When Information is NOT the Answer

by Andrew McAfee on July 20, 2009

My friend Don Sull and I met in HBS’s doctoral program, which we both slogged through in the mid 1990s. He’s now cranking out mounds of good work at London Business School, and also blogging for the Financial Times.

His current work concentrates on helping companies navigate their increasingly turbulent competitive environments, and his most recent blog posts discuss how IT can and should help with this task. I’ve written about this topic in an Harvard Business Review article last summer and a couple blog posts.

I’m thrilled to see a general management scholar of Sull’s caliber pay serious attention to technology issues. He’s encouraging his executive readers to take IT seriously, and offering excellent them advice. Please keep it up, Don.

In his recent posts on IT for execution, Sull concentrates on IT’s ability to provide needed information to business decision makers. As he writes:

Chief Information Officers, Finance Directors, CEOs and outside directors should be asking, and answering, a… fundamental set of questions: What type of supporting data do we need to make sense of a rapidly changing market? What other information is required to support execution? What organizational, behavioral, and cultural changes will we need to capture the benefits of improved information?

He also kindly cites my work on the incredibly agile (and hence wildly successful) Spanish retailer Inditex, parent company of the worldwide clothing chain Zara. Sull writes:

When companies start with the question of what data do we need to execute effectively, they can achieve a great deal without massive investments in IT. Consider Zara. The Spanish retailer surpassed the Gap in 2008 as the world’s largest fashion retailer. Zara leads the world in “fast fashion” a retail category pioneered by European companies including Sweden’s H&M and Britain’s Topshop. These companies track fashion globally, spot emerging trends, and translate them into new products. Zara can move a product from design table to store rack in three weeks. Zara, like other fast fashion retailers, succeeds or fails based on the quality of their market data.

Sull stresses that “Zara’s business model demands good information,” which is certainly true. But my work with the company (see this Sloan Management Review article and this case study) revealed something I found fascinating: Zara succeeds in large part because the company makes comparatively light use of market data and sales information, at least as these terms are commonly understood in the retailing industry.

The decisions about which clothes should to go which stores at what time(s) are probably the most important decisions made by any large apparel retailer. Most chains make them by collecting large amounts of daily sales data from stores, combining it with other hopefully relevant information, then applying a variety of statistical techniques to generate a forecast —  a quantitative prediction about what will sell. This forecast is used to push the ‘right’ items —  the ones predicted to sell — over time to each store.

Each retailer forecasts differently, of course, but I find their techniques broadly similar: they all gather lots of data, analyze it centrally, then use the resulting predictions to determine shipments to stores. In this model, the stores themselves have fairly limited roles: they are expected to record data accurately and send it promptly, then do their best to sell whatever headquarters decides to send them.

This seems sensible enough, and it also seems logical that as the business world gets more and more turbulent more and more supporting data will be required. This data will need to be acquired, analyzed, shared, and interpreted with ever-greater velocity, requiring ever-bigger computers, ever-faster networks, and ever-more-quantitative decision makers.

But Zara, operating in an intensely turbulent environment, does something totally different. The company doesn’t really generate a store-level sales forecast at all. Instead, it relies on its store managers to tell headquarters what they think they could sell immediately at their locations. Headquarters then gets as many of these clothes as possible to the stores as quickly as possible.

What’s more, the store managers are given very few quantitative or analytical tools to help them make their short-term predictions. They rely largely on intuition and experience, on walking the floor and talking to customers and employees.

Information technology is still critically important at Zara. The company uses technology to present store managers with a multimedia digital order form, and to transmit completed forms back to headquarters. IT is used heavily to support execution, in short, but not at all to assist with data-based analysis or decision making about getting the right clothes into stores at the right time.

Zara is obsessed with making good decisions about what clothes to stock, but has configured itself so that people making these decisions operate in what looks like a ‘data vacuum’ – a lack of aggregated, filtered, and massaged information from throughout the corporation. This is because the good information that Zara’s business model requires is not the kind that’s easy to digitally encode, transmit, aggregate, and analyze. Instead, it’s information that comes from watching, talking, and listening, then using the computer between our ears to pattern match, draw conclusions, and peer just a little bit into the future.

