Four building blocks for scaling insights — Part 4: How we train and support the organization to become self served with insight
In previous articles of our blog series “Four building blocks for scaling insights”, we described three building blocks that we use to create an insight-driven organization: a Data & Insight team consisting of the right people organized in the right way, the necessary infrastructure and tools, and a unified and structured way of working with our data.
While each of these building blocks is absolutely necessary to create an insight-driven organization, they are completely dependent on one last building block.
Your company might have world class data analysts, a perfectly organized and maintained data warehouse, and a really good business intelligence tool, but you will never realize your potential if your organization does not have the skills and motivation to use data for better decisions.
Hence, our fourth building block is Training and Support.
We believe that everyone in a company make better decisions based on insight, thus insight should be available for everyone too. We have worked really hard over the last few years to spread this philosophy, so we were delighted when even our French baker asked for access to our business intelligence tool (Looker).
We have established three guidelines for how we train and support our colleagues:
I will now elaborate on these three guidelines.
1. We answer questions quickly, and we ensure transparency and knowledge sharing by keeping them in open channels
We all know how frustrating it can be when our sense of flow is disturbed because we are suddenly kept waiting for someone else’s input on a problem we are solving. Getting stuck with work in progress is an energy drain and we might lose faith in our ability to solve the problems we are working on.
With this in mind we aim to help our colleagues as fast as we can in the way that suits them best.
In Kolonial.no we have established two main support channels for data and insight related questions: We have a support channel in Slack called #data-insight-support, and we arrange Data Hours where our colleagues can sit down with us for 1:1 support.
#Data-insight-support
Before we had established a Data & Insight team in Kolonial, and while Metabase was still our business intelligence tool, there were no designated people or places to turn to for help. Solution-oriented as our colleagues are, they created an open Slack channel where they could help each other with questions related to data and insight, including Metabase.
As the Data & Insight team was established, and the company grew, it became clear that in order for our colleagues to stay motivated to use data and insight in their work, we had to make sure that the questions that came into the channel were answered quickly and accurately. This could also solve the problem of questions being asked directly to specific Data & Insight employees, causing interruptions to their work, personal queues and lack of transparency.
Today we make sure that each week two people in the Data & Insight team have the main responsibility for answering questions in the channel, while the rest of the team monitors and helps out whenever needed. The questions range from the more concrete type, such as how often different data is updated, to the more analytical type, like whether or not we can confirm that a graph in Looker confirms a specific hypothesis.
This approach limits the constant interruptions for all of us in the team and forces us to get a view of urgent matters across all our users. That being said, we had to, and still have to, make some extra efforts to ensure that our colleagues use the Slack channel. When they contact us in direct messages we always encourage them to ask the same question in #data-insight-support, assuring them that they will get an answer there.
We experience that an important benefit of providing support in an open channel is that everyone in the company can read and learn from each other’s questions. Now we also see that our colleagues have started to help each other which is really great! However, with a growing organization and thus growing membership in our open Slack channels, we emphasise the importance of creating a company culture where no one is afraid to ask questions. When you ask a question, you do several others a favor as well.
Data Hour
While the #data-insight-support channel seemed to serve the purpose of answering data and insight related questions quickly and accurately, and thus allowed our colleagues to do as much as possible themselves, we still experienced the need to sit down with them sometimes. The Slack channel is simply not suited for all the problems our colleagues are working on, as sometimes they do not know how to formulate what they are stuck on into a specific question. Furthermore, while the answers we provide in the Slack channel make sense to some of us, they might not do so for everyone.
Therefore we established a concept we call Data Hour, where people can bring their laptops to a 1:1 problem-solving session with members of the Data & Insight team. The moderators of #data-insight-support host Data Hour during dedicated time slots at each of our office sites and/or in virtual meetings, where anyone at Kolonial.no can come to them with problems and questions. They are available to help people do things themselves, or sometimes understand complex problems better to scope the work we will deliver.
