Kerika’s internal product development workflow

It will come as no surprise to our users that the Kerika team uses the software for every aspect of the company’s internal operations, including, of course, our product development.

(There really isn’t anything better on the market for planning and executing product roadmaps — we know, because we have checked and made sure we are better than everybody :-))

We are sometimes asked about our own internal workflow: how we plan releases and manage our Sprints, and in particular how we have gone without using regular old email for over 4 years now!

Well, here’s a glimpse at our internal workflow:

Kerika internal workflow
Kerika internal workflow

We work in two-week long Sprints; we have found two weeks to be the ideal Sprint length since it provides enough time to produce something — at least some usable part of a new feature — while not being so long that we forget we are supposed to be working as a Scrum team.

Planning

Whenever we plan a new Sprint, we are always also looking forward as well: hence the columns titled “Sprint +1” (i.e. the Sprint that will come after the current Sprint), “Sprint +2” (the Sprint that will come after the next Sprint), etc.

This gives a view into the next several weeks of our product roadmap which makes sure we don’t approach any single work item with tunnel vision.

Now you might notice that “Sprint +3” in the list above is way larger than “Sprint +1” or “Sprint +2”: that doesn’t mean that we expect a sudden spurt in productivity (i.e. increase in team velocity) in 4 weeks time.  It just means that as we go further out in our planning there is more uncertainty about priorities and so the Sprints that are well into the future are not well defined — and that’s perfectly OK.

Quality Assurance

Within each Sprint we have a fairly conventional flow: Planning, Development and QA.  The QA step actually consists of three separate stages:

  • Code Review: we try to make sure we review all the code we are writing in any Sprint, unless it is a really trivial change we are making (e.g. changing the label on a button).
  • Deploy to Staging: we have a separate set of servers where we test our new releases.  This also happens to be where we have all our real projects running: we believe very much in the idea that “developers should eat their own dogfood”, so we run out entire business on the latest build that went past Code Review.

    This can be a little scary or frustrating at times: if there was a bad release to our staging environment, it can bring every project and board to a screeching halt.  But, from our perspective, that’s the whole point: make it everyone’s business to ensure that we produce high-quality code, and make it everyone’s priority to fix any problems that come up.

    This model of living on the bleeding edge of our product gives us a really good incentive to write good quality code!

  • The final QA step is Show & Tell, where the team formally presents the new features or bug fixes to the Product Owner.  With each new feature we will have identified a set of test cases, as part of the Planning phase, and these are used to formally check the new feature in a meeting attended by everyone on the team.

    Very occasionally something will get rejected at the Show & Tell stage, in which case the card gets moved back to Planning or Development and flagged as “Needs Rework”.  It’s more common for work to get rejected at the Code Review phase, not because it is buggy, but because it might not meet our internal coding standards.

Deployment

Once a feature or bug fix has passed Show & Tell, it is ready for Deployment to Production.

We have a continuous integration process for handling code changes — pulling them from our internal git code repository — but we don’t do continuous deployment. Instead we prefer to deploy on the last day of each 2-week Sprint.  We usually time this for Friday morning, Indian Standard Time, so that our developers in India can take one final look at the system working in production.

Documentation

We have an unusually strong emphasis within the team on creating documentation at the same time as we write code.  Many small companies skip documentation because they think it will slow them down.

In our own (sometimes bitter) experience, skipping documentation is false savings: if there are problems to be fixed later, or even if a feature simply has to be extended in the future, it’s very hard for even the original developer to recall the logic that she used to write the code in a particular way.

What’s important to note, however, is that we don’t have very big documents: most documents are less than a page long, because they refer to very specific work items.  But we have thousands of these small documents, since every feature we have ever produced, and every bug we have fixed, has been documented.

And thanks to Kerika’s very cool integration with Google Docs (we use Kerika+Google), managing these thousands of documents is very easy: just open the relevant card for a specific feature or bug fix and you will find all the relevant documents as well:

Documents attached to cards
Documents attached to cards

For new features we always have a short Planning Document that identifies any existing modules that will be affected and provides an outline of the new code that will be written. Here’s an example:

Example of a Feature Planning document
Example of a Feature Planning document

There may be other documents created, depending upon the complexity of the new feature, but even the smallest new feature will have at least one planning-related document that’s written before the code is written.

Bug Fixes

For bug fixes, we have a Repro, Cause & Fix document that methodically analyzes the cause of a bug:

Root Cause Analysis
Root Cause Analysis

The most important section of this document is the Introduced Since piece, where we trace the origin of the bug back to its source, to understand what new feature or bug fix we were originally working on that resulted in this new bug appearing.

This methodical root cause analysis, which we do for even the most minor bugs, helps us identify patterns in our code writing that we need to improve.  Sometimes it can even point to bugs that we haven’t discovered yet: the root cause analysis may identify an assumption that we have applied in other places, all of which need to be looked at to make sure there aren’t other variations of the same bug that haven’t been identified yet.

