Why a BI mindset can be BS

“All of your data, under one roof.”

So good in theory, but in reality, the marketing promise has written checks the hiney is unable to cash.

Data centralization is something that people have been working on for 50+ years. Our digital data warehouses are only one of the more current attempts at the myriad ways businesses have tried to get it all in the same place so we can see the whole picture. And you know how much of a struggle this effort has been for us in digital? It’s always been that way. It’s actually always been worse.

Unfortunately, the promise that data centralization makes will never be realized. The promise is that centralization will get us away from the clutter of decentralized data and give us one place to build from. I don’t think we will ever get there.

For one thing, the data fights centralization. In all fairness, it is very, very hard to centralize data and link information from our various systems (CRM, point of sale, web analytics, web marketing/advertising channels, call center, etc.) And once the data is linked, it’s incredibly hard to collapse trees of data where metrics don’t match up (web uniques, customerIDs, credit card #’s, email addresses, invoice #’s, TV and radio performance data, coupon redemption, finance metrics, etc.).

This is HARD. Being a CIO would suck.

The reason I don’t think we will get there is because today, linking data and creating models almost always means sacrificing fidelity. Take cross-channel attribution, for example. The problem is that revenue is duplicated across the reporting delivered at the channel level: email reports $10,000 and search reports $10,000, but we only sold $15,000. Many tag management solutions today “fix” that problem by choosing which of the tags to fire, excluding the other tags. But is this a fix at all? We just lost fidelity into the revenue that each channel participates in. If two different agencies manage this media, they just lost some ability to understand how their media influences revenue; they lost some ability to optimize and boost results, because now they are only working with part of the story.

These tools fail to take outside influences into consideration like the creative, changes in budget, visitor history, etc. They teach marketers to assume that media channels intrinsically are “closer” or “further” from the purchase, or more/less related, and that this will be consistent over time. In other words, they suggest that because email was an early touch and search was a late touch last month, it will continue to be that way in the future and that’s how you should optimize the campaigns. Hogwash.

Yes, it is true that guns don’t kill people and tools don’t tell stories; people do. But when we create technologies that “fix” problems, we are stating that what comes out of the tool represents the truth, the whole story, or something like that.

The promise of BI and data centralization is that when information is married and central, we get a better picture of what’s happening and can make better decisions. It’s this fallacy that has led your company’s management to make decisions that make your head spin. It’s that false promise that leads businesses to invest millions in a solution that blurs the focus and liposuctions critical details that will influence decision making.

Am I saying that progress hasn’t been made? No. Immense progress has been made, and data centralization is a hugely important thing. But it isn’t the answer. It’s one of two barrels we need to employ to understand what is happening. The other is the specialist view. The highly trained and sometimes myopic perspective that crafts per-channel, per-market, per-issue solutions. Specialists on their own are highly dangerous. Generalists on their own are, too. It’s only when both fire together and you can take superb, but realistic solutions to market that you win.

This is just a, “where do you spend your time,” idea. If you are investing an inordinate amount of your attention into centralizing, de-duping, common-keying, etc., at the cost of per-channel, per-market, per-idea optimization, I believe this isn’t the best use of your time, as a ninja analyst.

It’s your choice how to tackle optimization. You can start at the central: Omniture Genesis, multi-channel attribution, etc., but you MUST go to the individual branches. You can’t make decisions from the trunk. Or you can start at the branches to find opportunities, but you MUST look at the trunk to make sure that growing one branch won’t tip the tree over.

“Simplification” through centralization is an illusion. What we do is hard, and it’s going to stay that way. It’s just a decision on how we’re going to spend our time.

No Trackbacks


  1. Story of my life. Well said. I especially feel bad with those in the UK who have set their primary key in their BI tool to IP address. I think there is more truth to this in the agency world than on client-side, though. While there will always be a certain degree of variability in any company, there is so much more volatility when an external client is calling the shots (and let’s assume your managers aren’t pros at pushing back on bad requests). The bottom line is that BI tools will never be a silver bullet. You’re always engaging in a long-term relationship with whatever platform you use to “centralize” your data. It requires an analyst with hours, a DBA with hours, and some kind of manager with hours… at LEAST. Then you have to have the infrastructure to host it – and if you don’t, there’s an additional operating cost to host it on a cloud.

    3 FTE’s, 1 BI tool, 1 cloud server, and countless hours of training later, you have your BI tool ready to show to the client! By then, you have a backlog of work that you need completed ASAP because so much time has been spent starting this up, you’ve missed 3 or 4 requirements/prioritization meetings, the 3 FTE’s finally get the first report up and it has been 2 months and it’s now barely relevant. You present it to the client, get feedback, and schedule a followup design session. Your analyst finally finishes the dashboard and it’s fine – but we need to ensure half the company has permissions to the report. 6 hours of intermittent, disruptive emailing later and you finished the permissions list. Not just that, but you have created a file that you can paste into other reports! … Well, that list is only good for reports that have a certain prefix. Oh, you didn’t prefix? We need to create a naming convention. Oh, we also need to develop a process to submit requests because we have too many in the pipeline with no context. Hmm… Well, I guess we’ll put report #2 on hold.

    Posted July 25, 2011 at 8:36 pm | Permalink
  2. I’m not sure exactly why but this weblog is loading very slow for me. Is anyone else having this issue or is it a issue on my end? I’ll check
    back later and see if the problem still exists.

    Posted January 9, 2013 at 1:58 am | Permalink