Thursday, August 20, 2020

It's time for lean philanthropy: a case study

I’ve long been an advocate of using Lean Startup principles to advance non-profits and philanthropic organizations, whose work is so urgently needed at this moment in history.

As in many areas of the crisis, I have seen a number of new organizations founded to help alleviate hunger and food insecurity in the wake of COVID-19. SF New Deal and Frontline Foods were early leaders. In this blog post, I want to share the story of a lean pop-up organization called HelpKitchen. Its creation was a chance to see how Lean Startup principles can be applied to philanthropic work as well as high-growth startups. 

Early on in the crisis, I recruited Justin Mares to build a team to run this experiment. Below, you can see how he and an all-volunteer team quickly validated a need, tested assumptions, and built a lean, fast-moving organization that has served over 170,000 meals in the weeks since it was founded. I hope the lessons he shares here will help others involved in philanthropic work during these challenging times.


In April, the world was about a month into the COVID-19 pandemic. As workers were furloughed and laid off, millions of Americans found themselves newly food-insecure and hungry, without any idea of where to get their next meal.

Lines at food pantries were hundreds of cars and hours long, as the food security system struggled to keep up with unprecedented demand. Some food banks saw demand 3-4x normal, just as their volunteer base shrunk due to worries about COVID-19 infection.


 Minneapolis, Minnesota: Cars line up at a drive-thru food pantry. Neil Blake/Grand Rapids Press/AP

The food security system was (and unfortunately still is) completely overwhelmed and struggling to meet the massive influx of need.

That set the stage for HelpKitchen - a non-profit organization we started in the last week of April. Since then, in about 3 months, we’ve fed over 170,000 people and injected over $1mm into the local economy in San Francisco and Detroit. Our goal is to try and ease the burden lockdowns have placed on both local restaurants and the food security system.

We’re hopeful that by sharing our story, others in the tech and philanthropic world can get some ideas for ways they can use lean principles to help their communities through this trying time.

How HelpKitchen began

The idea for HelpKitchen began with a conversation between Eric Ries and Jeff (founder and CEO of Twilio) and Erica Lawson. Jeff, Eric and Erica saw the massive increase in demand the food security system was experiencing, and wanted to do something about it.

As soon as Eric and the Lawsons had agreed to tackle the problem, Eric began talking to people in the tech community who were willing to help.

We connected in a WhatsApp group for tech people involved in COVID response projects. I was in the group due to my involvement (with Brent Summers) in launching GiveLocal, a site where customers could support their favorite restaurants through COVID lockdowns by buying gift cards. As soon as I saw Eric was looking for help with a food security project, I wanted to help. So did the two friends I looped in, Brent (again!) and Jeff Nobbs, who I’d worked with at Perfect Keto.

Sketching Hypotheses

We began by creating a few hypotheses, based on  what we saw happening in the food security system:

  1. Restaurants were hurting due to lockdowns: less foot traffic, less revenue, and facing the prospect of laying off many staff members. At the same time, they had fully licensed kitchens and access to purchasing food at bulk discount prices. After all, buying food in large quantities and selling it at a markup is their entire business.
  2. Food banks were seeing increased demand, but struggling to meet that demand. They were also experiencing a volunteer shortage due to concerns over the spread of COVID-19. To make matters worse, many food banks'  sources of food supply (extra restaurant food, food company donations) were disappearing
  3. Food insecure individuals fell into two camps. In the first camp were those who had experience with the food security system and understood how to navigate it. The second was made up of millions of newly food insecure individuals who had no experience navigating the system and needed a meal. This second group included many people  who had recently been laid off or furloughed.

After looking at the various stakeholders involved in the food security system we had an idea:

Why not have restaurants feed people struggling with food insecurity?

The model was simple: we’d build a SMS texting tool that matched the food insecure with a free meal from a partner restaurant, all covered by donor dollars. By doing so, we’d keep restaurants busy (and their employees working), inject funds into the local community, and ease the burden on food banks.


Validating Assumptions

Having come up with our hypotheses, we then created a series of assumptions to validate as we moved forward:

  1. Restaurants would happily participate in a program that generated revenue and gave back to their community.
  2. We could build a technical backend that matched food insecure individuals with partner restaurants.
  3. Those struggling with food insecurity would be willing to share some information and text a phone number to get matched with a partner restaurant to pick up a meal.
  4. This system would provide a scalable answer to the food security problem, which would resonate with donors.

In the spirit of lean methodology, we set about proving these assumptions, starting with the restaurants.

Fortunately, Jeff Nobbs (one of the pillars of the HelpKitchen team) is the co-owner of Kitava, a restaurant in San Francisco. He helped finalize our assumptions, and then went and interviewed several restaurant friends. After some positive feedback, we set a 24-hour goal to get 10 restaurants on board. If we couldn’t get 10 restaurants to commit in one day, the idea likely wasn’t compelling enough.

Fortunately for all involved, restaurants responded extremely well - we had nearly a 100% participation rate for restaurants we reached out to.

