A Q&A with John Pastor, Director of Business Technology at Pfizer
Editor’s Note: John recently presented at a Snowflake webinar. Below are highlights from the conversation (edited here for the blog format).
Pfizer is an American multinational pharmaceutical and biotechnology corporation headquartered in New York City.
Moderator: It’s been an amazing few years for Pfizer. Why don’t we start with getting us a quick picture of where Pfizer is in regard to data strategies for these complex times? Obviously, the Pfizer vaccine has been an incredible success.
Pastor: Thanks for mentioning our success. It’s been great. It’s been an amazing time to work at Pfizer knowing how big a risk we took, and there is a lot of pride in our company right now. It’s great to have played a very small part in it. We really had a very patient-first mindset even before the pandemic, so that’s even more solidified now. And our data strategy, and the various initiatives my team works on, as always, really support the high-level direction of the company.
And that is that Pfizer is really trying to be a trusted partner and a place to go for the best information. So how do we do this or how are we looking at this? It’s still a hybrid model. The face-to-face remains very important and probably always will be. But omnichannel and remote engagement continues to play a bigger role. We want to be like Amazon for pharma, which is tough in a regulated industry, so that’s quite an ambition, but we’ll get there. And to get there, we are asking what do our analysts need, and how do we use data and analytics to become better? Our commercial systems have always had embedded intelligence, at least for the last five or six years or so. We’re trying to make them smarter with AI and ML output, so really leveraging the analyst community.
Q: What sort of benefits have you seen since Snowflake has been put in place?
Pastor: We are running more cost efficiently and with better performance. And we’ve gotten out of the data-wrangling business—spending less time rounding it up and more time analyzing it. One particularly interesting area is with analytical data workflows and data sharing. With our analysts that are writing SQL or using apps that push down SQL, we ask them to take their data workflows and run them on Snowflake, so we can really leverage the analytical and processing power of Snowflake. And their pipelines are running faster and more consistently. So, not only are we saving time, but if something took 20 minutes to run today, we can be confident it will take 20 minutes to run tomorrow. Some of our open source platforms, we’ve seen jobs that run for 12 hours take 20 minutes, or even five minutes, on Snowflake! Huge benefit here because if an analyst can run 20 iterations in the time they used to spend for 10, they gain serious think time. And we are starting to test Snowpark and looking forward to that.
Q: Can you share a bit about Snowpark for anyone who doesn’t know?
Pastor: Yes, Snowpark is a capability that allows you to run various sorts of data science workloads, whether Java or Python, etc., directly in Snowflake. So you do not have to necessarily convert any of your routines to SQL calls because, again, this goes back to that core philosophy that we spoke about that you shouldn’t have to bring your data to some other utility to run your code to run your transformations, but rather be able to bring your code to the data where it lives. (Note: Snowpark for Python is currently in preview.)
Q: Can you discuss your commercial strategy for data sharing?
Pastor: The area of supply chain is one that offers huge benefits, obviously up and down the supply chain. Typically, this is done for free, so I’ll stay out of whether it should, could, be monetized, though there are those conversations. If you’re sharing things like your inventory positions, warehouse withdrawals, consumption at a pharmacy—there’s lots of benefit to sharing all that with your manufacturer. You get advanced information as to demand fluctuations. Now we’ve all lived, as consumers anyway, through a year and a half of empty shelves, shortages, inflation, and not being able to always get what we want. So clearly, there is a ton of mutual benefit to collaborating up and down the supply chain. And clearly the more we can share and collaborate on data, the smarter we’re all going to get about getting product to where it needs to be when people need it.
Q: Can you talk about ways that Pfizer is collaborating today, whether it’s with distributors, the payer community, regulatory authorities, and any ways Snowflake is helping that collaboration?
Pastor: We collaborate today or we couldn’t run our business. When we think of data collaboration, we think of other healthcare organizations where we trade data with, such as syndicated data partners, the co-promoter partners, the regulatory agencies, and even some of the more open data sources. It’s easy to just think old-school and say, hey, we need to go get that data and move it into our environment. But we are going through a big mind shift here; we realize now the amount of energy and cost in moving data (on-premises) is high for pretty low-value activity. We’re looking forward to shifting more effort into analyzing data where it resides. I don’t need to move terabytes of data into my warehouse—I can use it where that data resides by leveraging my partner. We’re getting access to data and setting up those data shares across domains at Pfizer and with our partners, whether it’s a syndicated partner or a co-promoter. We’re now having those conversations all the time.
Q: What sorts of new data are actually being made available by all of these different channels? And what is the management strategy for this data today?
Pastor: We have the capability today to collect everything. We shouldn’t be throwing data out unless it’s really totally useless. So we do take the data-lake-style approach, the data warehouse approach, where we collect it, whether we know yet what to do with it or not. The other piece is we really focus in the last couple of years on global approaches, especially around the omnichannel data. So when we collect and publish data, we’re trying to think global. The biggest project is our global interaction repository, which is a core model that stores all of our digital interactions, including the (sales) rep base. So we’re seeing that our analysts can use that data globally using the same analytical workflows, using the same techniques, and they’re not reinventing the wheel for each market in each region. So that’s a big push for us and we’re starting to see value there.
Q: How do you deal with the data residency and the data localization requirements and balance that with the fact that you have consumers who need access to this in a convenient, secure, and compliant way?
Pastor: Obviously, we know what we can bring data across and put it in different locations. A lot of times that’s just for location, just to make things a little bit faster for the data analyst, depending on where they’re working and what they’re doing. Our data strategy includes looking at what’s our Snowflake environment going to look like, what’s our enterprise data going to look like, and we approach dealing with all that knowing we want to put data in different places and secure it under one broad security model.
Q: Pfizer is a big organization, but do you have any tips for a smaller clinical organization as to where they should start leveraging data to find quick wins to increase champions internally?
Pastor: First, I’d look at what’s in Snowflake Marketplace and fire up a very small virtual warehouse and test it. And smaller companies should be thinking about partnering with a big company. Ask the bigger companies you work with about their Snowflake capabilities. You may be able to send Pfizer your data, maybe there’s an opportunity to collaborate.
I think this is worth expanding on. What you are touching on, John, is that even if you’re not a Snowflake customer, but you partner with someone who is, you have the ability to spin up what’s called a reader account for the context of data sharing. So if you wanted to partner with an existing Snowflake customer and we wanted to share data, you can spin up a reader account (with no setup or usage costs) where you can isolate the compute resources that you have to use, even setting up chargeback models. That way, you can dip your toe in the water slowly and get a taste for working with Snowflake and for what collaborating with a partner might be like.
To view the webinar, click here.