This episode features an interview with Anahita Tafvizi, VP and Head of Data Science & Business Operations at Instacart. Instacart is the leading grocery technology company in North America. In this episode, Anahita explains Instacart’s unique 4-sided marketplace and reveals how she has built and structured her data science team to fuel product-led growth. By understanding behavioral patterns and outcomes, her team helps guide product strategies that make the online shopping experience convenient, efficient, and personalized.
This episode features an interview with Anahita Tafvizi, VP and Head of Data Science & Business Operations at Instacart. Instacart is the leading grocery technology company in North America
As a senior executive at Instacart, Anahita drives key operations and strategic decisions across all company product pillars and ensures data investments are aligned with the long-term business strategy. She leads a team of over 150+ Data Science and Strategy individuals across all company product lines including consumers, shoppers, advertisers, and retailer products. Previously, Anahita was the Director of Finance for Google Commerce, Retail & Travel, as well as the Head of Finance for YouTube Ads and Head of Analytics & Data Science for eBay Ads. She is passionate about building high-performance data and strategy organizations with a focus on agility and impact. Anahita earned a Ph.D in Physics from Harvard University.
In this episode, Anahita talks about structuring her data science team to reveal opportunities for new efficiencies that guide Instacart’s 4-sided marketplace, her approach to hiring the leadership team and overseeing 150+ employees, and reveals recent data science initiatives fueling product-led growth.
Anahita Tafvizi is currently the Vice President and Head of Data Science & Business Operations at Instacart. As a senior executive of the company, she drives key operations and strategic decisions across all Instacart product pillars and ensures data investments are aligned with the company’s long-term business strategy. She is passionate about building high-performance data and strategy organizations with a focus on agility and impact.
“How can we make the experience of buying groceries on Instacart not just more convenient but also more efficient and delightful? To inspire product strategy, we spend a lot of time trying to understand patterns of shopping so we can build personalized experiences.” - Anahita Tafvizi
(01:17) Anahita’s path to data science
(03:04) Instacart’s 4-sided marketplace
(04:56) Structure of the data science team
(07:10) How business teams can unlock new insights
(09:51) Benefits and drawbacks of virtual teams
(12:45) Data needs of product-led growth
(14:30) Key data science techniques, tools, and skills
(16:30) Recent data science initiatives fueling PLG
(19:16) Instacart's data maturity
(20:20) Data access for business context
(22:00) Approach to hiring data science leaders
(23:30) Career growth paths in data science
(26:37) Increasing internal talent bench
(27:59) Driving efficiency in an economic downturn
(32:08) Key insights on grocery delivery services
(34:35) Takeaways
Announcer: [00:00:00] Hello, and welcome to the Analytics Edge, sponsored by NetSpring.
Thomas: The Analytics Edge is a podcast about real world stories of innovation. We're here to explore how data driven insights can help you make better business decisions. I'm your host, Thomas Dong, VP of Marketing at NetSpring. And for today's episode, my co host is Vijay Ganesan, co founder and CEO at NetSpring.
Thanks for joining me today on the show, Vijay
Vijay: Great to be here, Tom. And I'm really excited about today's episode, talking to Anahita. She's got great academic credentials, very distinguished and very diverse professional background, so we're really looking forward to this.
Thomas: Yeah, absolutely. Especially around today's topic, about fueling product led growth with data scientists.
Our guest today is Anahita Tafvizi. Vice President and Head of Data Science and Business Operations at Instacart. Instacart, of course, is a popular on demand service that delivers groceries in as little as an hour from local stores, where users can [00:01:00] place orders on the website, iOS, or their Android app.
Anahita, we are delighted to have you with us today. Welcome. Thank
Anahita: Thankyou so much, Thomas. Thank you, Vijay. I'm very glad to be here as well. Thanks for having me.
Thomas: All right. Well, Anahita, you joined Instacart in early 2021, and today you lead the data science and business operations teams. Can you tell us a little bit about your career path and what led you to data science and over to Instacart?
Anahita: Yeah, sure. I started my early career in a combination of academia and consulting, which helped me learn on one hand about Technical problem solving, and on the other hand, about the world of business and organization and leadership.
I moved to the Bay Area in 2011 and have since been in tech in a variety of leadership roles at the intersection of finance, strategy, analytics, and data science, and I've primarily focused on marketplaces, e commerce, and advertising fields. [00:02:00] Um, the last role I had at Google, I led the finance team for Google travel and commerce and shopping.
And that's where I learned firsthand about the challenges of achieving the positive unit economics in grocery e commerce. Um, I started talking to the Instacart team early on in the pandemic, and I very quickly built a very deep appreciation for what they have built, and the incredible mission of the company, which is to provide access to the food for everyone.
One of the most basic needs in the society. I'm also a working mom and I've been, and continue to be a very strong consumer of the product and have experienced firsthand. The value that it delivers to me and to my family. And then it was also in the pandemic. So, you know, like watching how Instacart and Instacart shoppers really enabled providing food to people who were sheltering in place.
