The latest technology and digital news on the web

Powered by

Scrum didn’t work for my startup — so I advised an addition activity administration method

  • Tech
  • Software development process
  • Product (business)

Scrum didn’t work for my startup — so I advised an addition activity administration method

Built In
Story by
Built In

This commodity was originally appear by Built In.

Nowadays there are more opinions on how to advance software than there are firms absolutely doing the development. Some people call programming a science. Others call it engineering, a craft, or even art. Just as coding’s 70-year history has produced hundreds (if not thousands) of languages, it’s also spawned endless processes and frameworks purporting to have the best model for allegorical a team to build software.

In recent times, no framework has seen broader acceptance and more hype than the scrum accomplishing of agile. Agile — at its core — is very simple: a focus on collaboration, functionality, and adaptability over formality, documentation, and all-encompassing planning. But just like software development overall, there are more variations on agile than you can shake a stick at. Scrum, one such variant, has proven very popular.

Since co-founding 4Degrees three years ago, I have spent a lot of time with my team alive out what software development action would work best for us. Agile all-embracing makes a lot of sense: It ensures that we’re talking to customers, accommodating internally, and blockage nimble; it’s a good match for our core values. Scrum, on the other hand, really didn’t match our needs very well. While scrum may be absolute for many organizations, we found that it didn’t suit us — and likely isn’t a good fit for most other startups either.


In scrum, there are three audible roles: the artefact owner (who decides what appearance should be prioritized), the abstruse team (who absolutely builds the product), and the scrum master (who coordinates the process). Right from the get-go, this analysis of labor didn’t make sense at 4Degrees. In the early days, we didn’t even have that many people on the team, much less all adherent to a specific type of functionality development. My co-founder, Ablorde, and I absitively early on that we would share the role of artefact owner in the organization. But I was also doing most of the actual programming. If we were to have a scrum master, that apparently would have fallen to me as well. Clearly it didn’t make sense for me to be  of the audible scrum roles.

We absitively to set up our team a bit abnormally than the scrum structure. Ablorde and I still share artefact owner responsibilities, though our ultimate goal is to bring in a head of artefact (similar but hardly altered from scrum’s artefact owner, which is often tied to the business in some other official capacity). We anticipate the head of artefact and the other people in the “product” alignment as arena the role of both artefact owner and scrum master: analogous across engineering, design, sales, marketing, and anybody else to figure out what appearance should be prioritized and then bringing the team calm to execute.


Oftentimes, scrum is anticipation of as a very fast-paced framework for software development. And indeed, when alive from a avalanche action with six-month-long development cycles, then deploying every two weeks feels like lightspeed. Not so in the startup world. When you’re just accepting started, you might be a absolutely altered aggregation in two weeks. Planning to update your assembly appliance only twice a month is a recipe for adversity and black customers. At 4Degrees, we deploy  more frequently: about two to five times per day. That allows us to get out important updates bound and divide work into more atomic units with fewer dependencies. Accordingly, we do not use any formal “definition of done” — tasks are “complete” when they’re live in assembly and at least abundantly tested.

Accordingly, we don’t anatomy our artefact development cycle into “sprints,” one of the most axiological apparatus of scrum. The idea is to have the full team focus on one botheration for a set period of time (most often two weeks). That just didn’t work for us; we couldn’t take two weeks to put out new functionality, and many times, it just doesn’t make sense to have the whole team focused on one set of tasks. Our back-end architect may be doing some data science analysis at the same time as our artist is administering user interviews for some abeyant new functionality slated for weeks down the road.

Instead of sprints, we anatomy our development cycles into one-week and one-quarter increments. Every Monday morning, we check in to hear the latest on what barter are saying and prioritize/discuss the tasks that the team will be alive on for the week. The affair helps anybody stay up to speed on what’s happening, but it’s not the end-all-be-all: Tasks and priorities often shift over the course of the week as we abide to talk to barter and iterate on the code we’ve deployed.

In accession to our weekly cadence, we take a step back once a analysis to think about our belvedere a little more strategically. We assay our chump insights and begin major initiatives that may advance the project. That effort after-effects in a roadmap of new functionality for the quarter. That roadmap is often more aspirational than realistic: Changing advice and accommodation over the analysis mean that oftentimes half of the roadmap won’t be completed on time. But that’s OK, it still provides us a path to follow to make sure the belvedere is headed in the right direction.


The structural differences amid our action and scrum apparent in a very altered access to how work is organized. Scrum prescribes epics, stories, and tasks to ensure that the chump angle is translated into functionality. Belief (“As a user, I need…”) have become an abominable authentication of scrum, and even agile more broadly.

At 4Degrees, all abstruse work is simply recorded as tasks: line items in our Airtable database. Chump centricity is the first of our five accumulated values and is deeply chip into our process. As such, we haven’t seen any need for artificially diction tasks as if they’ve come beeline out of the mouth of a specific user.

Rather than relying on team accord for bringing tasks into a sprint, we use a centrally bent antecedence and point value. Specifically, whenever a new task is added, I will assign that task a antecedence and point value. Higher antecedence tasks are sorted to the top of the task tracker and accommodate a slight bump to the “raw” point value of a task. Point targets are set and abstinent at the alone level rather than the team level. We’ve found that this works better because our team is so small and almost disconnected in its responsibilities. We have a front-end engineer, a back-end engineer, a mobile engineer, and a designer; each of these team associates has almost little adeptness — much less desire — to guess the points that a task should be worth to accession else on the team.

Do what works

Due to scrum’s popularity, many practitioners have abandoned that the agile acclamation says “individuals and interactions over processes and tools” and “responding to change over afterward a plan.” While scrum works for many organizations, it has been force-fed to all startups, even when it’s far from ideal.

The most important abstraction to take from agile is being quick and acknowledging to things that don’t work. For many startups, the overly structured and almost slow nature of scrum just doesn’t make sense. Instead, they’ll be better served by taking a faster, more accepted access to experimenting and accession out a software development action that is best suited to their specific needs.

Appear July 1, 2020 — 09:30 UTC

Hottest related news