The Zappos holacracy, Edward Tufte’s sparklines, and an 11×17 printer

by philg

Atlantic magazine’s October 2015 issue carries “Are Bosses Necessary?” One of America’s more useful business school professors is quoted:

What’s enabling this shift [to Zappos-style holacracy], argues Thomas Malone, a professor at MIT’s Sloan School of Management, is simple: falling information costs. In his 2004 book, The Future of Work, Malone broke the history of organizations into three stages. In stage one, information is expensive to convey, so most decisions are made face-to-face in necessarily small firms. As communication costs begin to fall, Malone explained to me, we reach stage two, in which “it becomes economically feasible to send information to a single, central place for decisions to be made.” That is, the large, centralized hierarchy becomes possible. Then comes stage three: “As communication costs continue to fall, there comes a time when it’s economically feasible to bring information to all points, so in some sense, everyone can know everything.” In this third stage, the benefits of bigness can persist, but its traditional handmaiden, hierarchy, doesn’t have to. (Indeed, when the volume of information grows large enough, trying to direct its flow upward for evaluation can slow everything down.)

[See Malone’s “Semi-Structured Messages” paper from 1986.]

I want to push one of my old ideas… combine Edward Tufte’s sparklines (see explanation in this 2006 posting) with a 1200 dpi 11×17 color laserprinter. Take every number that matters to an organization and print it out every day on an 11×17 sheet of paper that is given to every employee. For a web site it might be “new registrations yesterday” printed in green if it went up from the moving average of the previous week, printed in red if it went down. Next to the number would be a Tufte sparkline to put the number into context. The next line down could be “500 Server Errors,” followed by “Average pages per session,” etc.

What do readers who work in companies with at least 50 employees think? Would it be helpful to arrive at work and find an 11×17 sheet of paper across your keyboard with perhaps 150 metrics of the company’s performance?

18 thoughts on “The Zappos holacracy, Edward Tufte’s sparklines, and an 11×17 printer”

  1. Meh. I worked in a Fortune 500 manufacturer with a few dozen plants across North America. One of the plant managers publicly posted operating numbers in his plant – it didn’t make much difference in a mature industry with thin margins and little room for optimization. However, shared metrics to solve specific problems did produce good results (define, measure, analyze, improve, control). There was an attempt to compare operating numbers across plants, but you’d be surprised how hard it is to get disparate teams doing almost identical work to agree on common measurements.

  2. I like sparklines, I like your idea of a daily, easy to overview, concise graphic 1-page summary using sparklines (and generated automagically from an aggregating app that fetches up-to-date values from e.g. selected corporate spreadsheets and other internal and external sources… sounds like an advanced Excel macro job). But, despite a background in graphic production, I have difficulties visualizing such a Compact “How Are We Doing” report. Were I soliciting comments or otherwise attempting to sell that idea at large, I’d have started with a few mock-ups (hand-assembled & formatted samples) of such, rather than description of, to make sure that commenters are seeing exactly what you are thinking. How hard could it be?

  3. I find I my over 500 person company that this kind of metrics charts provides the ink blot for clueless management to argue about ot ignore without any real effort to understand. The DoD does a lot of these kind of program health or risk cube charts, but the brass still wants a narrative weekly summary (because the next level wants that too, so easier to cut and paste it than think). To stop that rant, your point was charts for a leaderless company, and it is definitely helpful to see important trends over time, but hard to use them for the future.

  5. John: Definitely off-topic but it involves Robinson R44 so it meets the moderation policy! Can I open a beer can with a helicopter skid? I don’t think so. Sometimes we practice hovering with the tip of a skid on the tip of a cone. One of the instructors at East Coast Aero Club is great at this. I would rate my ability as “fair/rocky.” It is not easy to divide one’s visual attention between the skid tip (necessary to hold position and/or open a bottle) and the horizon (necessary to hold a steady hover).

  6. I work for a 900-employee organization. As an MBA, I find my organization’s performance vs. competitors very interesting and helpful in performing my work duties. However, about 850 others at my workplace wouldn’t understand such metrics or care, and are not paid to do so; providing them with such information would be a waste of paper.

  7. My thought is that this would provide too many measures. I’d think that focusing on the “vital few versus the trivial many” might be a better approach for most businesses.

