Are you an Excel guru?

[vetran]

Remember the term ' decision support systems'? That word ' support' is important here. I don't like the idea of systems deciding on big things, but supporting decisions. I don't believe that ethical or emotional aspects of decision making can be put into algorithms right now, and even if they could, hmmm....difficult one that requires a lot of thought from people who are much brainier than your average IT manager.

Also, the word ' improve' is very contentious when it comes to systems that decide on something as emotive as asylum procedures or child protection; one person may have an entirely different concept of 'improve ' to another. Who's right?

yes but once a significant part of work is undertaken by the system the users who performed it manually are reassigned/ sacked/ replaced with cheap labour. You have one chance to get it right, if you don't then it spirals into a massive crater.

[Mich the Tester]

yes but once a significant part of work is undertaken by the system the users who performed it manually are reassigned/ sacked/ replaced with cheap labour. You have one chance to get it right, if you don't then it spirals into a massive crater.

The more complex the system, and I'd argue that almost any system is now part of a huge, emergent complex system, the harder it is to get it right first time, and the harder it is to judge what's right.

[Scrag Meister]

If you're on a 25 year mortgage, you may want to have another look at that.

mortgages!!

[Brussels Slumdog]

What happened to Lotus 123

I was a super Calc and a lotus 123 guru until Excel came on the market then I ended up on the bench as "Excel" did not appear in my CV
These days all I need to know about Excel is how to download a SAP report to an excel spreadsheet and perform lookups for data migration and calculating my expenses

[vetran]

The more complex the system, and I'd argue that almost any system is now part of a huge, emergent complex system, the harder it is to get it right first time, and the harder it is to judge what's right.

agree completely now how do you convince the business?

[Mich the Tester]

agree completely now how do you convince the business?

Get them to read some books they probably don't understand.

[bless 'em all]

The more complex the system, and I'd argue that almost any system is now part of a huge, emergent complex system, the harder it is to get it right first time, and the harder it is to judge what's right.

So a system with multiple inputs, calculations and outputs can become less accurate over time - particularly if there is a circular data route where the outputs from the system form a basis of subsequent inputs from other sources. How do you manage the 'drift' from having or generating truly accurate data to a point where your data is accurate to a +|- tolerance?

How would you even recognise an unintended diviation from a natural one in a system which might handle millions of records and tens of millions of transactions per year? It'd take some serious number crunching to separate these from the general 'noise' of d2d transactions.

[Mich the Tester]

So a system with multiple inputs, calculations and outputs can become less accurate over time - particularly if there is a circular data route where the outputs from the system form a basis of subsequent inputs from other sources. How do you manage the 'drift' from having or generating truly accurate data to a point where your data is accurate to a +|- tolerance?

How would you even recognise an unintended diviation from a natural one in a system which might handle millions of records and tens of millions of transactions per year? It'd take some serious number crunching to separate these from the general 'noise' of d2d transactions.

Indeed, standard approaches fail to work as things get more complex. you could use statistical analysis methods of running gazillions of transactions to analyse deviances, in combination with exploratory methods analysts set out to expressly create nasty problems while switching some connected applications on or off. Another method is to inject deliberate errors into code and then watch how data behaves in other parts of the system or other connected systems; this can gives a lot of insight into the potential flows of other errors, but it must be done in a controlled environment. PMs hate it but it can help find faults that really are in the design.

[NotAllThere]

I do SAP. There's only one version of the data.

Anyway, must crack on. Got to find out why the data in the global cubes doesn't reconcile with that in the regional cubes...

[Mich the Tester]

I do SAP. There's only one version of the data.

Anyway, must crack on. Got to find out why the data in the global cubes doesn't reconcile with that in the regional cubes...

Am I the only one who suspects a contradiction here?