As I wrote here, Zara believes that the relevant knowledge for fast fashion forecasting isn’t general knowledge (the kind that can be digitized), it’s specific knowledge (the kind that can’t). Three critical business design considerations flow from this belief. First, Zara spends almost no time on store-level sales forecasting and other similar kinds of data analysis. Second, it has moved decision making down very low in the organization, because this is where the relevant knowledge is. And third, it gives these decision makers very little market data or other forms of general knowledge.

So in addition to the IT-related questions Sull lists in the quote at the top of this post, I’d add one more:

For this decision, what’s the relevant knowledge?  What’s the mix of specific knowledge and general knowledge required to make this decision well?

If this question is not asked, the danger is that executives will assume that all or most of the relevant knowledge will be general knowledge, and will therefore get to work digitizing, analyzing, and sharing it. Zara’s success shows how beneficial it can be to question this assumption.

I want to be clear: I believe that in many if not most situations business decision makers are well-served by the kinds of information served up by computers. But in at least some cases they’re not. Companies that do a good job of figuring out which cases these are will have an edge over those making the blanket assumption that more turbulent times call for greater reliance on data.

Have you seen other circumstances where classic market data is just not that useful?  Leave a comment, please, and tell us about them.

  • http://www.pretzellogic.org Sameer

    Another use case, one that I'm currently consulting on, is localized CRM for franchise based businesses. Relying only on a centralized marketing campaign design and ad spend campaign for 1000s of stores for a franchise such as the local Super Cuts or Jiffy Lube can be pretty inefficient.
    There's power in local input from a store (as Zara does) but also in collaboration networks amongst clusters of local stores to inform marketing spend. This reduces ramp up time for new stores, prevents duplicate marketing efforts and allows for volume ad spend buying at the local level (amongst many other benefits)

  • http://www.bennettstrategy.com/ John Bennett

    Great post. Is the title misleading, though? The store managers are still sending “information” to headquarters. It’s just not statistical data or the type of information associated with BI.

    Incidentally, how do they convey this information about what’s hot and what’s not? Phone conversations about styles and colors? Or is that where the sophisticated ordering system comes in? Do all the stores share a fashion vocabulary? Does the company train its store managers much, or does it really treat them as a diverse crowd that in effect crowd-sources a solution for inventory?

  • http://lifesays.com/ Sridhar

    To make a decision and make it well, facts and opinions are both needed. They move from being data to information to knowledge. That is the reason why it is called “Information Technology” and not “Knowledge Technology”. Computers can only transform data into information – but it is our experience and knowledge that can transform it into something useful and actionable.

    Thanks for the great post.

  • http://www.ribbonfarm.com Venkat

    This is a view I strongly resonate with, and I am perhaps more extreme than you. There is work in cognitive science (eg. Gerd Gigerenzer, “Gut Feelings” http://www.amazon.com/Gut-Feelings-Intelligence…) that highlights how too much information can make decision-making worse. Even “relevance filtration” etc. are suspect strategies, because relevance criteria come from mental models that in turn represent preconceived patterns of thought.

    So there is a strong trade-off in being a data-driven decision-maker and a strong instinctive decisionmaker. The right way to break this tradeoff is to aim to deliberately present limited information in somewhat disorganized/random ways so that there is room for some creative right-brained pattern-recognition to happen. The disorganization serves to scramble any preconceived models that might be loaded in your RAM.

    A simple example is the presentation of a table. Size is not the only variable (depending on the context, you may need 7 or 7 million…). Ordering is important too. Any rational sort on some column biases the mind to think in certain ways. There are advantages to random or otherwise arbitrary (eg. lexical order) presentation (which is why browsing libraries is still more fun than following an Amazon recommendation trail. Libraries sometimes juxtapose things in interesting random ways).

