July 17, 2011
Badly targeted ads are worse than untargeted ones
The recent launch of Google plus has got me musing on targeted advertising again. I quit Facebook a while back, as the whole experience had just become too annoying. Part of that was the ghastly spambot games like farmville, but part was the dreadful advertising that was becoming more and more intrusive. Social networking sites appeal to advertisers because you can precisely target your market, and hopefully get a better click through rate; but I struggle to imagine adverts less likely to interest me than the ones that Facebook showed me. Diet pills? Tooth whiteners? Plastic surgery? A “free sparkly pink princess iPhone” for 26 year old women who click this button? I think not. Patronising me is not generally a good way to get me to part with my hard earned pounds – or make me return to the site where the adverts are. In my case, the advertisers wasted their money, and Facebook lost a user through inflicting inappropriate advertising on them. I’m now encouraging all of my friends to move to Google plus, which harms both Facebook and the companies who advertise there. If they had shown me ads for, say, a subscription to the Harvard Business Review it could have been a very different story!
I guess the moral of the tale is that an immature targeting model can be actively harmful. Companies need to target advertising very, very carefully (predictive analytics tempered with common sense is usually the best way to do this) and check what kind of adverts their intended market are already seeing – and clicking on. If it’s a sparkly pink princess anything… Well, you won’t find me there! Lets hope that when Google plus introduces advertising they find a way to help advertisers do a better job.
May 18, 2011
May AIM meeting – SAP Data Services
I’m off to the monthly meeting of the Avon IM group tomorrow – this month Richard Munn will be giving a presentation on SAP Data Services – The future of large-scale Enterprise ETL? And we’ll be ensconced in the Lonsdale room at the Bath Royal Literary and Scientific Institute admiring the dinosaur bones from about 7. Hope to see you there!
April 21, 2011
Unsupported Activity
So I mentioned last time that I was going to try running an SSRS report from data processed by an SSIS package…. It turns out that this is unsupported by Microsoft for a very good reason. I made the configuration changes, created my SSIS package, added it to my report….. And BIDs crashed as soon as I tried to view it. It crashed when I tried to run any of my old (unchanged) reports too. I undid the configuration changes, and about half of my reports still crashed. It took a full re-install of SQL Server to sort it out. *sigh* There’s a moral in there somewhere…..
April 10, 2011
SQLBits – the aftermath
Just got back from SQLBits – it really was a brilliant conference, so thanks to everyone who organised it, and all the speakers who generously shared their knowledge. I learned something from almost all of the talks, but there were some particular highlights…. Marco Russo made an incredibly compelling case for the BI Semantic Model (BISM) which is going to form a major part of the next release of analysis services, and I think I’m going to need to start learning DAX in earnest! His talk on SSAS best practice was also really useful. Jess Meat’s talk on Microsoft BI was good fun too, and I didn’t realise that Visio 2010 can connect up to data and be used to build simple dashboards! I’m going to have to have a play with that when I’m back in the office. Allan Mitchell also gave a super talk on StreamInsight – it’s a technology with such potential, I’m sure it’s going to be re-writing the rulebook on real-time dashboarding very shortly. The schedule was pretty packed, so I couldn’t go to all of the talks I wanted to see….. But thankfully the kind organisers will be putting videos of all of the talks up on their website for free!
I met up with my favourite DBA Richard Munn (perhaps better known as SQLMunkee!) while I was there too, and he has inadvertently given me an idea to make a particularly troublesome report run faster – I’m going to try pulling the data into the report by way of an SSIS package! That’s the great thing about these community events – everyone there has interesting tricks and valuable experience to share, and even going to the bar is an educational experience! I’ll let you know what happens to my report….
April 7, 2011
SQLBits – oh I do like to be beside the seaside….