After the bug has been fixed, the Repro, Cause & Fix document is updated to reference the git branches where the code fix can be found.  This completes the circle of careful analysis, careful execution, and methodical review that we strive to adopt (and improve continuously).

The End of Email

We stopped using regular email towards the end of 2013. There was never any formal decision to renounce email; instead there was a formal decision to “eat our own dogfood” in a very serious way.

And as we started to build our entire company using our own product, we found that using Kerika is so much better than using email that there really wasn’t any need for email any more.

So people simply started drifting away from email. There was no explicit decision or formal movement to do so, people just realized, on their own, that email is an exceptionally poor way of managing work within a distributed team.

Conclusion

Considering the size of our team — and entire company, for that matter — you may be surprised by the degree of formalism we have adopted, and the heavy emphasis on analysis and documentation.

We spend at least as much time on analysis and documentation as we do on “pure coding”, and if we add in the Code Reviews and other elements of formal QA, the actual coding time is a relatively small portion of our total expenditure.

But trying to skimp on analysis and documentation really doesn’t pay off, in our experience: if you are building a company and product for the long-run, as we very assuredly are, you need to build it right, not just fast.

Updating the look of Board Settings

One final (?) bit of restyling, to make all of Kerika consistent with our new look-and-feel, has been updating the Board Settings dialogs.

Board Settings
Board Settings

The functionality is essentially the same, but the appearance is cleaner, lighter and more in keeping with the Material Design standards we have (mostly) adopted.

(We say “mostly”, because there are some elements of Material Design that we find unattractive.  For example, for the on/off toggle switches we prefer the iOS style buttons.)

Column Settings
Column Settings

The Column Settings dialog has also been restyled, and looks nicer and cleaner. The example above shows a board that uses Work-In-Progress Limits.

Tag Settings
Tag Settings

And the same with Tag Settings: we have a restyled color picker, and better messages for warnings when tag names or colors might clash.

Enjoy.

New feature: alpha sort of cards

At a user’s request, we have added alpha sorting of cards on Task Boards and Scrum Boards:

Sorting alphabetically
Sorting alphabetically

This is most useful if combined with the Auto-Numbering feature in Kerika, that can automatically insert a number at the beginning of each new card: the alpha sort will sort all the cards in the column by their number.

Usability tweak: reordering Board Attachments

We have made it easier for you to re-organize all the Board Attachments on your Kerika Task Boards and Scrum Boards: you can grab any of these and drag up or down the list to re-order them:

Drag handles on board attachments
Drag handles on board attachments

Use the drag-handle shown on the left edge of the attachment to drag it up or down.

Dragging board attachments
Dragging board attachments

(And, by the way, this feature is also available for Card Attachments.)

New feature: partial sorting of cards in a column

We have improved the sorting feature for cards on Task Boards and Scrum Boards to allow for partial sorting: if you select some cards within a column and then do a sort, the sorting action will apply only for the selected cards.

Sort options
Sort options

This will make it easier to organize very large boards, e.g. where a single column may contain a hundred cards or more.

New feature: we made it easier to use Card URLs

Every card (and every canvas) on a Task Board or Scrum Board in Kerika has a unique URL, but most of the time you might not notice the URL shown in your browser’s address bar is changing as you open one card after another.

(And if you are using Safari on Macs, you definitely won’t notice this since Safari hides most of the URL anyway.)

These URLs can be helpful in many ways: Kerika recognizes them as pointing to other cards or canvases, and makes these links an easy way to connect up different work items.

Here’s an example of a card URL that’s referenced in a chat message:

Referencing Card URL in Chat
Referencing Card URL in Chat

Any URLs that are included as part of the chat or details of a card, that reference other cards, canvases or boards within Kerika, are automatically recognized and presented as useful links as you can see from the first chat message shown in the above example.

To make it easier to get these card URLs (and to help you notice that they are important in the first place), we have made it possible to grab any card’s URL with a single click:

Card URL button
Card URL button

Clicking on the new Card URL button that appears on the top-right of each card’s detail view makes it possible to see, and copy, the card’s URL with a simple click:

Getting Card URL
Getting Card URL

We have made it easier to grab the URL of an entire board as well, from your Home Page:

Board URL
Board URL

Try this way of creating links between related work items, across all your Kerika boards.

An improvement to Views: eliminating Templates

As one of our users pointed out, Kerika’s ViewsWhat’s Assigned to Me, What Got Done, etc. — shouldn’t include any cards from templates, just regular boards.

People who use templates on a regular basis often pre-assign cards in the template: for example, an employee on-boarding template that involves HR tasks may be preassigned to a specific HR employee.

Our initial implementation of Views included cards from Templates as well, which led to a misleading impression of the amount of work, particularly unscheduled work, that was waiting for a particular person.

That’s fixed now: Views will automatically exclude cards from Templates.