A MARKET GONE STALE

In 2019, seven out of 10 hamburgers and nine out of 10 hotdogs consumed in Mexico were made with Bimbo buns. In the same year, street food vendors (dubbed “Special Channels”) represented 33% of the company’s total sales. But between 2020 and 2021, as eating patterns shifted through COVID and beyond, and with competition from Big Fast Food as fierce as ever, that percentage fell by 8% – resulting in an annual deficit of $192M USD for the company.

BRIEF: HELP BIMBO CUSTOMERS SELL MORE STREET FOOD

How could Bimbo support its commercial allies’ recovery from the economic downturn and increase sales, while creating and strengthening brand awareness among both food vendors and end consumers alike?

OBJECTIVES

1. Increase visibility and drive traffic to street food carts, shops, and stalls across Mexico.

2. Recover 10% of sales participation in Special Channels.

THE GREATEST GUIDE

The world’s first fully generative search and signage system designed to put 8,402 of Mexico’s most creative hamburger and hotdog stands on the map.

CREATIVE FOOD DESERVES CREATIVE DATA

From octopus to grasshopper, the creativity of Mexico’s street food is staggering. It’s also highly diverse, with each of these 8,402 stands selling unique and local recipes. Our data collection and application of it to mapping and generative AI reflects the variance and creativity of the food, while respecting a rich, traditional visual language. The food isn’t generic – neither is our technology solution.

THE NOT-SO-SECRET SAUCE

To drive word on the street, we turned to 18 earned television appearances, radio coverage, foodie influencers and +340 social media posts, and digital ads to get people exploring local food vendors via the Google Maps API-powered website.

AUDIENCES

Jocheros and hamburgueseros – the street food vendors that Bimbo’s served for more than 78 years – and everyone who enjoys their creations.

KNOWN DATA, NOVEL PURPOSE

We started by collecting data from thousands of street food vendors across Mexico, both directly, and using the IVY and RTM data management tools created by Bimbo’s business analysis center. That data had previously only been used for placing purchase orders for buns.

NAVIGATING INVISIBLE WALLS

The commercial application of generative AI at scale is challenging; throughout development, we discovered invisible walls everywhere. For example, generative platforms are likely to render burgers that look like BigMacs, suggesting heavy bias in their training data. Because the street food on our menu doesn’t look like ubiquitous fast food, and often uses ingredients not well documented on the Internet, we had to push at the edges of what AI could actually render, through rules-based decisioning.

BRAND RELEVANCE

Bimbo has supplied buns to millions of food vendors for more than 78 years, building trust and loyalty with its clients. To support their culinary endeavors, we saw an opportunity to give them exposure and attract their target audience, unlocking the power of big tech for small business.

AUTOMATED ART DIRECTION

To produce 8,402 data-driven point-of-sale posters we combined Mexico’s rich tradition of sign painting with the scale and efficiency of artificial intelligence – we built up an innovative, iterative workflow between human art directors and our AI models. Rather than bespoke art direction, our approach was rules-based (e.g., the rule of thirds), which we rolled up into a programmatic creative system. For example, our automated drop shadow had to work for both long, wide hotdogs, and narrow, high hamburgers.

GENERATIVE STACK

OpenAI CLIP – image-to-text, to create prompts from selected images

OpenAI Dall-E2 – text-to-image to produce base food paintings, backgrounds

RmBGAI – for automated close cropping

Pillow – Python for compositing (essentially programmatic Photoshop)

Wand – Python bindings for ImageMagick, for compositing

Stability AI – upscale API, for increasing resolution to poster-size print specs

PROGRAMMATIC PRODUCTION DESIGN VIA PYTHON

Drop shadow

Vignetting layer

Rough border layer

High-pass sharpening filter

Scanned, hand painted-type

SCALE

National, with 8,402 hotdog and burger carts served across Mexico’s 31 states.

OUTPUT

We created a world-first workflow designed to turn data into prompts and prompts into food paintings, all are masked and combined with generative backgrounds. Dynamically typeset from hand-painted fonts, automatically retouched and branded, then ingested into our system as separate layers. Ultimately, we generated 42,010 unique point-of-sale materials for Bimbo’s small business customers – five versions for each vendor, all combined, up-resized and output to final specs – they received via WhatsApp ready to print.

BUSINESS IMPACT

• +12,200,000 impressions across Mexico

• +77,000 online map visits to date

• 42,010 unique point-of-sale images generated for hotdog and burger carts, shops, and street vendors – five versions for each

• +23% increase in sales among Bimbo’s “Special Channels” compared to the 10% target

• Set a new +180% week-over-week nationwide sales record

STREET CRED

By putting 8,402 of Bimbo’s customers on an online map accessible to all, we not only supported the economic livelihood of thousands of families across Mexico, we also created a strong halo effect for the Bimbo brand. The Greatest Guide established an authentic connection between a multinational bakery giant and Mexico’s vibrant community of independent street food vendors and foodies. It also strengthened commercial relationships, including the trust and loyalty that Bimbo has built with some individual food vendors for more than 40 years.