Hacker News new | comments | show | ask | jobs | submit login

This is a surprisingly common trap to fall into. I worked at a startup in which the CTO criticized my dev team for not having a perfectly linear burn down chart. Every week in our retro we had the same discussion about how to improve the shape of our burn down chart; we tried putting higher valued stories at the front of the queue, involving QA earlier in the sprint, and cranking out 1 pointers mid week to tweak the graph. We rarely achieved the perfect shape. On top of this we also had one week sprints, which made it difficult to plan for the future or work on spikes, so we ended up doing very little high-level, architectural planning.

I remember one week my team cranked through their allotted stories so fast that we ran out of work two or three days into the sprint. We decided to take on a couple extra easy stories that we knew we could finish before the end of the week. Rather than being praised for doing extra work, we were criticized for "introducing volatility into the sprint". In other words, we messed up the shape of the burn down chart.

I'm currently working at a place that has virtually no process, which has it's own challenges, but I'm happy that I'm not arguing about burn down charts anymore :)




> This is a surprisingly common trap to fall into. I worked at a startup in which the CTO criticized my dev team for not having a perfectly linear burn down chart.

You are describing this far more charitably than I would. I would describe it as idiocy.

For some (idiotic) reason some management types have this belief that when they measure some metric, it should fit some ideal shape when they put it on a graph, whether or not that makes any sense.

This is exactly the fallacious thinking behind stack ranking where people persistently abuse/misunderstand the whole concept of normal distribution by changing employee evaluation scores to match a normal distribution. It's like these people took a statistics class and only remembered the shapes they saw (and really only one shape) and none of the concepts underpinning those shapes.

Anyone who subscribes to this dangerously simplistic mode of thinking shouldn't have a position of leadership. Unfortunately many times everyone in leadership subscribes to the same cargo cult so they can't hold each other accountable.


We had a client pushing us on our velocity. Our team was doing a quality-first approach which took far more time than the other team's approach to pitch stuff over the wall to QA and wait for bugs to roll in. Our leadership took the deliberate approach of gaming the points system to improve our appearance while keeping the team 100% focussed on deliverying quality work at the same pace we had been all along. We didn't change a thing except inflate our estimates and the client was happy :)


> We didn't change a thing except inflate our estimates and the client was happy :)

Or, in other words, you follow the good old "Scotty engineering estimates" ;)


I also had this happen at a company I worked at, so as the leader of the software team I had a secret meeting with them and told them going forward we would create fake burndown charts for the CTO. Totally worked. He never complained again. For all I know the CTO was reporting on burndown to his peers and superiors and was trying to tell us to make it look better.


I'm dealing with a similar problem, except on 4 week sprints. It's not management, who doesn't care, it's the team itself which continues to be FIXATED on a burn down chart, to the extent that they massage story points, avoid doing work late in sprints, and are way too concerned with burndown in general.

Shipping valuable features without regressions is my metric of success. Not y=mx+b applied to story points.

Burn down charts are misused far too much to be considered generally valuable.


This all sounds exactly like Solzhenitsyn's descriptions of Soviet factory production, where the work near the end of the month was always rushed out the door to meet quotas.

The way he tells it, no one ever got in trouble for subpar products. But if production numbers were down, someone was going to the gulag.

Apparently, consumers were aware of this and would try to obtain goods that were produced near the start of each month.


this is exactly the CON of the burndown chart. it's a guide, nothing more. I wouldn't say its bad to learn from what the chart gives you, except if its trying to get a linear burn. In fact.... one may argue there is value in underfilling sprints (from a capacity standpoint) and then overdelivering, every time. otherwise known as... momentum or GTD


This was happening at a startup? How common is this kind of process slavery in startups?


Yes, it was a VC-funded, early-stage startup.

On top of the large amount of process and the one week sprints, we also four different dev teams each working on their own microservice app which was tightly coupled to the other apps in the company. Every team had to deploy at the same time, and if one app went down, the other apps went down with it. The company's platform was essentially an ecommerce site that could have been one medium sized monolith, maintained by a single team of 5 or 6 developers, but instead we had 20+ developers, several dev managers, a scrum master, and a large QA automation team. I guess you could call it premature optimization.

