates value fast when it’s structured in a way that processes for leveraging data & tech don’t limit the #value the data & tech have to offer. The better we design digital transformation end-to-end, the more value we will get from any individual component. ??We will get way more value by architecting efficiency use cases together with customer experience use cases and by looking at commercial value cases and supply chain / R&D at the same time.

Having a clearly defined architectural approach for your #digitaltransformation that will support all current and likely future business objectives will help to avoid the “skylights in the floors” and “unknown number of rooms”.

What architectural approach should you take?

What you personally prefer: taking a train / bus OR taxi? Good question.?If we look at our digital transformation as a trip, intuitively we might prefer a train/bus approach. Trains are discrete events and we can plan specific places, stops we want to make, level of comfort we want, cities to visit on our way, and we know exactly when we will arrive ( assuming there are no delays). We can evaluate our train experience after we finish the trip.

However, the train “architecture” model is a less desirable for #digitaltransformation efforts. Why??In this model we are prescribing on-train activities, stops, timing?and don’t pay attention to what is happening before and after the trip.?We are “automating” the train trip for our business users?and we are leaving them manual pre-trip and after the trip work. ?Lets say we are delivering a connected omni-experience platform as a part of our digital transformation.?We can automate all MarTech, channel and content integrations, automate data delivery.?But we are not automating our pre-trip activities. We still manually start from asking our business users what they want to accomplish and how to prioritize various choices. After we deliver CRM data integrations, we deliver reports to sales or marketing to possibly motivate them to make decisions and come back to us to do another project.?In the case of train-architecture, the digital transformation is viewed as People => Data & Tech => People.?So, we really miss many automation opportunities with “people” parts. ?If we continue considering the digital transformation as a series of “trains”, ?we will be missing up to 70% of the transformation efforts, we will be leaving money & opportunities on the table.?Also, lets don’t forget about people introduced errors.?If we don’t see good results from omni-targeting, this could mean our data models don’t work correctly and we selected wrong channels?or bad content. But this could also mean that we scoped our efforts wrong, or salespeople refuse taking advantages from data analytics.

So, the taxi architecture approach for digital transformation in my opinion is preferable.?We still need to decide on a direction of our destination. But we have an ability to do unpredicted stops, ride back to pick up passengers we missed, pick up new passengers we didn’t plan for, or we can even decide to take a different road because we get new information. Every new trip starts where we finished, and we don’t need to ride back to the train station to board a scheduled train.

Productionalization architecture for data analytics & tech is a circle with people at the center and data & tech products around. Productionalization is less about people, it’s more about operating and adjusting data &tech.?We don’t measure our success by sending another report. We measure success by where we are in a comparison to initial parameters ( our location on a map during the trip). ?Our initial parameters should be viewed as opportunities to modify directions and not as very specific requirements to be meet.

If our digital transformation will focus on automating every part of our data cycle including collection of requirements and getting data directly into actions & decisions, we will create more opportunities for our people to spend time with the data and enjoy results of data driven actions. If we use less people time to do manual pre-work for requirements, our people will have more time to monitor results.

Let me give a very specific example.?CDOs will understand me and appreciate. ?Data mastering and data integration tasks take up to 90% of the work, many people days and years for requirements, manual data masters and we still end up with non-dynamic #datadelivery. The data architecture gets outdated sooner than we can enjoy results. ?There is no point to continue such manual efforts if we can use ML & AI automation to master the data. ?We can automate masters with machines, without creating people driven requirements.

Conclusion: the major challenge for #digitaltransformation efforts is not to choose specific tech, tools and harmonize data.?The major challenge and benefit are to architect our digital transformation in a way that you can deliver automated, continuous revision of the business value solutions. ?This will help to constantly discover missed opportunities with original assumptions as well. ??Business value solutions are more important than tools or tech. ?Not every #digitaltransformation needs to #productionalize all data analytics & tech. But as more we can productionalize as better outcomes we will deliver with less risks at less costs. Decisions to digitally transform the business are usually ?articulated in terms of business goals and what we want to accomplish as the business. Data analytics and tech are viewed as enablers ?of the transformation. However, if we drop fantastic data & analytics “products” into an inefficient, non-automated environment, the data and tech won’t do anything but let us look sexy as we increase costs and fail. The major determinant of success is to architect #digitaltransformation in a way when we make data analytics & tech a part of our business. This will be a real productionalization.