Crash Test Dummies – Using realistic Test data

I was watching a programme about Crash Test Dummies yesterday. I’m on Holiday this week so I’ve been watching a lot of rubbish TV for the past week, so I thought that this was just another random programme that would keep me occupied for half an hour or so.

It didn’t click with me that this was a form of “Testing” until right at the very end of the programme until I the narrator said the following:

A dummy is just a dummy unless it provides information

What a great quote. This is also true of Software Testers. In my opinion, a Tester is just someone clicking buttons unless they provide information.

What fascinated me more about this programme was the history behind Crash Test Dummies. Before they were invented, they used to use Human volunteers, dead bodies and even dead animals! As morbid as this sounds, imagine that you have been assigned to this project and that this is your test data.

I feel quite sorry for those Human volunteers. I’ve been in a few car crashes and, trust me, I wouldn’t like to do this over and over again throughout the day! The trouble with using a Human is that I don’t think they could have possibly tested this to its limits as this would have surely meant killing the volunteer! So all that using a Human Volunteer proves is that, yes, it works between these set of variables, which we knew before hand as we didn’t want to kill him, but we have absolutely no idea what happens after that. When I test, and try and test the worst case scenario and use some extreme data just to see what happens. What would happen if you crashed the car at 100mph into a wall, for example.

Using a dead body would allow you to carry out those tests without fear of killing the driver, but I also don’t think that this would be an accurate test. Crash test dummies check for things like brain damage and head trauma by measuring the force of the impact, but this wouldn’t necessarily show using a cadaver. To me this is like using a good set of test data which lacks certain realistic elements. I suppose this would be like testing the system using valid, but inaccurate data. For example, using “abcdefghi” in an address field.

I find the idea of using a dead animal quite laughable. Surely all this proves is that crashing a car at a certain speed is fatal if it’s being driven by a Sheep? I couldn’t think of a suitable, real life testing example of this. What do you think? Leave a comment below.

Linking this to Software Testing, the more accurate and realistic you can make your test data, you’re more likely to find defects that occur out there in the real world once it’s gone live. An example of this is during the last big project I worked on. As this was a new system and their was no existing test data available, we had to make it up. We tried to make it as realistic as possible and had regular meetings to discuss the product and the requirements, so we had a fairly good idea of what the data should look like. After the first build, I would compare the data to like using a dead animal. We weren’t really sure that it was correct as we were still learning the software ourselves. After a month or so, our test data resembled something like the dead body. The data was fairly accurate and like how it would be used, but we knew it lacked certain realistic elements. It’s only now that we are in a position to get realistic test data, and I would compare our data now to using a human volunteer. In a another month or so, we will have a fully functional, super-duper crash test dummy of our own to play with!

10 thoughts on “Crash Test Dummies – Using realistic Test data

  1. I like the quote.

    I think what needs to be taken into account is what test data was available, the test data evolving and what you are attempting to learn.

    I agree that using a cadaver and using a test dummy will yield different information and they are designed to do so but I think that what information you are after needs to be considered as well and what you are testing. You could be testing seat belts, you could be testing air bags, you could be testing windscreens, etc. These will require different test data.

    If you are testing how windscreens shatter then a cadaver is probably better data then a test dummy. Shards of glass hitting a dummy in the face probably won’t give you much info but glass in a cadavers face will give you a lot.

    A dummy may provide figures (stats) on head trauma but if you can see your cadavers brains I’m going to go out on a limb (risk) and say that you really want to lessen the chances of that happening.

    The same with using a dead animal, a visible indication of damage to flesh. And also, people drive with their pets in the car so it’s a totally valid test.

    What would you consider realistic and unrealistic data? 9 people in a hatchback designed for 4 might be considered unrealistic but it happens, I know, I was a stupid teenager and I was 1 of the 9 many times.

    As far as I’m concerned anything goes, if it’s doable do it. Stick a dolphin in the driver seat and have at it.

    So Long, and Thanks for All the Fish

  2. Interesting blog post, that reminded me of one of my first testing jobs, using a cadaver. The project was to add a laser scan to remove any “movement” by a patient during an MRI Scan.
    I used to man the broom pole, and manipulate the cadaver in different ways during a scan, to see how well the software coped in removing the noose from the signal.
    The things you will do for rent when you are a poor student!


  3. “I find the idea of using a dead animal quite laughable. Surely all this proves is that crashing a car at a certain speed is fatal if it’s being driven by a Sheep? I couldn’t think of a suitable, real life testing example of this. What do you think? Leave a comment below.”

    In many cases using dead animals, live animals, or dead humans would be far preferable to a crash test dummy, and could provide extremely useful information.

    If you have chance, read “Stiff: The Curious Lives of Human Cadavers” by Mary Roach.

    Here’s an excerpt:
    “By and large, the dead aren’t very talented. They can’t play water polo, or lace up their boots, or maximize market share. They can’t tell a joke, and they can’t dance for beans. There is one thing dead people excel at. They’re very good at handling pain. For instance, UM 006. UM 006 is a cadaver who recently journeyed across Detroit from the University of Michigan to the bioengineering lab at Wayne State University. His job, which he will undertake at approximately 7 p.m. tonight, is to be hit in the shoulder with a linear impactor.”

  4. wat is tat pic supposed to mean. plus you spelled program rong its program. prograMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMM there is no “e” GET IT , NO EEEEEEEEEEEEEEEEEEEEEEEEEE JUST AN MMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMMM reply with correct spelling. You couldnt beat a kindegartner at a spelling bee. for that matter you couldnt beat a chicken.

    • Hi Felix.

      The picture in the post is of a Crash Test Dummy at a desk. I thought it was funny :)

      When referring to a Television programme, in English-English anyway, “Programme” is the correct spelling. When referring to an application, “Program” is the correct spelling.

      From memory, the American-English spelling for a Television programme is “Program”.

      I’m going to assume that you are a tester, so in future it might be worth keeping in mind that different countries may spell words differently to yourself. Another example I can think of is “Colour”, “Favour” and “Favourite”. Although these may look wrong to you (as you would spell them color, favor, and favorite), these are in fact the correct way to spell them according to the Oxford-Cambridge Dictionary which we adhere to in England.

      Hope this helps :)


    • Thank Jon. I’ve actually been looking for something like that for a while, so that has actually helped me out loads.

      In an application I’m testing at the minute, one of the comments from the test review was that my Test data needed to be more realistic in case it is shown to auditors.

      Thanks again


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>