There were also plenty of issues on the business side that prevented the company from truly succeeding. The place was a bit of a clusterfuck but it was an excellent learning experience.


Sounds like the CTO is trying to preemptively solve all the problems they had at their last company.


But haven't you heard the VC driven truth that it takes one woman 9 months to make a baby... But 9 women can make a baby in one month.


Burn down chart?


It's pointless gimmicky make-work that presents your upper management with the fantasy that development work is not a difficult bespoke engineering effort, but rather an assembly line like they studied at business school.

In order to pull off this sleight of hand, real work is broken up into elastically-sized multiples of a subjective metric called "points" which the team is allowed to estimate. This estimation game is key as it builds buy-in from the team by appearing to "give them a seat at the table." Chunking the estimations into sprints is a smoke screen to keep the scam going.

Having acquired the buy-in from the people actually doing the work, the estimates are immediately converted to commitments by tracking progress on a burn down chart and telling upper management that the project will begin at 100% work remaining and proceed in a steady, linear assembly line fashion to 0% work remaining after the time allotted.

Deviations from a steady downward slope are interpreted as failures of the development team which paradoxically reinforces the standing of the chart presenter because they are able to "show the numbers". Unchecked, this ultimately elevates them to diety-like cult standing.

As the flawed assumptions underlying the chart are never questioned after the initial implementation, the use of burn down charts ultimately ends with:

- Blue Pill Ending: the company shoots itself in the foot by firing the team for failing to make the burn-down chart linear

- Seller's Market Ending: the team gets wise to the con and leaves the company

- Red Pill Ending: the process manager is fired after heroic intervention by the technical leadership

A process manager that wants to show up with their PMI certification and their burn-down charts faces a precarious end-game.


It's a common metric used in the Scrum methodology. You start off the "sprint" with a set number of "story points" (say, 30 points), and end the sprint with zero points. When you graph the team's progress over the course of the sprint, it should ideally form a linear, downward slope.

https://en.wikipedia.org/wiki/Burn_down_chart


It's one of the misunderstood pieces of scrum to think that burndown charts should be linear. If a team of 5 has established a velocity of 65 story points / 2-week sprint, and they accept 5, 13-point tasks for the next sprint, you'd expect a burndown chart that would look almost horizontal for the entire sprint until the very end, when it would drop vertically to the x axis. You might get linearity if there were 65, 1-point tasks...


I think you're getting at the heart of why burn down charts are so pointless. The shape changes sprint to sprint, depending on what kind of stories you take on. If the sprint is entirely high pointed stories, your chart will be mostly horizontal. If the sprint is entirely 1 pointers, the graph will look like an exponential decay.

Since the shape is always different, there's virtually zero useful information to take away from it.

Some people in this thread claimed that burn down charts are useful for finding out a team's capacity. I disagree - you don't need a chart for this. If your team has a target velocity of 30 points and routinely fails to meet it, reduce the target velocity. Likewise, if the team routinely completes all of their work well before the end of the sprint, slightly increase the target velocity.


Exactly.

My insistence is always that what happens inside the Sprint is the concern of the development team and nobody elses.

However it's important to remember that the team itself should set their velocity in negotiation with the product owner. This shouldn't be set by some external force.


Agreed - that is part of my own distaste for them...


This kind of thing makes sense if you were doing something like digging a trench, where the progress can be linear because the task A is the same as task A - 1. Why would anyone think that given a variety of tasks that each would take the same amount of time?


I agree. Being on the receiving end of Scrum feels like being an assembly line worker. I totally empathize with the business people who need some transparency into what the developers are working on week to week. Software is pretty abstract. I get that. But obsessing over burn down charts is the wrong way to go about it.


No one does, you estimate each task to have a different number of points, which modify the task's contribution to the burndown graph.

I guess the idea is that points should correlate to time (questionable), and if you're estimating perfectly, that should give you a linear burndown graph. Which is stupid. What might be less stupid is that if you want to get some idea of how well your team is estimating tasks, compared to i.e. this time last year, you could look at the volatility (across several sprints) this year compared to last year, and a lower burndown rate volatility might suggest that the team has improved at estimating the duration of tasks.

However, when you just focus on the volatility as a metric, then people start optimizing for it, which is not the point at all.