After getting 10 restaurants to commit to participating in a pilot, the next question was: could we build a system that matched those in need of a meal with a partner restaurant?

The fastest way to get a system up and running was with a no-code solution. Brent, who previously founded a no-code education company, spun up a text-based intake system with Twilio. When a request came in, we used their Flex chat system to manage requests and responses. Then, we had an Airtable dashboard on the back end where we could manually match restaurants, manage capacity and pair individuals in need with a restaurant near their zip code for a meal.

Though this V1 buildout was incredibly manual - as it required each of us to monitor the Twilio Flex chat for hours each day, and hand-match individuals in Airtable - it worked well enough for us to validate whether we could use technology to pair individuals with restaurant partners. Assuming it worked (which it did), we figured we could simply build software later to automate the matching and texting process. Fortunately for all of the founders’ romantic relationships, we automated this in just a few weeks. To see how the system evolved, Brent put together a full writeup here.

Once we validated that restaurants wanted this and we could match individuals to a restaurant to get a meal, we decided to test our most important assumption - would individuals in need of a meal text a number and get matched to a restaurant near them?

To make sure we were only reaching those who were in need, we printed flyers with the HelpKitchen number, compiled a list of food drives in San Francisco, and handed out flyers made to people waiting for food in lines that stretched for blocks. And on day one of handing out flyers, we managed to feed 50 people at our pilot restaurant! Assumption: validated!

Lastly, we had to validate that this approach to building a scalable tool to address part of the food security program resonated with donors. After all, in order for this all to work it would require the generous donations of philanthropists and (ideally) the support of government programs down the line.

In this regard, we were incredibly fortunate that Jeff and Erica Lawson agreed to donate $2mm to fund our approach and prove this out as a valid and scalable way to help those struggling with food insecurity.

That was enough for us, and with those assumptions validated we went to work.

Testing the MVP

We landed on the below MVP approach to roll out and test:

For those who need a meal, they simply send a text to HelpKitchen and tell us what neighborhood they’re in and how many meals they need (up to a limit of eight if they have a large family). We then match them with a partner restaurant who prepares their meals for pickup. The individual then picks up the meals, gives the restaurant their name and last four digits of their phone number, and our work is done. .

For restaurants, they agree to partner with HelpKitchen, set a price per meal, and tell us what they’ll be making for the individuals we match them with. Restaurants then tell us how many people they can serve each day and we match them with food insecure individuals. The restaurant gets access to a dashboard that shows them how many people are going to pick up meals each day, and marks meals as fulfilled each time one gets picked up.

For donors, they donate to feed the hungry in their community while also keeping restaurants in business, and restaurant employees employed. They can be confident that each dollar of their donation goes directly to feeding a hungry individual.

On the back end, we got to work building out the MVP matching tool: first by having me, Jeff and Brent manage and respond to incoming messages while manually matching individuals to restaurants in the Airtable sheet. As soon as we got to the level of hundreds of meals per day, this became all-consuming, literally taking hours of our time to manually match individuals to restaurants and manage restaurant capacity. To scale further, we built out a team of volunteers and built an automated system that balanced restaurant capacity, individual requests and zip code assignment. Before long, we were feeding thousands of people per week. As the system needs evolved, we built more features to make sure we were helping people and restaurants in the most efficient way. This included implementing a “learn phase” to complete our MVP, which took user feedback from those picking up meals, sent that to the restaurants and led to improved meal pickups.

We also invested time in building out a set of Standard Operating Procedures (SOPs) for restaurant partners to use and reference as they worked with HelpKitchen. As the system scaled (and our tech back-end could match thousands of individuals daily to restaurant partners), we recruited more restaurants to the program and handed out more flyers at food banks. We also began to partner with organizations in San Francisco (like Glide) to reach those in their community who were struggling with food insecurity.

After just four weeks, we were serving nearly 20k meals a week. Pretty wild for an all-volunteer team that had hacked together a system with nothing more than Airtable, Zapier, Twilio and a bit of Javascript.

Where we are now

After validating that the HelpKitchen system can match tens of thousands of individuals with restaurants each week, and after 170k meals served in the last few months, we’re now focused on scaling the HelpKitchen approach to more cities across the country.

Today, HelpKitchen is live in two cities (San Francisco and Detroit) and in discussions with several others about expansion. We’re incredibly proud of the progress we’ve made in a short period of time and eager to continue helping more people.

If you think you can help us achieve our mission, please reach out or donate.

Since our end-of-April launch, in just over 3 months (and with no budget and  an all-volunteer team) we’ve served 170,000+ meals, contributed $1mm+ to the San Francisco and Detroit local economies, and raised $2.3mm in total for the HelpKitchen operation.

Thank you to our core group of volunteers: Sarra Mumayiz on the operations side, and David Head, Jason Dielman, Ryan Wang, Tom Monks and Ahmad Khan on the tech side. We could not have done this without you all!

blog comments powered by Disqus