I'll write separate posts about how to systematically create and how to measure je ne sais quoi (JNSQ), but for now, I just want to argue that it's something important to consider as a product manager or designer.
Imagine a hypothetical popular restaurant, Crawford's Tap Room, known to its patrons as CTR. CTR's owner, Crawford, gets visited by a vendor that promises to install software that will measure everything at his restaurant and help to optimize his revenue and profit. Crawford, being somewhat of a technophile and control freak, jumps at the chance to have detailed knowledge about his business.
Soon enough, reservations are online, servers take orders digitally, alcohol and food inventory is managed through a simplified supply chain management system, and all sorts of per-table analytics are available.
After pouring through the numbers for a few months, Crawford makes a number of changes to CTR. He raises his corkage from $15 to $25, finding that the corkage fee is largely inelastic. He fits all of his booze bottles with a measurement system that reduces waste from the extra large drinks that bartenders pour. He stops waiters from tasting wine, finding that he's losing hundreds of dollars a week on free glasses. He stops serving special entrees over $30, noting that specials under $25 tend to sell better. He uses the online reservation system to keep the restaurant slightly overbooked so he's full more often. And so on, with dozens of tiny operational changes.
For two months, his revenue is soaring and bottom line is healthier than it's ever been. Then, all of a sudden, there's a drop. Crawford analyzes the numbers again, but nothing he tries is working. Finally, at a loss, he surveys 100 of his top patrons.
He finds that:
- Patrons will pay the extra corkage fee because they're already in the restaurant with a special bottle of wine, but feel resentful afterward for being taken advantage of.
- Drinks have gone down in quality and seem to be stingy and weak.
- Waiters don't suggest good wine pairings anymore.
- They miss the exotic specials that CTR used to serve.
- People think that it takes too long to get a table, even when they have a reservation.
All in all, he finds that many of the optimizations he made to increase his revenue were made at a huge long-term expense: his restaurant lost its unanalyzable je ne sais quoi.
Now, I wouldn't argue that all of the changes that CTR made were folly. For example, a few unproductive waiters were fired and the digital order system helped to streamline the kitchen. The new data available to Crawford enabled him to make these decisions.
However, when Crawford relied entirely on his data, he forgot about why his customers were coming to his restaurant: gastronomic creativity, the best drinks in town, and a knowledgeable wait staff.
When laid out like this in a real world example, it's obvious that no restaurant could rely solely on the advice of an unfeeling digital system. A good restaurateur knows what kinds of dishes to serve, interacts with his clients, watches over the restaurant every night, and hires the right kind of bartenders, hosts, and waiters.
Unfortunately, many Web sites operate as if percentages, click-through-rates, and conversions are God. Numbers should certainly inform decisions and give product managers and designers information that was heretofore unimaginable in the physical world. But, if all decisions are made by tweaks that inch your numbers higher and higher, maybe you're spending too much time on the short term and not enough time on the creative features that affect your site in the long term.
Numbers are usually excellent at telling you "what," but don't get at the "why." Sure, that change made your bounce rate go down, but why? Revenue went up when you put more ads on your site, but are you pissing off customers? Conversions went up when you changed that button color, but do users think your site looks garish and unprofessional?
Also, especially in volume, numbers are good at showing how changes immediately effect your site. But, what about how things change over longer periods of time? Longitudinal numbers are really difficult to measure and normalize. Seismic shifts in your user population's attitude are usually evident after months, not a few days. If you're not sensitive to the qualitative feeling of your users, then you might make your numbers today, but not in the long term.
Again, I want to stress that numbers are important and can drive product decisions that positively influence users. But, when your left brain is wringing its hands trying to figure out how to inch up your stats, make sure that your right brain is engaged an equal amount in trying to preserve your site's je ne sais quoi.