  • http://twitter.com/teisenmann Tom Eisenmann

    Great post! Citing Don Sull's recent FT posts reminded me of a Harvard Business Review article, Strategy as Simple Rules, that he wrote with Kathy Eisenhardt (Jan. 2001). They argued that in turbulent environments, companies should formulate strategy using a few simple rules, rather than relying on comprehensive planning processes. They don't say whether firms that rely on simple rules tend to make relatively light use of market data, as in your Zara example, but the converse is certainly true: comprehensive planning clearly requires LOTS of data. It'd be interesting to explore whether Zara uses simple rules and whether we can generalize about the type of IT capability that best supports the use of simple rules.

  • http://caddellinsightgroup.com jmcaddell

    Still thinking about this post, because it raises a lot of interesting questions, but I'd take issue with the way you say one thing about Zara: “It has moved decisionmaking very low in the organization.” I might rephrase that by writing “Zara has moved decisionmaking out near the customer interface.” The most relevant quality of the Zara decisionmakers is not their place in the hierarchy, it is their proximity to the customers, their habits, stories and desires.

    I do think there are ways to harness this “specific” data, share it and even make collaborative decisions with it. There's a middle ground between huge objective numbercrunching and improvisational decisions out in the field, and in that middle ground (which is very untended today) is a lot of insight that can help both executives and first-line managers.

  • http://caddellinsightgroup.com jmcaddell

    Still thinking about this post, because it raises a lot of interesting questions, but I'd take issue with the way you say one thing about Zara: “It has moved decisionmaking very low in the organization.” I might rephrase that by writing “Zara has moved decisionmaking out near the customer interface.” The most relevant quality of the Zara decisionmakers is not their place in the hierarchy, it is their proximity to the customers, their habits, stories and desires.

    I do think there are ways to harness this “specific” data, share it and even make collaborative decisions with it. There's a middle ground between huge objective numbercrunching and improvisational decisions out in the field, and in that middle ground (which is very untended today) is a lot of insight that can help both executives and first-line managers.

  • http://www.fastforwardblog.com/?author_name=pthornton rotkapchen

    “forecasting isn’t general knowledge (the kind that can be digitized), it’s specific knowledge (the kind that can’t)” Love this story. Thanks for this data point : )

    It reinforces one of the critical elements missing from most initiatives (where they 'do' make sense) = the real-world research to validate the assumptions.

    As to “classic market data” — that's a whole 'nuther can'o-worms. We've been having several in-depth conversations about most of the key metrics being used not being relevant to the specific business model (http://twurl.nl/2txzo0).

    Again, assessing the relevance is a matter of 'design' and the related research of what makes sense for the business to best preserve their lifeblood — connection to their customers.

  • http://www.silverstop.eu/ szampon dziegciowy

    You can’t always be right, and You can’t always predict consumer reaction. There were some cases that showed the computer analysys is not everything. E.g. white pepsi went dead after few months of bringing loses.

  • http://twitter.com/dagb dagblakstad

    My view may be biased towards “long tail thinking”, because I've just finished reading the book (late adopter). By not relaying on data (favoring the hits), Zara's way of manufacturing promotes clothes in the tail. Since most fashion clothes is only sold for a short period, it is crucial to be the first one to provide them. Some of Zara's offerings sure will move to the head, but many of them will probably not sell that much individually. By beating the competition, by relying on cheap and fast manufacturing they can afford to offer both tail- and head offerings. I guess this way of doing clothes business is much more interesting for both employees and their customers which can check out different clothes at each store.
    My daughter visited a couple of Zara stores, and she was very pleased with them.

  • rajatparwal

    Zara's case is good from one perspective that organisation has a culture of making decision at tactical levels. But, Zara might be trading off by taking skilful and costly managers. It is good to see their business model embracing customer's needs and customer's requirements at different levels. But, important question is- How long?

    I would not consider “Information” to be useless or not giving good answers. If sense making could answer all the problems why would new technologies be invented (concept of variety is important here: http://www.techrust.com/humane/us-and-machines ). Also, if being close to customer was the answer, smaller businesses would have flourished at faster pace. In fact I would disagree to the fact that Zara's logic of supplying material at fast pace can keep customers satisfied too long, as Customer themselves are very well informed about the products (so many market comparison sites have popped up in the internet space)! It's a classical situation in which DATA vaccum is working fine for NOW.

