In the world of TV’s Parks and Recreation, few people are as concerned about their image and how they market themselves as Tom Haverford. The erstwhile entrepreneur loves to think up business ideas that range from the quirky (a submarine-themed club named “Club-a-Dub-Dub”) to the successful (tween clothes rental store “Rent-a-Swag”), to the plain bizarre (toddler cologne, anyone?).
Today, however, we’re focusing on Tom Haverford’s greatest failure — the “multimedia entertainment production conglomerate” Entertainment 720 (e720, for short). We’ll let Tom (and his co-founder, the eccentric Jean-Ralphio Saperstein) explain exactly what e720 does:
That….sounds like a lot. Unsurprisingly, having a scattershot business model leads to big problems, but perhaps none bigger than the fact that neither Tom nor Jean-Ralphio actually understands how to market themselves.
It seems like their idea is to simply spend inordinate amounts of money on “swag” but beyond that, they lack a clear vision about how to reach the right audiences, where to prioritize their efforts, and what are the best channels to market through.
The question is, could we hit the party switch on e720’s marketing by introducing machine learning? Let’s break down what e720 did, and how using machine learning could help them hone their strategy and potentially land an actual customer.
Let’s pretend for a minute that Tom and Jean-Ralphio didn’t doom their business from the start, and were looking to actually scale it the smart way. To that end, let’s ignore some of the more glaring problems (like buying a money printing machine and throwing around fake money at clubs) and focus on the things they actually did (mostly) right.
One thing Tom is great at is self-promotion. He knows how to make himself look good, and he loves swag. Indeed, e720 has all sorts of great promotional gifts — pillowcases, scarves, shirts, and even iPads. These cost a lot of money, however, so it’s crucial to make sure they get to the right place. Here is the first spot where machine learning models could help e720.
Let’s say the company is keeping track of engagement based on who is receiving specific swag. Some people respond better to pillowcases, it seems, while people who get iPads are surprisingly unlikely to engage any further than their free swag. Using observations like these, we can start gathering the data we need to build e720 their first machine learning model. To help bolster their predictive capabilities, let’s add external data to their new datasets. We can integrate a wide variety of sources, including:
Taken together, these external data sources and e720’s internally collected data could lead to a much better understanding of who they should be marketing to, and with which materials and strategies.
Next, it’s time to give Tom and Jean-Ralphio a harsh dose of reality. We’ll let Ron Swanson impart the knowledge:
The biggest reason e720 failed (aside from their inane spending) was a lack of defined vision when it comes to the services they offer. Tom and Jean-Ralphio want to be successful at, well, being successful but they don’t have the services established to draw in business. It’s not enough to simply offer everything — e720 needs to pick and choose what it’s good at. For a company that has offered to go as far as “installing a home entertainment system in your house” and then coming to watch a “Fast and the Furious” movie to make sure it works, it’s crucial to know which services could actually make them money.
The great news is that we can use the data e720 already (hopefully) have about their existing customers, and use the same external data we gathered for their targeting model. In this case, however, we’re going to be looking not at e720’s customers, but at the services they offer with a product blend model.
The bad news is that they don’t actually have enough of a track record to start using the model. However, let’s say they did, and make some extrapolations based on what we do know. We know they have a knack for planning major events. Check out this incredibly emotional tribute to the one and only Lil Sebastian:
Let’s say that additionally, they’re great at social media strategy. After all, Tom can’t even stop driving long enough to quit tweeting:
So let’s make a few assumptions to help us build a model in the absence of more solid data:
Based on this, we can start predicting where they’ll be most successful. Even here, though, the assumptions need some more context. So let’s enrich the model with the external data. We can compare e720’s top skills with demand in Pawnee and around Indiana to see which of their services might be in the highest demand. We can tap into social media for this, as well as search engine results and even online interactions. This will let us tap into the conversations occurring in real-time, giving us a gauge of demand directly from those that need services, as well as understanding which specific aspects are lacking or succeeding. This way, e720 can tailor their offerings to meet these circumstances.
Additionally, we can look at business data. We can examine the number of events held in different areas of the city to determine the biggest demand spots, and compare those to the search data we pulled earlier. We can also look at industry numbers such as total sales by event planners, or revenues for marketing agencies. This can let us know which services to prioritize, and also if there are seasonal patterns for e720 to exploit in terms of marketing their own company.
For all our fantasizing about helping Tom be a media mogul, e720 was doomed from the start for reasons unrelated to whether they used ML or not, and more to do with…well, we’ll let Ben Wyatt explain:
However, for burgeoning marketing organizations, data science is a great ally. Most organizations are already using analytics and already have massive stores of data. You don’t have to be afraid of data science, and you don’t even need to invest in building yourself a new team to handle it. There are already great tools that can do the heavy lifting and give you the insights you need to boost your marketing operations. It’s just a matter of finding the right platform, combining your data with the sources you need to get real value from it, and letting the insights flow.