From raw to optimized cost of a digital solution

Embrace the digital world

From raw to optimized cost of a digital solution

0

In the textile industry, one of the challenge is to be able to cut pieces of clothes from raw material and, in this process to reduce the amount of material lost that will not be used at the end. When I was working in this industry, our company was providing a software that was optimizing the placement of each piece to be cut so that it reduces the inter-space between them (that leads to loss of material). This process was having an important impact on the total cost of the activity.

The core principle of this activity was to understand that there are many ways to place piece of clothes on the raw resources, but some of them were having more efficient results. The saving realized by those results were justifying the investment on a product that will achieve it.

When you work on your digital solution, the fact that your resources are digital should not hides the fact that it is still important to be sure that the resources that you provision will be as closed as possible to what you need. The efficiency of your deployment can be measured by the delta between what you have really provisioned and what you should have provisioned.

But what are the factors that can cause those discrepancies ?

There a many of them than can be classified as

  • Internal forces
    • Redundancy between the components of your solution
      • Redundant data storage
      • Redundant data exchanges with external SAAS (Software as a Service)
    • Redundancy for functional needs
      • Redundancy for tenant (same service deployment for each tenant)
      • Redundancy for geographical needs (same service in multiple digital regions)
    • Temporary spaces
      • Spaces for automatic tests
      • Spaces for concept exploration
      • Spaces for investigation of customer issues
    • Inappropriate usage
      • Resources purchased by not used
      • Resources not used at their maximum capacity (only a part of the day or only a part of their capacity)
  • External forces
    • Fixed granularity of resources (for example only 2 sizes of hard disk)
    • Fixed granularity of licences (for example, purchase by batch of 50)
    • Provision out of possible discounts
      • Global discount like enterprise agreement
      • Reservation discount like saving plan on AWS
      • Specific discount like engineering resources versus production resources on Azure

Overall, those forces will push you to provision more resources that what you need or pay them at a higher price than you should.

Also, as your portfolio expands, if each digital product teams tries to solve those challenges by themselves, such a fragmentation in the approach will lead to even more discrepancies between your ideal target and your real consumption.

How to solve it then ?

You need to approach this subject with an internal platform business model.

Before looking at the specific value of the platform on that matter, let’s first review the platform business model itself.

The platform pattern is one classic business model pattern that have been heavily used in the industry as one of the main support to the digitalization of activities. The platform pattern focus on creating a bridge between the producers and the consumers. It matches the providers with the potential users or clients, providing resources and a channel for communication. However, platforms are not production pipelines as they don’t own any of the resources that are creating value.

There is an important literature on that topic, but for now, the main point is to understand that the platform will help match producers and consumers with the best offer.

The platform pattern is often used for external offer and has many variants.

A closer example to our subject, is the Groupon platform that addresses the optimization of the cost by offering focused offers from providers and benefit from group discount.

What is covered ?

In digital, this part of the platform offer (there is many others that we will look in further post), relates to the domain of FINOPS.

The definition of FINOPS on the foundation website is :

FinOps is shorthand for Cloud Financial Management. It is the practice of bringing financial accountability to the variable spend model of cloud, enabling distributed teams to make business trade-offs between speed, cost, and quality.

https://www.finops.org/

In details, the activity target to work on 3 aspects

  • Inform
    • Understand fully-loaded costs
    • Benchmark performances
  • Optimize
    • Enable real-time decision making
    • Optimize usage
  • Operate
    • Optimize rates
    • Align plans to the business

The value of the platform will be to realize those goals by allowing to move from financial data to data model. Or, in other words, to allow to optimize the cost by doing future forecast instead of just observing the present and the past.

Moving from data to data model

The sad part with production data, is that they are always a measure of the past. To be able to look at the trends of the future, you need to move from data to data model which are algorithms that can create new values based on previous ones.

There is many type of data model from the more simple (average trend, compound rate) to more complex (like neuronal network and business specific algorithms). The platform goal will be to create models that will allow from business data sources to forecast the future and then guide offer fulfillment.

