There is a global push for digital transformation and taking an “innovation first” approach, but, up to 80% of these initiatives will fail. Organizations approach data science and analytics platforms with the expectation of large projects that provide large returns on investment. However, the success of small projects can have a much broader impact on company-wide technology adoption.
In this post, we’ll explore how business leaders and technology champions can increase adoption and innovation through the use of small projects. Using these small projects can not only reduce the risks associated with larger, longer projects, but also help to upskill and motivate business users, subject matter experts, or those with a less technical background.
The Digital Transformation
If you look at the news or social media, there are often articles on becoming “technology first” or adopting data science, machine learning (ML), and AI — with new AI startups or large businesses launching internal initiatives (or even acquiring the startups!).
The driver is the expectation that data science, ML, and AI will unlock large benefits for these companies. Use cases and projects have often already been identified, quantified, and used to build the business case for headcount, tools, and resources.
In 2020, Capgemini reported that more than 85% of organizations fail to scale large projects in production. There are different reasons for this and ways to combat this, but for now we’ll look at the benefits of small projects as part of a company-wide transformation.
The push for new technology comes with a highly competitive job market, and many organizations are looking internally to bridge the gap with training and enablement. However, jumping straight into data science and new technologies simultaneously can be daunting, so breaking this transition into smaller, incremental steps can make it more accessible and foster that technology culture.
Looking Beyond Large Projects
To identify and prioritize use cases, you will quite often see some sort of assessment matrix, showing measures such as risk, difficulty/complexity, and benefit. Something like this:
The idea being that you can identify low risk/cost/duration and high benefit/reward/value projects, then prioritize bottom-right, top-right, bottom-left, and hopefully ignore the top-left ones (or work on ways to move them towards the lower right). Additionally, the projects people get most excited about quite often sit in the top right corner.
Quite often this analysis is reserved for larger projects with dedicated teams and budgets. These projects are often completely new initiatives, in that they are performing a task or producing an output not currently available. As a result, there are often unknowns and roadblocks, such as: the actual benefit (it’s hard to know accuracy of a model before you train it), the availability of suitable data, access to required hardware (sensors, cameras, etc.), and the other unknown-unknowns that pop up during projects. These can contribute to longer timelines and slips in meeting deadlines.
Micro-Victories From Micro-Projects
These risk-reward matrices often ignore projects that are classified as Business as Usual (BaU) — the ones with no additional allocated budget. However, there are tangible benefits to be gained from these micro-projects as well.
In “The Progress Principle,” Teresa Amabile writes about “the power of small wins” contending that these small wins — of meaningful progress — are important for driving a sense of purpose. By taking a more gradual approach to introducing new technology, and celebrating “micro victories,” business users gain both familiarity and ownership over their work. Amabile says, “At work, people develop an increasingly strong sense of self-efficacy each time they make progress, succeed, or master a problem or task.” This strengthens overall adoption by gradually upskilling users, providing self-paced opportunities for innovation, and allowing them to build on their own successes.
I spent some time with a team that performed alert triage. This team was highly efficient, tracking time spent on activities down to seconds. In this case, a small project to automate or accelerate even a small part of this process can make a big difference. When you are a team of 30, all triaging alerts of ~3-5 minutes each, all day, every day, finding a 20-30 second time saving adds up to a 10% efficiency gain or 3 years of effort across the team!
Not every team is tasked with reviewing hundreds or thousands of alerts per week, but there are lots of ‘hidden’ repetitive tasks — either repeated by a single user or repeated by users across the organization. These might include updating data/graphs each week in a management report, filling out timesheets, converting Excel spreadsheets into tables, or even project kick-off activities. They might seem relatively insignificant when you look individually, but when aggregated across each person of each team for the entire year, they start to add up.
The Resource Roadblock
With the global interest in data science comes a large demand for skilled workers, such as experienced (or even graduate) data scientists, which are difficult (or expensive) to find. To counter this, many companies have switched focus to upskilling their existing workforce, quite often with visual tools that have a lower barrier to entry than coding.
To provide additional expertise, a Center of Excellence (CoE) for analytics is also formed, acting as internal consultants for all things data science and analytics. For an example of a CoE and the Hub and Spoke structure in practice, check out the ebook on Rabobank’s data transformation here.
Larger data science projects promote the use of data analytics for only the select few who are tasked with completing the project work — usually people already from the analytics organization. A key to wider adoption is enabling the business experts who are already working with data but need, or want, to take the extra step.
Learning to Fly Crawl
It’s easy to get excited about pushing the limit on the type of work that can be done with analytics. While topics such as text and image analysis are very popular and can be extremely beneficial, they are not always great starting points, especially for employees who are both new to data science and to the tools.
What may be beneficial is starting with tasks that the employee/team does on a daily basis and already knows well. These don’t have to be complex or even involve any data science. The aim is to get people comfortable with using technology. Recreating an existing task into a new platform or technology is a great way to start. This allows users to start with mostly familiar data and processes — they know what data goes in and what results should come out.
Think of it like the Minimum Viable Product (MVP) from an agile development perspective, which is essentially a skeleton of the final product that still delivers a usable outcome. People who have built something have a sense of accomplishment and ownership, and will want to tinker and extend it over time. This doesn’t have to be planned from the beginning, it can grow organically.
A client showed a great example of a series of incremental micro-projects for management reporting:
- The initial project was to gather alerting data and make a table for the weekly meeting instead of doing it manually.
- While showing this in a meeting, one of the business owners asked if this could be done for their sources also, so the project became reusable.
- The manager asked if this could be automated, so it just ran every morning instead of weekly.
- The support teams asked if an email could be sent with the results, so they knew as soon as possible if there were data or environment problems.
- They started looking at adding forecasting and anomaly detection, so they could understand staffing requirements and identify other issues.
- A dashboard was built to show all of this and included in the monthly senior management meeting!
Not only did it grow to affect several different teams, but it was something that they never knew they wanted!
One of our largest clients said of digital transformation, “The key is not the technology, it’s really a cultural revolution,” and companies should be mindful that introducing new technology and cultivating that mindset takes time.
A previous post talked about capturing AI value below the waterline and looked at the gains from AI when looking at smaller AI projects beyond the initial use cases. In a similar fashion, we can look at what can be gained simply from applying new technology to automate or streamline existing processes, without initially looking for AI/ML enhancement. Combining these two approaches creates even more scale and value, but also extends the impact of the technology and can shift users’ mindsets of what can be achieved later on.
Maybe true organization-wide transformation starts broad and shallow, by identifying smaller, inclusive projects alongside the larger initiatives. This allows organizations to build adoption by capitalizing on the feeling of ownership and motivation from the accomplishments, and then to move beyond with innovation and collaboration.