Reorganising Fynd to Build the Next Wave of Products and Features
Since the first release of the Fynd App on October 22, 2015, we have launched some industry first features – Gravity View, Flashpay, Mix & Match, and Fify: Your Shopping Bot-friend, in addition to open-sourcing our pitch, our growth hacks, and our scaling efforts.
A few months earlier we took stock of our efforts and what we wanted to build next. A big realisation was for us was to build the next generation of products, for which we had to change how the teams operated, and also upgraded everyone’s skills to make them Machine Learning experts.
Working with the simple paranoia:
We have reorganised the engineering team from technology specialisation silos into something we refer to as Domains.
Domain is a collection of closely associated technical problems around a customer/business proposition.
This new structure will allow us to build focused features which is going to impact business metric foremost and also help build fluidity of people learning new skills.
Our quest is to continue building better and also to share as much as possible with everyone.
Why did we do this?
Initially, the teams were classically organised as technical specialisations like iOS App, Android App, Platform, Web etc. This structure is great when product development has just begun, but as the platform and product matures, a new set of problems needs to be solved. The solution to the problem might require new skills for which you can either hire new resources or cross-train your current team. Hiring new resources is not an optimal solution as a large part of the team will remain under-utilized.
Before we adopted the new team structure, we had a month-long Applied Machine Learning Bootcamp to update everyone’s skills.
The next generation of products can only be built if each team applies machine learning techniques to solve existing problems.
Subhranath, VP Engineering — lists the pro and cons for anyone pondering how it will impact the product roadmap. Be very thoughtful before you go down this path. We would recommend this only for startups with large engineering teams with fairly mature products.
Aligning the Domains
In the classic team structure, the team focused on tasks and not so many metrics of improvement. Now each domain has a metric tied to a pivotal event in the customer journey.
For instance, the Reco domain’s event metric of improvement is products viewed per session. To improve this metric the domain has to recommend the most relevant product to the user, which requires modification of the recommendation engine as well as the Apps. The domain is focused on solving the problem and will do whatever it takes to improve it.
Defining DONE
To ensure high-quality products are shipped consistently across all the domains we had to define what DONE meant. A common definition of DONE ensures everyone understands the quality expectation and estimates timelines accordingly. The definition changes between engineering and non-engineering domains but follows the same spirit.
For engineering domains,
DONE = Production Release + 7 Bug-Free Days
The 7 bug-free days ensures we don’t move to another problem without ensuring the shipped work is of high quality.
Open-Sourcing & Sharing
We are big supporters and consumers of Open Source technologies. Indian startups are big consumers but don’t actively contribute to the community. One of the reasons for starting this blog was to actively share our learnings, the culture and methodology of building the company.
We think we can do even more. From March 2017 we are launching a dedicated Open Sharing platform gofynd.io. Each domain will be open-sourcing core code of the Fynd Platform and Apps. We will also be publicly releasing 2 APIs for everyone to use
- Vision Auto Tag API: Post a product image to get product attributes
- Product Feed: Request product information based on location, gender, category etc.
We will continue open-sharing/sourcing as we strongly believe this will make the Indian startup ecosystem stronger. We learn from others and others can help us build better.
2016 was a very interesting year for the Indian startup ecosystem. Tough times are the best times, good times make us lazy.
Wishing everyone a great (tough) 2017.
#HappyFynding.
