The metric / visualization pirate 'rules'

Posted by Marcus Hammarberg on October 15, 2015

At my current client, the hospital RS Bungsu in Bandung, we are working a lot with metrics and visualizations. This has proven to have a profound effect in interest, focus and understanding for the entire staff.

During the last years a couple of “rules” (I’ll come back to the quote marks soon) has established itself. Based on our experience and feedback we have come up with a few guidelines that helps us to do better metrics and visualizations of them.

In this post I wanted to share our current set of “rules”. Hopefully this will be helpful to you and maybe even you can share your guidelines with me.

“This works for us” (™), “Your millage might vary” (™) and “Please see past the practice and look for principle that you might find useful in your setting” (™). All of those applies to this post.

These are more like guidelines

The Pirates of the Caribbean is a really great movie, I think. The other installments not so much, but the first (Curse of the Black Pearl) rocks.

There’s an very funny and recurring quote about The Pirate Code. Different people comes back to the importance of following The Pirate Code. But they seem to be very flexible in interpretation. The pirate Captain Barbossa explains it like this:

The code is more what you'd call "guidelines" than actual rules

We use our “rules” in the same way. If there’s a good reason to break them, that we can explain and justify, we’ll break’em. These are more guidelines than rules. Hence the quotation.

The “rules”

  • Visible for all, aka big
  • Understandable by all, aka easy
  • Focus on learning, aka trends
  • Has a least one target, so we know where we are going, aka has target
  • Changes often, at least once a week, aka moves fast

These rules are meant to bootstrap us to succeed. Or “Fall into the pit of success” as Microsoft put it a couple of years ago.

Let me make a few short comments about them here.

Visible for all

First and foremost we strive to make the metrics transparent. As in “show them to all the staff every morning”. This is a big thing in Indonesia, I can tell you. Quite the opposite from most management recommendation and trends. But it works for us.

If we’re going to show them to the staff they have to be visual. The diagram (above) we used up to now was 70 cm x 100 cm. We just taken that one down to fit more than one metric (another post…) and the diagrams is 50 cm wide. Below.

We put only one metric per diagram in (so far), to hinder cluttering

Our goal is that you should be able to tell me what the data explains from the back of the room. You do not have to be able to read every detail, but stuff like; are we above or below target, what is the overall trend etc. If not - the visualization is to small.

Optional: physical

By the way, if it’s going to be big we’ve found that it’s better to make it physical. There’s also something with the tangible nature of a diagram on a white-board that makes it more real than an Excel diagram on a projected screen.

For example; we update the diagrams, by hand, every morning. For good days, there’s cheers and rejoicing. For bad days… a more solemn, reflective mood :)

Understandable by all

Showing the data is not enough, the data has to be simple enough that everyone understands it. Showing data that people can’t make sense of is pretty useless.

For example in our hospital we have shown how the hospital is doing financially. This is abstract and pretty uninteresting information for most staff, actually. We also noticed this since at one point in time the hospital was really struggling and still no one reacted. There were yawns, looking at the phones etc. when the catastrophic data was revealed.

Therefore I tried an idea; by using average prices per patients and the data for our costs per month we converted this data in to “Target in number of patients served per day”.

To my surprise this didn’t have a great effect either. Most people still seemed uninterested even though our target was 134 and we were averaging 74 patients served per day. Just above 50% of our target. Really bad in other words

When I saw this I asked my colleague, that helped me with the finance behind the calculation:

134 that's our target - including making a little profit, right? What is the Break Even point?

It was 120.

So I added a new line (see the diagram to the left) and called it “Loose money line”, a more provoking title for Break Even point.

And then I said to everyone:

Everyday we are below this line the hospital lost money by being open

BOM! I’ve never seen a more collective jaw-drop in my life. Everyone got that.

Therefore we from then strive to make the visualizations of our metric in a way that is easy to understand for everyone. Our target is that everyone in the hospital should be able to tell us what the data is about and how we are doing.

Focus on learning and trends

All our metrics are to learn about how we, as a hospital, are doing. We are not interested in performance indicators for individuals.

This also means that individual numbers are not that interesting for us, we focus merely on the trends. For example; if our goal is to try to increase number of patient served per day, we will just take a step back, squint our eyes and see if the trends is pointing upwards or not. If we can’t see that we call it “flat” and need to continue to work towards the target.

An easy way to do that is to take away the outliers. In the diagram below there’s a few dips. Those are Sundays. Very few people goes to hospital on Sundays in Indonesia (and the rest of the world?). We have called those “Out of reach”, and simply don’t care about it when we analyze the trend.

Oooor… this might be a ripe opportunity for improvement. If we could become “that hospital that has excellent care on Sundays”… but now we’re leaving visualization country and entering management / marketing territory. I’ll leave that up to them.

Has a least one target

We’ve found it very important to have target, a goal for our data. Without it the data doesn’t make much sense to the people we are describing it to;

Oh! Number of Operations per day was 6 yesterday, average seems like 4 and trend is pointing up. Is that good?

Well it’s really good numbers if the goal was 5. But it sucks if the goal is 25.

By visualizing a target (often just a simple line) on the board it’s much easier to understand and make decisions based on the information.

Changes often

I’ve blogged about this one already so I’ll leave this short. We prefer data that is “too simple” but moves often rather than “very accurate” and updates seldom.

Here’s a few examples to clarify what I mean:

  • We think it’s better to ask many patients What do you like our hospital, graded 1-4 and get a daily score, than to do deep interviews that we get the result from every week or month.
  • We think it’s better to count the number of patients per day, knowing that they all represents average of average, than to get a exact financial report in the end of the month
  • We rather count the number of new patients per day as an indicator of marketing improvements than see an increase of the financial bottom line at the end of each month

Note that this goes for our visualized metrics. We still compare and follow up slower running metrics too, but that for another reason.

The reason we strive for quick moving metrics, even though they might be simplified, is that it makes the data interesting and relevant. If the data moves once per month people loose interest.


Using these “rules” we have found that the visualized metrics we are using becomes much more relevant, useful and interesting to everyone that we show it too.

Underlying which metrics we choose lies a strive to only do things that will make an impact on the things that we are tracking. That’s another blog post, but basically we ask (thank you Dr Hubbard):

What decisions are we going to make based on these data?

And then, once we’ve decided what to track:

What do we think will happen with the XXX line if we take this action?

These guidelines has helped us a lot. I hope that you find some of this useful.

Published by Marcus Hammarberg on Last updated