The Double Diamond for Data Analytics: The Discovery Space
New life events usually lead to reflection and my moment of reflection now is a change of jobs. During my leaving dinner, my team jokingly put together a goodbye card in which they nicknamed me ‘Mr Double Diamond’. Jokes aside, in this post, I want to take you along the journey of establishing a ‘new’ way of working for data analytics based on this double diamond. Not only is this way of working user-centred and flexible enough to embed many different disciplines, but it also establishes data analytics as a serious workstream in many a business track. Before moving on though, let me first guide you through the pain points that I experienced that kickstarted this endeavour. In the true spirit of design thinking, you start with a thorough discovery of problem or pain point, before you move on to the solution — in this case, the framework. To keep this post manageable, I’ll keep it limited to the discovery space of the framework for now.
The problem space
Over the last couple of years, I have been an observer and actor in many data projects. Some were better organized and more effective than others, but many of them had ill-defined ways of working and were somewhat low on project management. While doing some searching around, I became a little dissatisfied by many of the established ways of approaching a data & analytics project (e.g. crisp-dm). I found their focus to be somewhat ‘narrow’, this while many tracks in the real-world are usually more holistic and cross-functional in nature. Here, data analytics is sometimes no more than a ‘workstream’ and needs to be aligned to the other disciplines and workstreams of the wider track.
With time, this initial dissatisfaction became enriched by 2 other observations:
- I found many data scientists and data analytics professionals to be quite ‘methodology’ focused. Many were ‘algorithm centred’ and tended sometimes to forget about the actual business problem at hand. There is little exploration or reframing of the initial problem statement, this while usually the initial business question is just the tip of the iceberg. Through some digging and exploration, you would usually uncover a whole new world that makes the initial business question seem like a mere draft. If you would have continued to train an algorithm straight away, at best you would have ended up with a solution that has little relation to the actual problem at hand.
- Many business and design tracks tended to shie away from quantitative data and analytics and had a preference for the more qualitative data sources. I’ve seen plenty of tracks enriched by user observation, surveys and interviews, but devoid of more quantitative data sources or insights from algorithms, this while different perspectives can usually be quite enriching.
Based on the above observations, I started to wonder. What if we could somehow align the worlds of business (read: product & design) and data & analytics and create synergy instead of dissociation. What if we (“How might we”) could bring in one framework that is flexible enough to incorporate many workstreams and known widely enough to spark recognition?!
The solution space
I guess the introduction already spilt the beans. Without further ado, find below the double diamond framework tweaked sufficiently for data analytics tracks.
The framework contains 9 different steps and 5 milestones but can be customized further for every situation. Let’s go through the discovery space of the framework at the hand of an example of marketing automation for mortgage churn.
- Point of departure
For every track, you need to have a point of departure. This is basically the initial briefing that will kickstart your track and provide the preliminary direction all the next steps. Usually, this point of departure briefing contains some of the following elements (this is not an exhaustive list)
- A title and description of the initiative
- The goal of the initiative
- The initial challenge to solve
- Target group & Value
- Desired timeline
In the case of the marketing automation for mortgage churn initiative, our initial description was ‘Mortgage churn orchestration’. You’re probably thinking ‘Orchestration’ what do they mean with this in the context. Well, you’re not alone, so did we. So it really stood to reason that we had to explore this description without the business owner. Apart from that, knowing that this track was part of a larger marketing automation program, we knew that one of the main target groups that we should provide value to were the marketeers. Lastly, we knew that we had a horizon of some 4 months to come up with a minimum viable product (MVP) which I will call a prototype in this context.
2. Discover
Having our initial project briefing, we moved out to discovery. Here you focus on the current reality or the ‘what-is’. We already knew that one of our target groups were the marketeers, but we started thinking of other personas for whom this track may be relevant. We went through the following steps:
- We came up with a list of personas that were involved in the current marketing campaign processes such as communication specialists, database marketeers and the campaign engineers.
- We started mapping the campaign process through multi swimlane flowcharts. Here each swimlane constituted the steps that each of these personas had to take. We created this flowchart through ‘co-creation’ interviews with these personas
- We then moved on to putting pain points & points of opportunity post-its in the flowchart. These post-its would be essential input for ideation
3. Insights acquisition
Usually, the insights acquisition phase flows logically from the discovery. Here though, you take the gaps during your discovery as a starting point to acquire additional insights. In this phase, you answer the questions that you had no answer to yet during your discovery phase and for which it is really hard to come up with an answer based on existing (desk) research. In our current track, we went through some of the following steps:
- We thought of additional ‘perspectives’ to look at the challenge. In the discovery phase, we heavily relied on the personas involved in the campaign process, but at the end of the day, there’s also the ‘bank’ perspective (read: business owner) as well as the ‘end customer’ perspective
- From each of these additional perspectives, we started answering a range of research questions. For instance, from the bank perspective, we really liked to get a sense of the business case. If we talk about churn, how big is this problem? What are the main reasons for churn here etc? From the customer perspective, we really liked to know some more on why are people are churning. Here we consulted a lot of qualitative data sources.
4. Define
As you can probably tell from the previous 2 phases, we did a lot of research on the challenge at hand. In design thinking terminology, we were ‘diverging’ and creating a whole range of research questions. The define phase is about converging which comes down to bringing all this research back to the essentials. What are really your key insights? What is your new problem statement after having looked at this bulk of research? Here, it would really make sense to visualize your findings first, before you synthesize and interpret them. A range of visualization tools can help here such as ‘as-is customer journey maps’ or ‘personas canvases’. At the end of this phase, you’ll also have your first milestone: ‘The discovery review’. Let’s have a look at this milestone moment for marketing automation mortgage churn.
A. Discovery milestone review
- Review of personas and problems to tackle
In this review moment, it makes sense to answer some of the following questions:
- Who are the key personas that you are doing this track for:
- The marketeer
- The communication specialist
- The database marketeer
- The campaign engineer
- The marketing business owner - What are their current problems
- Campaigns are marketing plan-driven
- Our set-up barely allows for multi-step campaigns - How are they solving their current problems right now (alternatives)
- New campaigns are created to ‘imitate’ multi-step
2. Requirements on how to move forward:
- You’d need a high-level understanding of the future state and how this should look:
- A reality where campaigning has become data-driven instead of marketing-driven - You’d need to have clear requirements and design criteria for this future state that can feed the ideation that follows
- Algorithms should help with making the channel selection
3. Business case:
- A clear link between this track and strategical relevance:
- The track was strategically relevance up-front - The track is sized in monetary or strategic value:
- We did a lot of sizing, but these numbers are confidential - An indication of the expected usage of the potential solution built in the solution and implementation space:
- We had the business (marketing here) on board, so we had full support that whatever prototype we would make, we would at least test through a proof of concept to see whether it would work
Some finishing remarks
The discovery milestone review milestones for us marked the end of the discovery space. In reality, we had quite some movement between the various steps here. We sometimes went back from defining to discovery and insights acquisition, so it was by knows means linear. After all, we were still trying out when we did this specific track. However, all the insights that we acquired, visualized and wrapped up in the milestones review would be the input of then solution space. They would be used to ideate on possible solutions and to kick-start our MVP design. I, however, customer segments promised you upfront that I would keep this post somewhat manageable in size, so for more information on the solution space, the data science toolbox in the discovery space as well as experimentation methods for solution validation, please stay tuned.