  8. 500 person late stage technology startup. I think it would be interesting (or at least interesting for a while, and maybe pretty distracting for my team of engineers in the short term), but I don’t know what the long term if any would be, or if it would even be positive. It’s just hard to predict. If you gave this type of info to everyone at Apple in 1999 would you increase the chances of getting the iPad in 2001 and the iPhone in 2007? It seems kind of… secondary or even tertiary to that kind of outcome. A lot of things can’t be reduced to easily-measured numbers about the current business. Analysis of current business is kind of the opposite of vision for the future, depending on the size of the current business and the time scale.

    On the other hand, once a product is out in the field it becomes possible to generate some interesting metrics that might some relevance for what you are trying to do in the future. For example, suppose you have a complex product with a number of identifiable subcomponents which are the responsibility of various engineering teams. Suppose further that the product has a lot of software in it. Suppose further that this enables you to identify the rate of customer-encountered defects by product component, normalized to the size and complexity of the various product components and by the magnitude of the various defects. (These component sizes would be reasonable engineering ballpark estimate by people knowledgeable about the product. You don’t just only count lines of code or whatever. But let’s posit that this is a solvable problem.) This tells you something, maybe a lot, about your people, your code base, your testing approach, etc, if you are willing to look at the data and come to understand what it is telling you. Well, I guess you have to be willing to ask the question in the first place, and not everybody is.

    So sure, some numbers on my company I’d love for somebody to put in front of the entire company on a daily basis:
    – customer found defect rate by software component (about 30 numbers here)
    – revenue rate and margin by hardware model
    – revenue rate for various licensed software

    Hell I’d settle for getting these once per quarter. As it is, I settle for informal estimates obtained by seeking out individuals across the company (and looking at the bug database directly). How is that for a company communication culture? I guess my company is in some sense a counterexample to the general notion that efficiency of information exchange matters at all. Apparently the company can be successful by providing almost no company-wide metrics to anybody, as long as everybody is building the right thing and is doing a reasonably OK job of it.

  9. You know you are just going to end up with measures such as “lines of code” and “number of bugs” which soon will be “optimized” so that they look great but contribute nothing.

  10. Daily is probably too granular. I know sparklines are intended to provide context, but they don’t work that well in my experience…too small, no scale specified, scale different for each one.

    Corporate dashboards seem more promising. No paper, arbitrary data options, ability to intersperse informational bits that ordinary line workers will seek out of their own volition.

    I have worked for one company that experimented with them, but the implementation was limited. Plasma screens at the elevator bank with no intranet availability for depth on the data. But the real problem was that we had no useful metrics to monitor. We were an incubator of sorts, with half a dozen funded but pre-customer startups. The only real metric we had was burn rate, and that would have been demoralizing.

  12. I worked at a large (10k+) company that used metrics successfully across many different lines of business. Management decided which metrics matter to who, employees in various roles used metrics on the web to help them make decisions and do their job. Go to the intranet, and you’d see a custom dashboard with the metrics that mattered to you, and that you’d get held accountable for. It’s easier to focus on a concrete goal, like get a metric as close to zero as possible.

  13. Maybe keep it simple with one actionable metric per day?

    Showing sparklines in a fish-eye lens view could work well. Metrics can be crudely modelled as being hierarchical, e.g. company makes $X profit > sales makes $Y sales > salesperson makes $Z sales. So you could show each worker a big sparkline for their own responsibility and smaller sparklines for those responsibilities above and below them.

    e.g. You’re a salesperson (a leaf in this tree), so you get a big view of your own sales and above it are small sparklines showing company and sales dept. Or you’re manager of sales dept, so you get a small sparkline showing company, big sparkline showing your own dept profit, and small sparklines for each salesperson).

  14. Would it be helpful to arrive at work and find an 11×17 sheet of paper across your keyboard with perhaps 150 metrics of the company’s performance?

    My company has over 100,000 employees around the world. So most of the metrics of the company’s performance are meaningless to me because the metrics regard things I never see and have no knowledge of. Even our site is so large that most of its metrics are divorced from the work I do every day. Such a sheet of paper would be completely meaningless to me unless I chose what was on it.

  16. Printing it out seems unnecessary. We instrument everything with statsd (https://github.com/etsy/statsd) or statsd compatible formats and use Grafana (https://github.com/grafana/grafana) to display the data. One team may be concerned with signups and another may be interested in performance of an api they’re responsible for. Each team can craft their own dashboards with real-time metrics that pertain to them.

  17. MJT: Paper at 1200 dpi is much higher resolution than any screen, so my theory is that more information can be conveyed quickly via 11×17 paper. Also it seems that a piece of paper left on a chair or keyboard is more likely to be viewed than a dashboard that requires visiting a browser bookmark.

Comments are closed.