    One of the most slanderous argument made by Analysts: “Over-analysis leads to paralyses and sometimes information overload can be painful”. I would counter-argue these statements by asking again- WHAT was analysed, WHY was it analysed, WHEN was it analysed and HOW was it analysed. If the process design let the DATA pass all these filters, useful INFOMRATION was translated to higher levels, else, it was meaningless DATA (not Information). That's why I disagreed with the name of your post. I believe you, like other analysts, confused Information with data.

    Also, I would like to draw your attention towards “Concept of Strategy” at different levels. Zobaggy has pointed out what happens at different levels (http://www.hcss.nl/en/download/176/file/Militai…). When it is difficult to achieve Ends by the Means (available), different behaviours are observed at different levels, with increased level of complexities. What worked as 'simple rule' is again one of the states called “Strategy as Rules”, but there are even harder points, edges of Chaos, which needs a wider perception of understanding and deeper logic of problem solving.

    Arie de Geus, MD of Shell, talked about importance of Memory of future in his book- The Living Company. He said, “Managers can see the signals for change in time.” But I would add that managers also need “powerful tools” to SEE these signals from future and logically change it to meaningful information to act proactively in a pacefully changing environment. Pragmatic thinkers like DeBono, explained Seeing-Exploring-Judging as an important tool for decision making and I would say that information helps to take step1 and technology enables to store this information as memory of future. In the end it's the people of the organisation, who can make decisions and trade off before eliminating the options I have explained.

  • http://oatranslator.com/ Oscar

    Hi… on third paragraph / last line (as of 8:25 PST on Jul 29, 09):
    reads: “and offering excellent them advice”
    should read: “and offering them excellent advice”
    hope it helps (there may be more, but I stopped reading there)
    Oscar (oatranslator.com)

  • http://info-architecure.blogspot.com driessen

    Very interesting post, Andrew. I think we run into this daily. I just wanted to point to two books that underline the point you're making. One is 'Blink' by Gladwell. And the other is 'The Social Life of Information' by Seely Brown and Duguid. You've definitely read them. In 'The social life' they have that nice example of a control tower. The tower is fused with IT. But when they really tried to understand the PEOPLE that work in the tower and how they make decision, they arrived at interesting results. These people used sticky notes, easily processed loads of data, etc.
    And 'Blink' is loaded with examples of 'split second decisions'. Based on processing lots of information in the past, becoming expert at something. Gladwell shows that some can make good decisions even though all the (objective/general) information in the world seems to be against them.

  • staffing225

    Valuable information posted.
    For more jobs visit http://www.staffingpower.com

  • lihaoxj16

    tiffany jewelry
    Choose, buy and shop for on sale tiffany jewelry including Tiffany & Co Silver Necklace, Pendants, Bangles, Bracelets, Earrings, Rings and Accessories.
    tiffany co
    Tiffany Jewellery offering bangle Jewellery, bracelet jewelry, eardrop jewelry, necklace jewelry, ring jewelry, finger ring jewelry and earring jewelry
    tiffany
    tiffany and co
    links of london
    links london
    Tiffany Style Silver Jewelry: Rings, Earrings, Necklaces, Bracelets and more Tiffany Jewellery at low prices.

  • http://www.searchofficespace.com/uk/office-space/soho-serviced-offices.html Rent Offices Soho

    Thank you for this blog post. The commenter, then, becomes the critic who shapes the writer's next content. So I want to encourage you to keep on blogging in the same style as you writing style is really perfect.
    ________________________
    Liverpool Street Serviced Offices

  • http://pulse.yahoo.com/_NMRTQCJMDRPD4RISOHGFA23PWI A K

    I’m sure Zara has updated their POS terminals by now. The case studies you have listed are both dated 2004. I understand the point of those case studies were to tell us that success can be achieved by using IT effectively and not to over-use IT. But upgrades to IT is necessary today. Any idea if Zara has made any changes to their information system?

Previous post:

Next post: