Rule 10: Know thy farmer.

(Part of the "Analytics Is Eating the World" series, an insider’s advice on data and analytics for CEOs, boards, and general managers. Featuring food metaphors.)

If a server should ever present you with a slab of wood supporting a row of tiny spikes, upon each of which a pickled vegetable has been impaled, you would be justified in thinking that you’d grown to the size of a giant or been abducted by Moby. But it’s more likely that you’ve embarked on the four-hour tasting meal at Blue Hill at Stone Barns, a farm-to-table restaurant in the Hudson Valley.

You may have had the good fortune of obtaining a reservation in winter, in which case you experienced a number of delightful sensations even before the mini-vegetables showed up. You arrived at the restaurant through fields blanketed in snow, walked in silence into a dramatically lighted stone courtyard, and entered a convivial space to the sounds and smells of a crackling fire. (The management of this blog sincerely regrets the clichés, but facts are facts.)

Many of Blue Hill’s ingredients come from the on-premises farm and other local producers. Even if you’re a longtime farm-to-table enthusiast, you’ll find the food at Blue Hill remarkable. For one thing, it’s disproportionately made of vegetables, as foretold by the pickled bits that got the Oliver Cromwell treatment. Carnivorous readers may be experiencing a moment of anticipatory dread, but the meal turns out to be very satisfying (and suitable for omnivores).

Chef Dan Barber grew up in New York City and spent summers working on his grandparents’ farm in the Berkshires. He trained as a chef at the French Culinary Institute before starting Stone Barns, a nonprofit food education center, and Blue Hill, the for-profit restaurant on its premises that was recently ranked second-highest among U.S. restaurants on the S.Pellegrino 50 Best Restaurants list that we have discussed before.

“Sustainability” is a word that a lot of businesspeople use today, generally referring to a mix of economic, social, and environmental practices that are believed to strike the right balance between the needs of the present and the needs of future generations. Barber uses the term “farm-driven cuisine,” advocating such practices as

• Encouraging diners to ask detailed questions about the way their food was grown, beyond where it came from and whether chemicals were used
• Creating a dialogue among chefs, eaters, and farmers to transmit feedback and influence decisions all the way back to the field
• Buying local ingredients optimized for the specific attributes a chef cares about (such as taste and nutritive value, as opposed to uniformity and durability)
• Having shorter supply chains and having food spend less time in storage, during which quality can degrade
• Resisting oversimplifications such as “organic is good” and “awareness of seed genetics is bad”

If leaders have the right mindset, analytics projects can embrace their own version of sustainability.

I was once asked to create a wealth-based segmentation for a retail bank and asset manager, with a goal of accurately estimating the wealth of every household in the United States. My team had three choices: we could use different touchpoints to ask bank customers directly about their assets, purchase an estimated wealth score for all U.S. households from a data broker, or blend one or more external scores with locally sourced information about customers.

We chose the third approach, building a segmented model that allowed us to predict household-level liquid assets by first looking, as much as possible, at home-grown information the bank had on its own customers before supplementing it with outside ingredients. When buying external data to estimate the wealth of non-customers, we set up the model so that it would not fail if the external data feed got disrupted in the future.

The resulting approach was more accurate than the ask-a-customer approach, was robust to disruptions, and was beneficially different from what the bank’s competitors might know about the same households, since we benefited from proprietary information that was available only to this bank.

This approach can be generalized to other cases where a team needs to build a new forecasting model. Four lessons from the “know thy farmer” approach are:

• Avoid black-box approaches; keep track of sources and document the lineage of how the original underlying data gets turned into predictions
• Keep multiple sources on hand, even if the redundancy may seem inefficient in the near term; it reduces risk in the longer term and allows for experimental improvements
• Take a longer-term view on the sustainability of your data: will it still be there under a more restrictive regulatory regime? Under different commercial pressures?
• Plan for outages and ensure that your model will work even if there’s a data disruption lasting a week or a month

At Blue Hill, the menu necessarily varies by the season, and sometimes varies by table, but it generally ends with a final presentation of something sweet that matches the overall theme of the evening. Here I will leave you with a bracing glass of grappa instead: Richard Cook’s canonical piece on how complex systems fail.

This article first appeared on the Braff & Company blog at braff.co. Subscribe for automatic updates.