Picture from Fine Country Lifestyle – Devon Farm Shop
I’ve done the same presentation about psychology, seduction and persuasion several times. It changes a bit here and there depending on the audience. I include a few more contemporary examples, add a few gags and throw in the odd bit of data to back things up.
At one part of the presentation I talk about how adding largely meaningless adjectives to products makes them more appealing – so pepper becomes hand-cracked pepper or we add a provenance like Suffolk honey. I’d always think of these off the top of my head during the presentation and, for the sake of a little humour, would try and invent outlandish examples to make the point (an in so-doing probably dilute it). Then last year I was watching the inestimable Stewart Lee when he amusingly parodied the craft-beer industry with some ludicrous names: Gandalf’s Memory Stick, Hogwarts Bukkake and it inspired me to keep doing the same gag.
I must have been holding on to this idea for a while and I got chatting to Dare’s technical director, Charlie, in a cab. Charlie’s got an academic background in English and a similar sense of humour so we naturally came round to the idea that this generation of novel food labels could be done in a random fashion. It seemed so simple to concoct the recipe: take a foodstuff, add a method and a provenance and the result takes an ordinary staple and turns it into a farm shop or artisan product that can be sold with a healthy mark-up.
Persuasive marketing nomenclature, automated with a tinge of comedy.
So we (well he) started building it. A simple JavaScript took items from three arrays (lists of data) and combined them at random in the order: Provenance, Method, Foodstuff. It worked quite well. But, thinking about the old adage of garbage-in, garbage-out, we noticed that some of the combinations didn’t work.
Does it feel right?
At this point we should stop and consider what we mean by work. It’s quite subjective, but you have to think about it a bit. The comedy is about the combinations appearing almost right but a bit outlandish. If you go too far toward the outlandish then it just feels wrong. In some cases this is obvious – the pairing of methods with foods that don’t make sense hand-reared houmous, pulled brioche, grass-fed asparagus. So we started to think about what it was about these pairings that made them wrong and how we could eliminate them. Do you, for example, identify a matrix where methods applicable to foods are deemed ok/wrong? So hand-reared is relevant to all animal products? line-caught is relevant to seafood only? Or do you simply manually edit the list to exclude methods that are too niche? The trouble with doing that is that you reduce the serendipitous moments that make this work. Trying to avoid creating a behemoth that relies on learning or crowd-sourcing inappropriate pairings I set about building an Excel sheet with a series of lookup tables that allowed me to fettle with the source lists and try out combinations without relying on a very busy Charlie to repopulate his script.
Syntax is important
Creating the spreadsheet opened up even more questions. Taking a leap from an unconnected musing I had on Twitter last week, it occurred to me that order – syntax – is an important part of the output. Food will always come at the end but does changing the position of provenance affect the humour or the apparent luxury of the item? To use an example, is Newlyn fried corn a different product to fried Newlyn corn? So the method seems more artisan and niche if it’s Newlyn fried (presumably only a handful of people know how to fry the Newlyn way) as opposed to the corn being from Newlyn and then simply fried? It’s almost the difference between an item being at the bottom of the prestige retail hierarchy and the top.
Aside: Could you put the following retailers in hierarchy of perceived prestige? Tesco Finest, Waitrose Seriously, Marks & Spencer, Whole Foods, Borough Market, Artisan Farm Shop, Selfridges …
Provenance and terroir
Looking at the list we’d made for provenance it was clear there were two things going on. Once was about the association a place had with the growing or raising of food and the other was about what this meant by association. So the concept of terroir is that the geography, geology and climate of a place affects a foodstuff. It’s hugely important in wine and coffee to know the place it’s come from, but also in items like meats or vegetables (Hereford beef, Norfolk turkey). It gets more complicated when you add in the method of preparation or the regional significance of a recipe (A Bakewell tart, a Cornish pasty) or get super-niche and choose a specific producer Blacker Hall quiche. Consequently, the list we compiled is composed of places that have strong associations with food – largely agricultural counties, coastal locations and regional recipes. I then scoured a list of Britain’s top 50 farm shops and delicatessens for examples of artisan-sounding producers
What’s a method, what’s a foodstuff?