It was really impressive to see all of that. So the combination of all of these, the strong belief that I had in the mission of the company, and then the really interesting and challenging data problems that is in the four sided marketplace that [00:03:00] we are in, really drew me to join the team.
Thomas: Yeah, so speaking of those data challenges, so Instacart has built a very unique marketplace here is what you guys describe as a four sided marketplace.
Could you describe that a little bit more for us and really drill into what the unique challenges that presents? For sure,
Anahita: yes. This is one of the most exciting part of our jobs. Instacart is, as you said, you know, frankly, one of the only four sided marketplaces that I know of, by which I mean, not just They're one, but four core audiences that we care about serving.
So the four sides are, um, consumers. So say Tomas, if you make an order on the Instacart app, then we have the shoppers, um, which are the people who pick up the groceries and brings them to your house, then retailers, um, which is a place that you order from. So say Costco or Sprouts in this example. And then lastly, we have our brand advertisers such as Pepsi, for an example.
Many of the decisions that we make in terms of our product innovation, or our [00:04:00] marketing, or our partnership impacts more than one side of this marketplace. Um, so let me give you an example. An improvement that we make in our batching algorithm, so batching algorithm is an algorithm that we use to connect available shoppers with, uh, potential orderers.
This impacts both shoppers and consumer experiences. So for example, the higher the number of orders that we put in a batch, the higher the economic opportunity for the shopper as it increases their utilization, but it may result in a delay in the consumer experience. So we have to always be thoughtful about, you know, how to balance.
these two different dynamics. And often we are able to create minimum situations for these different stakeholders. And that's obviously what we always strive for. But in the events that we aren't able to do that, it's very important that our team uses data to help articulate these trade offs between these different stakeholders and between these different metrics that we have at the company so we can influence and enable these decisions.
Vijay: So, Anahita, this is quite a challenging environment, you know, this four [00:05:00] sided marketplace. How is data science organized at Instacart to tackle these
Anahita: challenges? Yeah, this is a very good question. So, first of all, our data science team is a central organization. What it helps is that it enables objectivity, it enables functional excellence, and it enables professional and career development for individuals on the team.
But then we structure our team to map to each of... These sides of the marketplace that I talked about. So we have teams that are dedicated to consumers, to shoppers, to retailers, and to advertisers. And then our partner teams in engineering and product and design are very similarly structured. So as an example, we have a cross functional team that focuses on search experience in the Instacarts app.
So this includes a product manager, an engineering team, a designer, and a data scientist. And so the advantage you have of this model is, as I mentioned, the central team enables objectivity and functional excellence, but then it also helps because they are super embedded [00:06:00] in this each of these cross functional pillars.
So it helps result in subject matter expertise and seamless cross functional collaborations. One other thing that we have recently tried to do is what we call virtual teams. As our product has matured, what we have learned is that most of the important problems that we have actually exist at the intersection of these different pillars.
For example, we want to make sure that we are offering a product that's affordable to all of our consumers, and this requires partnership across all four pillars, right? So for example, a lot of the discounts that we offer to consumers, they come from our brand partners or they come from our retailers. Or as an example, if you find efficiencies in our fulfillment platform, then we want to pass those savings back to the consumers.
And then at the end, the consumer team has to figure out what's the right way to surface these deals to the consumer. So they're discoverable by them. So this is an interesting example that it actually like over, you know, basically goes across all of these. pillars. And for that, we created a virtual team so that [00:07:00] instead of relying on organizational, um, you know, structure, if you will, to solve this problem, we can bring experts across these pillars to work together and break these silos.
Vijay: So at Instacart, if a product manager or a growth manager, It has certain hypotheses on opportunities for new efficiencies. What is the process they follow to unlock these insights? Do they have access to self service tooling? Do they have embedded analysts? How do they go about doing that?
Anahita: It's a very, very good question.
So when we have like very common questions that we've had before these, we have all automated them and we have self serving tools that all of our product managers have access to primarily use Tableau and mode analytics. And the reason that we have chosen them is specifically because they have the ability to tailor and filter to the user needs.
So we have most of our common questions and stats obviously automated. Yes, our product managers have access to that. And then you have like more ad hoc, you know, strategic [00:08:00] insights or larger strategic insights questions. And so we have two primary touch points with leadership on that. The first one is our biannual planning process.
And then the second one is our weekly business reviews that we have with our C suite and with our functional leads. So just to give you a little bit of additional color on that, for our planning process, actually our data team is like very much in the driving seat and our planning really starts and our road mapping really starts with data that supports sizing and prioritization across different initiatives with similar methodology so that as a leadership team and as a company we can make the best decisions in terms of what initiatives we want to do and what are the things that we want to prioritize for the company and then from there we go to product road mapping.
So that's again, Incredible touch point for us that we have that like really helps us enable data driven decision making as part of our planning process. Um, and really drives what we do as a company. And then in addition to that, we have, uh, this weekly business reviews [00:09:00] that we use to, like, review key trends in data, whether it's, for example, impacting of marketing spend or product adoption, and then we use these to influence execution velocity, uh, or iterations on the product roadmap or, for example, launch decisions.
