The tricky thing here is that once a system gets automated it’s all too easy to kind of let it keep running, and there’s no better way to lose money quickly than to have a bad decision algorithm keep running in an automated fashion. So I think it’s very critical for managers to understand what’s behind these things—what is the underlying logic. All of these systems and these algorithms make assumptions about the world. In the case of the financial crisis the assumption that was violated was that housing prices would continue to go up and people with subprime mortgages would continue to be able to refinance or pay them back. And when you see the world changing in some critical way like that, that’s the time to step in and say “you know, our models aren’t fitting anymore.” Now there are semi-automated tools that will let you look at your predictive models and say, “the fit isn’t quite what it used to be,” and kind of let you know that your fit is decreasing over time—the fit of your models to the actual data. But that is not true for rule based systems. It’s only true for algorithmic based systems, and even those systems are not very widely used. A few banks have applied them, but it requires a level of sophistication and caution in the use of these that most organisations haven’t really mastered.
That’s Tom Davenport in a podcast interview with Frank Comis at the McKinsey Quarterly (requires free registration). The interview and its accompanying article, ‘Rethinking Knowledge Work: A Strategic Approach’, also by Davenport and discussed in a previous Analyst First post, comprise Week 1, Day 5 of the CORTEX MBAnalytics program. The core of his argument is that the free-access model—which has for the last two decades allowed high-end knowledge workers the freedom to use their mobile phones, surf the internet, and do what they like with office productivity software—has not increased these workers’ productivity. According to some studies, productivity has in fact declined. Productivity gains have, however, been realised over this same time period through the deployment of structured-provisioning systems to low-end knowledge workers (case management, workflow, document management, business process management, and so on). Davenport summarises that we have, if anything ‘over-automated at the low end and under-automated at the high end.’
His conclusion is that more “structure” needs to be integrated into high-end knowledge work, but in advocating this he’s acutely aware of the challenges it introduces: worker resistance, demotivation, and reduced loyalty; performance measurement and management ambiguities; and the hazards of automation (on which he is elaborating in the quotation above).
I argued in the previous post that Davenport’s ‘Rethinking Knowledge Work’ represented a change of direction from his landmark ‘Competing on Analytics‘ body of work. Having now listened to the podcast, I’ve changed my mind. I think he’s actually asking the logical next question: Why are so few organisations competing on analytics? Davenport has spent years studying and understanding how analytics is being used by a handful of firms to carve out new markets and disrupt existing ones. He’s now attempting to reconcile that demonstrated potential with the overwhelming lack of productivity across the vast majority of the “knowledge worker” cohort, of which he himself is a member. Instituting value extraction from analytics, as he says “requires a level of sophistication and caution… that most organisations haven’t really mastered.”
Related Analyst First posts:
About usAnalyst First is a new approach to analytics, where tools take a far less important place than the people who perform, manage, request and envision analytics, while analytics is seen as a non-repetitive, exploratory and creative process where the outcome is not known at the start, and only a fraction of efforts are expected to result in success. This is in contrast with a common perception of analytics as IT and process.
Tags in a CloudAIPIO analyst first Analyst First Chapters analytics analytics is not IT arms race environments big data business analytics business intelligence cargo cults collective forecasting commodity and open source tools complexity data decision automation decision support educated buyer EMC-greenplum forecasting HBR holy trinity human infrastructure incentives intelligence model of analytics investing in data lean startup literacy management culture MBAnalytics operational analytics organisational-political considerations Philip Russom Philip Tetlock prediction markets presales R Robin Hanson Strategic Analytics tacit data TDWI Tom Davenport uncertainty uneducated buyer vendors why analyst first