Especially if it’s a conference room beside the seaside…
I’m off to SQLBits again this year – should be good fun. I’m particularly looking forward to Jamie Thompson and Chris Webb’s sessions, and I’ll definitely be going to see Jess Meat’s session on “Microsoft BI – More than just SQL”… It sounds very like the talk that James Adman and I gave last month at the AIM group! It’ll be interesting to hear a fresh perspective on that, and hopefully I’ll come back with lots of new and exciting ideas 🙂
November 4, 2010
When quality is a dirty word
I’ve just had a lightbulb moment on – of all riveting subjects – software quality. I’ve always hated lengthy quality plans that no-one reads (least of all the customer) and testing that focuses on producing spreadsheets “showing” how well the software has been checked, rather than actually making sure the users can use the system. However, until now I’ve accepted the premise that – however tedious it might be – this was the way to get high quality software. I’m reading a book by Alan Weiss, and he makes the excellent point that quality isn’t the absence of defects, it’s the presence of value! It is totally irrelevant what the software designers and developers think is quality, what matters is what the client thinks – and all they care about is the value added, software that does what they need it to, not code coverage or specs that rival war and peace. Most software quality initiatives are completely backwards. They focus inwards instead of outwards – mandating twenty levels of integration testing and 98% code coverage instead of demos, delivering early versions for feedback and measuring customer satisfaction. How stupid of me not to realise this before! Obviously buggy software is never going to impress a client, but I feel like I’ve been missing a really big part of the picture.
Now I’m going to go and try to find out what criteria our customers judge our software by. And then i’m going to have an argument with our quality department….
October 2, 2010
Complexity – that which does not kill you…
The many headed hydra of complexity is usually seen as a thing to minimize or – ideally – avoid completely in business generally and in business intelligence systems in particular. This aversion to hard problems means that there is money to be made by the companies who can face up to the challenge….
Historically businesses have tried to deal with complexity by reducing it – offering a suite of standard products instead of a bespoke system, locking in customers, merging with competitors…. However, in our increasingly connected world this is becoming less and less possible. Barriers to entry are dropping for just about every kind of business all over the planet, customers are searching more widely for organizations that offer exactly the services they want at the best possible price, and the companies that can’t keep up with these demands are dropping by the wayside. To quote John Naughton, the old strategies “are unlikely to work in our emerging environment, where intelligence, agility, responsiveness and a willingness to experiment (and fail) provide better strategies for dealing with what the networked environment will throw at you.”
In terms of BI, to keep the business competitive some business users need to ask very complicated questions, to have the freedom to follow where the data leads, and those questions may change on a daily or weekly basis. And if the IT department can’t supply what they need, the business users will try to build their own systems in spreadsheets…. It’s quick and flexible and you can have as many copies of the spreadsheet as you like all doing slightly different things and it can link to other spreadsheets and no-one can remember how it works and the data is different in different copies and there’s no way to know which one of them is right any more… and down that road lies madness.
These kind of requirements really don’t suit traditional IT development with it’s careful upfront requirements spec; flat, static reports and long development time. Agile development – with it’s small, regular, fast deliveries – of a data warehouse and self service BI through a semantic layer like Business Objects suits them much better. This kind of project needs good integration between the IT team and the business users though, which may be a challenge in itself. Still, new, interesting, and potentially profitable ideas are likely to appear if the two sides can be persuaded to work together.
The complicated questions are likely to involve some pretty in depth analysis, and really, the business users don’t want to be chasing after the IT department every time they want to experiment with a new calculation. Cubes delivered through something like SQL Server Analysis Services or power pivot give control back to the users through the familiar interface of excel – without the danger of a spreadsheet daisy chain or many corrupted copies of the data. They could even be published on to share point for the rest of the team to share, and although they’re unlikely to be the sort of polished report you’d want to show the CEO, he’ll be more impressed by their impact on the bottom line.
So sometimes complexity can be a business benefit – for anyone who’s brave enough to grasp the nettle.
Some references that I found useful when writing this:
James Adman on Drowning in Spreadsheets: http://www.b-eye-network.co.uk/view/11482
John Naughton on how the Internet is changing the world around us: http://www.guardian.co.uk/technology/2010/jun/20/internet-everything-need-to-know