Why is it important ? To benefit from several providers offers, you need to be able to provide a picture of your future need and sometimes (like volume discount), this forecast will engage you in regards to the contract.

For example, a machine reservation on AWS can bring from 20% to 60% of discount but will engage you from one year to three years.

To build such a forecast, general providers (like Azure, Cisco, vmWare or AWS) can help you on 2 aspects

  1. Analyze the current use of your resources
  2. Forecast the use of resources with the historical measure

But for the second point, forecasting your need cannot be only based on the history. Without taking into account your internal needs and plans, your decisions will be as reliable as driving a car by looking at the rear view mirror. The past gives you insights based on the experience but cannot anticipate your business evolution (what is in front of you)

The platform offer

In that part of the platform offer, the producers are the resource providers (Cloud, SAAS and licences) and the consumers are the digital products

The platform value will be realized by the following activities

  1. Create the consumer profile
    1. Current footprint and consumption of resources
    2. Future deployment of the digital product to customers
    3. Evolution of the demand for the digital product
    4. Need for tenant isolation
    5. Need for performance and redundancy
  2. Create the provider profile
    1. Identify the catalog
    2. Identify the provisioning model (for example by batch of 50 with a monthly adjustment)
    3. Identify the possible discounts
  3. Aggregate the consumer profiles into a company profile
    1. To define the most optimum profile
    2. To be able to report macro-trends and consumer trends
  4. Setup macro decisions and redistribution filters
    1. Global volume contract
    2. Specific contracts
    3. Centralize justifications for discount

An example

As this can seems abstract, let’s look at the simplified example of the company Acme that have two digital products :

  • One product “Loggy” is helping employees to log their time in the day
  • One product “Monny” is doing the financial consolidation of the transactions of the company

As their products are only deployed on the East Coast of the United States, they are only running in one time zone.

Let’s also imagine that Loggy and Monny used the same type of resources which are 6 virtual machines.

Due to this, almost all resources used by Loggy are during business hours from 8am to 6pm, and almost all the resources for Monny are used outside business hours from 8pm to 4am.

It the resources for each product are not optimized, the resources will be only used :

  • 10h out of 24h on 5 days out of 7 days for Loggy => 29.7%
  • 8h out of 24h on 5 days out of 7 days for Monny => 23.8%

By analyzing the profile, the platform can also figure out that the used hours for Loggy do not intersect with the used hours of Monny.

In this case, the platform can suggest two optimizations :

  • Apply a standard auto-parking strategy for both (resources are stopped when not used)
  • Auto-deploy Loggy for 7am and destroy the resources after 6:00pm, and do the same for Monny

Selecting one or the other strategy will depends on the compliance of the application to the digital standards (like 12 factors) and other architectural considerations.

After this first optimization, the platform can also, based on the profiling performed, detect that both products demands are planned to be stable over the next year, and then propose to cover those machines by a standard reservation of 1 year, bringing an additional 20% discount

Overall, on this simple use case, the processing cost of Loggy will be decreased by 76.3% and the processing cost of Monny by 80.96%.

What is important to note is that the platform do not interfere in the willingness of the application to provision its resources. Its value is to optimize the matching of the need to minimize the cost of the fulfillment

To conclude

In a moving economy highly impacted by the digitalization of activities (32% intangibles in 1985, 68% in 1995 to 84% in 2015), the cost optimization of the digital solution is a core pillar of the success of the digital products.

The company needs to have a high focus on it to entertain the development of its digital products, but this should not refrain the digital product itself.

By applying a platform business model on the subject, the company can federate its products under an optimized budget envelop while avoiding to generate the gate keeper phenomena of a traditional pipeline.

The platform model is recognized in the industry to provide the advantages to be more efficient, benefit from the network effect (get better as user panel grow), can scale more rapidly and benefits from economies of scale

Also, as the cost optimization value is delivered, the platform can also provide with the same data, other values like internal accounting, generate standard deployment model or measure of the efficiency of the provider contracts.

© 2024 Red Phoenix AI.