Related to our thinking about ordering and the awkwardness of pairings it became apparent that the foodstuff could be the array that includes a variety of methods specific to that food. So, instead of simply putting pork we could add pulled pork to the list. We could have scallops and hand-dived scallops. This would mean that we wouldn’t need to worry about hand-dived pork coming up but we could keep the fancy-pants descriptor of hand-dived to make the scallops seem more interesting. It’s fair to say it had stepped away a little from the original plan to have a simple 1+1+1 = 3 pattern (but that was about to have another twist anyway). We started to think a bit more about what constitutes a food and that complicated dishes don’t work so well as items that are atomic or simple but this wasn’t clear cut. Bakewell vanilla-infused cupcakes works but Jersey broiled yoghurt doesn’t. For every decent example involving brioche, sourdough, quiche, pasties there were far more decent examples involving single ingredients – asparagus, quinoa, lentils, beans, chicken. Once again, order plays a part here and having categories might help solve this. Hold that thought.
Something extra
Finally, after about two days fiddling about in Excel and chatting to Charlie we decided to throw in another part to the concatenated string, a garnish perhaps. We had a randomly-appearing descriptor that affected the overall product. It could be vegan or gluten-free or giant. So, not so-much a method or a provenance but in the spirit of the type of thing that gets added to nomenclature to change the perception Clearly the taxonomic importance of vegan/gluten-free over micro/giant is worth bearing in mind. It many cases it works wonderfully: Giant sugared Herefordshire pudding in others not so well Salt-Baked Pommery Vegan Steak Pies, so it’s fair to say that becomes a matter of user preference. Which leads us neatly on…
Getting it out there
After a while you realise there’s loads more you can do and several of these things made great sense. I always loved the Urban Spoon app that helped you find a restaurant matching a series of criteria at random, the trick was that you could lock down the most important part of your criteria – for example price, and then leave the random bit to choose the genre, location or both. It strikes me that this might be a nice add-on to our generator. You might lock-down the foodstuff and just play around with random combinations of qualifiers – the most fancy chicken product you can find for example. Then there was the consideration that this could have a crowd-sourced element; users could work in volume to rate the best combinations or highlight ones that don’t work. Clearly this would mean a lot more coding effort than we could afford to spend. What about supporting unique URLs for each combination so they could be shared or copied straight into a tweet link. And finally, what about categorisation? would this be better if you could focus-in on drinks, ingredients or prepared products like quiche, cakes, pastas.
Everything’s a remix
Back to reality and I realised fairly early on that this wasn’t that new. There are about ten thousand ‘generator’ sites that compose sitcom and film character names, craft beers and, perhaps channelling a little of the Bill Bryson observation on British place names, a village name generator. What I rather like about all this is that it seems to be most effective with our wonderful language here in Britain. I hastily trimmed out provenances that weren’t British and have tried to keep the foodstuffs a little native, scattering a bit of brioche or salami here and there does work but one must be parsimonious. when the strings get a bit long and they pick up quite specific methods like -infused or cold-pressed it can definitely feel a bit Heston Bloodyhell (sic)
To what end?
So, where does this leave us? Perhaps one day Charlie and I will get a public facing version up, designed hopefully around a style that befits the point-of-sale references we see in hipster marketplaces. A tool that uses some of the functionality we’ve mused about and ultimately becomes a playful little twitter stream. I like the idea that you could run this for 6 months with a voting mechanic, gather the data and establish a shop somewhere in a quaint Cotswold market town (Greater Drowsisle?) that sells products derived entirely from this output.
In the meantime it has given me a great chance to revisit ontological thinking, nomenclature and linguistics and logic. Any opportunity to play around in those fields can’t help but contribute to my understanding and enjoyment of the job I do on a daily basis.
A selection of how it works (or doesn’t).
- Irish air-dried kale
- Ballymaloe thin-sliced mackrel
- Hand cut Suffolk micro couscous
- Fermented Worcestershire buffalo
- Pressed Derbyshire giant pheasant
- Castleford dried rye bread
UPDATE: Now showing on Twitter … @shinyplums
UPDATE: A Daily Mail headline generator and a direction to consider the writings of Brian Wansink concerning food psychology , thanks to Juliet Hodges.
UPDATE: Try it out for yourself with our artisinal food generator
John Gibbard 