The theory is that a developer can estimate the approximate difficulty of a given task relative to other tasks and thus give a coarse estimate of how long a set of tasks will take. One is supposed to use this estimate, compare it against the amount of work actually accomplished in a given period, and then project ahead against estimates work remaining to get an idea of when the project will be finished. This is intended to help managers set expectations with customers and cut or adjust features if the date is too far into the future.

Unfortunately, what most organizations actually do is use the progress estimates to try to bludgeon developers into working faster, with entirely predictable bad effects.


An actual serious answer:

It's a line chart that shows story points starting at the top left in the beginning of the sprint. Which shows all points to be done, trending towards 0 points with 0 days left in the sprint, in the lower right.

Linear means that the same points are done each day to allow the line to be perfectly linear between the top left and bottom right.


Sounds like exactly the sort of not-actually-a-real-metric metric that would entrance a number cruncher. Forget those moments of excellence where everything gels and you could get days or weeks ahead... Sigh.


It's a metric that can help identify patterns within a team. It only works if everyone is honest with the amount of work they've done, and they consistently report time in the same manner.

A pretty common burn down chart would have a somewhat linear slope in the beginning, a somewhat less sharp slope in the middle (mid-sprint slump), and a sharper decline to close out the sprint strong. Notice I said common (more likely to happen) and not ideal, because there truly isn't such a thing as an ideal burn down chart without context. What is ideal to one person/team may be completely different to another.

I personally track my time spent in an IDE slinging code or in app testing using an external tool to help me identify patterns. Purely in terms of time spent in IDE/app, Fridays are my least productive day by about an hour. Mondays are my second least productive. I am at my peak on Wednesdays. Thus I try to shuffle heavy loads to mid-week and the easy, gimme-task for Mon/Fri.

Just like a burn down chart, there isn't some ideal way productive time spent at work should appear on a graph. Well, I guess ideal is 100% of the day in IDE/app , but productivity would plummet. Humans just are not that efficient. Graphing my time gave me insights on how my week/month plays out. Burn down charts should do the same at a team level.


Actually it is a good metric. However the metric is not a linear shape, but within 20% of linear. If you are not within 20% you need to do something about it. In general once you are within that further effort towards perfection is not desirable.

Remember the real goal: money. Sometimes it is money from sales, sometimes it is money saved. (even in case of a charity where there are higher goals money is a proxy for the real goal since it can be applied to the goal in some other way). If management can predict with reasonable accuracy when features will be done they can translate that into how much it will cost. Then they compare cost vs expected rewards (expected rewards is the job of marketing) and decide if they should focus on feature A, B, or both.

Note that many managers fail to understand error bars. There is no way to know exact numbers. However you can predict your likely error, and if the error is too high you can spend more money to reduce the error.

I recommend the book "how to measure anything" for more detail.

In the mean time when management wants perfect linear burn down charts, there is only one way to achieve them: overestimate your stories, finish them early and then go home. If you are paid for a 40 hour week you should average about 15 hours a week, but once in a while you will need to work a 40 hour week (60 hours every 10 years or so). Most management considers this unreasonable (for obvious reasons), but if a perfect linear slope actually is that important to them they will agree.


Sounds about like my work habits.


This whole time I thought it meant how long your money has going to last.


Are you thinking of the term "burn rate", perhaps? https://en.wikipedia.org/wiki/Burn_rate


I think so it was from the show Silicon Valley. I know way to base your facts on a show. But I did hear that the show had at least one real engineer advisor.

Even so with regard to work vs. time what determines the time? Launch date?


This seems to have triggered the most cynical and disillusioned thread I've ever read on HN, and there have been a couple. Well done. It would be interesting to fish out a few other HN reviews of business practices, but I don't even know the correct terms to search as they generate instant disinterest.


It's a project management strategy named for how it makes you want to burn down the office.


Nice excel graph with the number of "points" burned each sprints.


The unhealthy fascination with burn down charts and Scrum in general continually amazes me. I have witnessed the same debates about what Scrum points system to use so many times and witnessed teams of intelligent people waste days of developer time earnestly discussing relative point values.


Oh theory... may it die a timely death


The trick is to retrospectively rescale the work estimates.


Did you try reverse engineering the burndown?




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact

Search: