SNAP Analytics (2) - Purchase Patterns

Roughly 15% of the United States population receives SNAP funding to help pay for food and beverage items.  We know that when SNAP (food stamp) funding is released in each state (see SNAP Analytics (1) - Funding and spikes)  this is accompanied by significant sales spikes on some products,

If 15% of all shoppers visit your store within a 2-3 day period you should see a sales spike on  everything they buy, SNAP funded or not . So, why do we not see a spike on everything?  Why are some spikes so much bigger than others?

The chart below shows (simulated) daily point of sales data for a SNAP-responsive product in one state.  The horizontal axis shows days of the month, the vertical shows a sales 'Index' relative to the average day.  You can see that sales around the 1st and 10th of the month are roughly double what they are on other days.

If you also know that this state releases SNAP funding on the 1st and 10th of each month, you might assume that the SNAP shopper takes their newly-charged


card and within 2-3 days spends the lot.

A relatively high proportion of SNAP funding is spent quickly and this ties well with the idea of a "stock-up" trip.  (If you have the capability to see basket size by date, you should be able to confirm that baskets around SNAP release dates are substantially larger than otherwise.)

So why do I think that this does not represent all SNAP spending?  

Some products are just not good candidates for a once-a-month stock-up trip.  Milk for a month?  I don't think so.  Bananas seem to go soft in my house if I forget them for 1-2 days.  A month's supply of a product may take up more room that I have available in the cart, car,  refrigerator, freezer, or store cupboard.  Some of this will have to wait. 
Some products are more attractive for stock-up trips: larger sizes of frequently consumed products that are stable (on shelf, in fridge or freezer) and perhaps also with "treats" that can be purchased while there is a little extra money available.
According to the USDA, in 2011,  the average monthly SNAP benefit per household was  $284.   Remember that this is $284 spent on SNAP eligible products only:  leave out  non-food/beverage items, hot foods, ready-to-eat items, alcohol and tobacco.  Can it be done?  Yes, but its going to be tough to fit into one shopping cart or in your car or in your kitchen.   $284 is the monthly benefit for the average household of 2.1 people.  Could a family of 4 realistically buy even most of their food once a month?  Even if the SNAP shopper could buy all their food and beverage  items in one trip, they still need other grocery items, paper goods, cleaning products etc. that takes up additional space. 
Finally, the  countrywide adoption of EBT cards, rather than paper vouchers, means the SNAP shopper can spend as little as they need right now without losing any of their benefits.  (Something the similar WIC program is still working on in most states).

Despite the big spikes in sales we see for some products around SNAP funding dates, the SNAP shopper is not buying all their monthly supplies in one trip.

  Some products will be much more responsive to SNAP funding than others because they fit well with the SNAP shopper's trip-type and taste preferences.

So, how do you know if your products are responsive to SNAP funding dates?

   If you have access to the payment details by basket it's a slightly simpler process of querying your data and correlating across to SNAP release dates.  If you have daily point of sale data you need to build predictive models against total sales rather than SNAP specific sales (Do you need daily Point of Sale data?).  In either case, you are dealing with very large quantities of data and need the right tools and the knowledge to wield them effectively (Bringing your analytical guns to bear on Big DataData handling - the right tool for the job).

If you do not know which products, stores and dates will see spikes in demand how can you ensure product is on-shelf?  Ignoring SNAP may be costing you sales.

If you're ready to get started - call me.

SNAP Analytics (1) - Funding and spikes.

Back in August I took a quick look at SNAP, the US government's "Supplemental Nutritional Assistance Program", formerly known as "Food Stamps". (see What's driving your Sales? SNAP? ).  

In 2011, approximately 15% of the US population received SNAP benefits that they can spend on most food and beverage items in store.  SNAP funding has doubled in the last 3 years.

SNAP can create large spikes in demand at the store and yet, because of the way these funds are distributed , this is typically hidden from analysts looking at aggregate data. (see Do you need daily Point of Sale data?... )

If you do not know which products, stores and dates will see spikes in demand how can you ensure product is on-shelf?  Ignoring SNAP may be costing you sales.

This is the first in a series of posts covering Analytics around SNAP and opportunities for driving incremental sales.

The table below shows the days of the month (highlighted in red) when each state distributes SNAP funding (click on it to enlarge):

US SNAP funding patterns by state and day of the month.

At the top of the table, we have the States that distribute all their funding on just 1 day of the month. Out of the 54 States, Districts and Territories shown just 10 of these distribute on one day and (thankfully) they are not the ones with the biggest sales. But, if you are selling a SNAP responsive product you will want to ensure you plenty of stock in-store and on-shelf on the first for these states. 

The States are ranked in terms of the impact SNAP distribution is likely to have within each state: the size of the sales "spike". Fewer SNAP distribution days and the spike will be higher. Perhaps less easy to explain but the closer that SNAP distribution days are to each other, the more their shoppers overlap in store and the higher the sales spike. Consequently, Utah with 3 dispersed distribution days may have slightly lower sales "spikes" than New Jersey with 5 distribution days in a single block.

Somewhere along the line between Nevada (ranked #1) and Missouri (ranked #54) SNAP stops mattering to you because the distribution of fund is so dispersed through the month that you see no sales spikes at all.

74% of funding is distributed on 10 days or less and 10 day distribution can still generate, on average, a 20%-40% increase in sales $$.

BUT, some products are more responsive to SNAP distribution than others; some stores will have many more than the average 15% of their shoppers eligible for SNAP. So,

within the same State expect huge variations in the size of demand spikes. It may not be the average that's causing a problem.

Do you know which products, stores and dates are at risk? If not, how do you know how much demand went unfulfilled?  


SNAP Analytics (2) - Purchase Patterns