2. We customize our training content to match our colleagues’ different skill levels
As late as 2019 Data & Insight had only one main course in the Kolonial.no course portfolio, namely Introduction to Looker. This course was basically a “one-course-fits-all” type of course, with just minor content adjustments depending on which business areas the course participants belonged to. The course worked just fine, and most of our colleagues gained a basic understanding of what Looker is and how it can be used. However, we noticed (and confirmed with feedback) that there was a large variation in what participants learned from the course, depending on their technical and professional background.
We want everyone in Kolonial.no to benefit from insight based on data, regardless of position, subject area and professional background. To achieve this we identified the need to develop a course portfolio with several levels of difficulty, and learning materials that covered not only how to use our BI tool, but also how to make decisions based on data and analysis. Early in 2020 we started concept development, and in the fall 2020 we launched Kolonial.no’s Data University.
The Data University is a learning program containing material on how we use data in Kolonial.no, and how our colleagues can use data in their daily work to make better decisions and build better data products. The content is divided into different topics, and has three levels of difficulty: Basic, Intermediate and Advanced. We have specified learning objectives for each course, and described in detail which skills are required to reach each difficulty level.
We now offer courses within five main categories:
- Technical tools for exploring data (Looker / Amplitude)
- Evaluating data / Experimental design
- Decision making
- Other technical tools used in our Data Scientist / Data Analyst workday (Kubernetes / Docker /Git)
- Forecasting
Everyone in Data & Insight contributes both to create course content and to arrange the different courses. We have also recorded most of them to make them easily accessible. We continuously adjust which courses are included in the Data University, and what their content is, as we get feedback from the participants and as subject areas develop. We experience that our embedded model makes this process a lot easier as the embedded Data & Insight people can identify competency needs across functions, and work together on how we should fill them.
Currently, our Data University consists of the following courses:
And so far we have received great feedback, but as always there is room for improvement.
3. We encourage and motivate our colleagues to utilize data in their daily work
Motivation and a feeling of achievement is just as important when working with data as it is for most other aspects of life.
As we are passionate about ensuring that data and insight becomes an instinctive part of our colleagues’ problem-solving process, we must trigger their interest and self confidence in gaining insight through the use of data. We also try to lower the threshold for what can be called insight as there is often no need for a large and complex machine learning model. A simple presentation of historical data can also be a very efficient eye opener.
To trigger interest in data and lower the threshold for what can be called insight, we have created an open Slack channel called #interesting-insight.
In this channel everyone is invited to share small and large nuggets of insight that might be useful to others or sometimes simply just fun.
We try to motivate and remind our colleagues of the importance of using data by always bringing our data and insight mindset in meetings and discussions. But it is very important to find the balance point where we motivate people and give them valuable insight, without challenging them so much that they lose their motivation or self confidence in using data.
We have now taken you through our three guidelines for working with training and support: We direct questions to open channels in Slack and answer them quickly, we customize training to match ability, and we encourage use of data / insight.
Let us now summarize the four building blocks in scaling insights that has enabled us to become self served
First of all we, have the right kind of people in the Data & Insights team, and organise them close to the decision makers. At the same time we are a close knit group that is well coordinated and learn from each other.
Secondly, we have over time moved to a mature stack of insight tools put together in a modular architecture. The main decision that is difficult to change at this point is the fact that it is SQL based. Despite many contributors, we have control as we are using version control and we strive towards single source of truth through our transformation in DBT and metadata layer in Looker.
With more than 321 tables in Snowflake we make it manageable by having standardized on development practises. Strict naming conventions and taking a lot of best practises from software engineering has increased our productivity and control.
Finally we make sure our decision makers use the data and tools in the right way by delivering timely and dedicated support face to face and in Slack. We continue to close competency gaps by providing in house training that we develop ourselves.
Feel free to check out other post from my colleagues in Product & Tech here on Medium such as our scalable way of working with development in Kolonial.
We are also looking for great people to join us, and have posted new positions in the Data & Insights team.
