Developing for the OTTO platform: What have we achieved so far?

In this blog article, I describe the following phases:

• Step 1: Preplanning
• Step 2: MVP until February 2019
• Step 3: Status Quo

Of course, such a large project is not easy. At the beginning, a lot of questions arise, for example:

• What exactly is to be developed?
• How do we want to organize ourselves?
• Which technologies and architectures do we want to use?
• What do we call the project?

Of course, you can spend months or years on these questions and still never come up with the perfect plan. Many of us still know this from the waterfall world. That's why it was clear to us early on that we wanted to develop in an agile way. But agile doesn't mean haphazard either.

We therefore started a two-month preliminary project with about 30 people from IT and business. Using the method of Domain Driven Design we approached the question of what capabilities we needed as a platform. Despite a lot of prior knowledge, it was very helpful to think about all the topics in a completely new way.

Understand the Business:

First, we got to grips with the method of Event-Storming and worked out which steps have to happen one after the other in order to create an optimal business process. In the process, we basically created a wall full of ideas that we collected on sticky notes. From left to right, these sticky notes lead us through the customer or partner journey:

With this information, we were equipped to define a Minimum Viable Product (MVP). For example, an MVP for us was the development of a live deployed software for the marketplace to gain experience together with our first beta test partners. This gave us immediate, direct feedback and allowed us to implement improvements quickly.

We cut as many corners as possible at the beginning to keep the complexity small for now. For this MVP, that meant we developed only two payment types and no shipping costs as a first step. This allowed us to deliver results quickly. However, we were sure that with our agile approach, specific features could be added gradually - in a decentralized way, i.e. without joint (and sometimes time-consuming) coordination of all teams.

For this MVP, we have set a target date of February 2019. And this much I can proudly reveal at this point: We kept the deadline.

Names do matter!

Names often seem like smoke and mirrors. But they help a lot in communication and identification with one's own work. In analogy to the Lhotse project , we have given this project the name DeepSea. For us, the name symbolizes that we are not only renewing the part that is visible on the surface, but also the part that takes place beneath the surface. And there, in the deep sea, there is also much that is unknown.

In addition to the question of content, we also thought about how we wanted to carry out the project organizationally during the preliminary project. It was particularly important to us that the teams can concentrate on themselves as much as possible and are not constantly burdened with status reports. Nevertheless, we want to keep our managers regularly informed about the status quo so that they can provide us with the best possible support where necessary.

These were our six recipes for success:

We started with the definition of our Technical Manifesto. In it, we set out our self-image and the demands we place on our software development. Basically, it is divided into two sections: macroarchitecture and microarchitecture.

Macroarchitecture

In order to support decentralized product teams, they should also be as technologically independent of each other as possible. Therefore we have defined the following rules:

• Each team develops self-contained systems. This goes hand in hand with independent deployability and the extensive renunciation of synchronous communication.
• Each team maintains its own infrastructure in its own AWS accounts.
• Each team should be able to deploy at any time. Usually with continuous deployment
  

Microarchitecture

• Within their area of responsibility, teams are free to choose technologies, but their responsibilities require them to independently consider available expertise, potential synergies, etc.
• At DeepSea, the technologies used are actually very diverse. Very commonly used technologies are Java, Terraform, DynamoDB and MongoDb and SQS/SNS.

If you're interested, I'll be happy to go into this point again in a separate post. Just leave a note in the comments below the article.

In software development, MVP means Minimum Viable Product. In other words, a product that is as small as possible but still adds value to the business. The goal is to be on the market as quickly as possible and to create a basis for continuous further development.

We started the MVP phase (approx. 6 months) with three product teams. We started from the beginning with the selected specifications from the pre-project and therefore also started immediately with bi-weekly sprints. Thus, we were able to show executable software already after the first sprint. After about three months, we were so well-rehearsed that we created three additional product teams and were thus able to develop all the required products for the MVP. In order for the three new teams to benefit from the experience already gained, people from each of the existing teams were transferred over.

In the last two months before the deadline, the joint testing phase with the first partners began, which actually went very smoothly. Although we were all deploying continously and therefore technically already live for a long time, the rollout date was still very exciting. Because that was the date when we actually flipped the switch in the otto.de webshop, with which the new partner products could actually be found and ordered.

The development of this MVP was very instructive for us, because it allowed us to test a lot of things without having to consider all the complexity. Our first two partners were also very satisfied.

The first weeks after our MVP, we focused on delivering some features that we felt were very important to make the MVP "enjoyable". Since then, however, our main focus has been on partner growth. This includes simplified and, above all, automated registration as a partner who wants to sell products on otto.de, as well as expanding our product ranges. I'll tell you more about that in the next article.

Is the most exciting phase already over? Not at all! The focus is now shifting to the topics of partner UI, which is particularly essential for smaller partners, and the further development of logistics services to further simplify processing for our partners. So this incredibly exciting journey has just begun. If you found this article interesting, please write it in the comments.