A good example of this is the way that we approach and dissect marketing spend on a week to week basis. Um, every week with leadership, we look at what channels are working, how our audiences are responding, and then we change, based on that, we change our targeting and our spend allocation accordingly.
And because of this, we're able to move very quickly and we're able to iterate through our insights on a more regular cadence with our, with the leadership. It's
Vijay: interesting what you said about how, for planning, a roadmap starts with data, which is not very common. It's a sign of a data mature company, so.
Yeah, you mentioned virtual teams. What are the opportunities and drawbacks to virtual team versus a more formalized organizational structure?
Anahita: Yeah, so this was the first year that we used them [00:10:00] and it's definitely taken us a while to build that virtual team muscle, if you will. We've had a lot of challenges at first because the teams are forced to rely a lot more on influence over direct ownership.
Um, and they have to navigate a lot of ambiguity, especially at the beginning when we were standing up these virtual teams. But, you know, there has been a lot of upside. Virtual teams allow us to be a lot more nimble. They enable us to tackle the biggest opportunities head on. They empower the teams to look at the problems more holistically rather than within the silos of their orgs.
And I think it's been really healthy and beneficial for us, especially for more mature products, as you said, and in a way they're actually less disruptive because we can pull the right people with the right skillsets into this virtual team. And we don't need to change reporting lines every time we want to kick off a new project or like go deep in a particular strategy.
You know, as you said, there has definitely been challenges that has been associated with that. And, you know, just. Maybe to share with you [00:11:00] some of the learnings that we have had, I would say we found that there are three things that really help set them up for success. The first one has been establishing very clear ownership and operating model for these teams because we can't really rely on org charts and existing operations.
We need to be It's listed about who's doing a lot, what are the milestones, what are the communication lines, et cetera. The second one is that we've also learned to make hard decisions about prioritizations. You know, virtual teams, like any other team, they only work if it's considered someone's full time job, not like something that they do on nights and weekends.
So this has meant for us that we had to deprioritize other work and then be very explicit about that and communicate that very explicitly. And then lastly, I would say empowering the team. Just like any other team, virtual teams work best when they're not top down and the teams feel empowered to make decisions and rely on their best judgments.
And this can be hard because the [00:12:00] mandate of the virtual teams oftentimes comes from the top, but it's very important to quickly make sure that the team takes the reins and has full ownership of the problem space. Yeah, it's
Thomas: really interesting, this concept of virtual teams, especially today. It feels like org structures are almost a thing of the past.
We need to be much more cross functional and customer centric organizations. It doesn't matter, you know, which function owns, it's the customer experience that matters at the end of the day. And so PLG, product led growth is this new emerging business strategy where. This really embodies that concept of the virtual team.
You need to have marketing, working with product, with sales, customer success, support, you name it. Would you characterize Instacart's new go to market strategy as product led? And then, how would you describe the data needs in particular for such an organization?
Anahita: And so it's actually interesting. I guess I'll build on the comment that Vijay made earlier.
So I would say rather than thinking of us as product led or sales led, [00:13:00] I would like to think of Instacart as data led. And you know, I'm sure you would think that I'm biased, but really pretty much everything we do follows the framework of understand, identify, execute, and then measure. And then really data shapes our understanding and prioritization.
And then it allows us to whether our hypotheses were correct. So as I mentioned in the previous question, we have these touch points, like our planning process, like really starting with data and really using data and consistent methodologies, which is one of the other benefits of having a central organization to help build these prioritization decisions.
And like really deciding what the company works on is like the data team is on the driver's seat at the beginning of the planning. And I would just use this as an opportunity to give a huge credit to our CEO and our COO, whom I report to, who have really pushed us to become a company that is. It's proactive in our insights and it's extremely data driven.
So it's like it really been that empowerment that we've been really able to transform the organization and us into a more data driven company.
Vijay: [00:14:00] You hear about product led, sales led, there's a lot of discussion around it and uh, you know, it's an interesting, refreshing perspective to say, look, we think of ourselves as data led.
Thomas: Yeah, we love that. Data. Data science, um, you know, both Vijay and I come from a, a data science formal background academically, uh, so it's very refreshing to see data science organizations emerge at many companies these days. So we want to drill in a little bit more here into the meat of data science.
Would you maybe describe some of the key data science techniques that you're using at Instacart today? And what skills in particular are you looking for as you build out your team?
Anahita: So I would say, um, the core skill sets on our team have are things like experimentation, causal inference, modeling. We also, obviously, as I said, we use a lot of data analysis and insights to help improve our core understanding of the experience and decision making and areas to improve in terms of your question about the tools.
Um, so we really ask our scientists. [00:15:00] to think about what is the right tool for the most important problems they have to use. So it could be a model that helps route shoppers more efficiently as they drop off order, or it could be an A B test to evaluate different versions of the product. In terms of our experimentation tooling, we actually have our own in house experimentation platform that we have built, and it has like numerous state of the art techniques.
Some of these, for example, includes power calculators, covariance adjustments to reduce variance and noise. Dynamic and adaptive experiments, and then various analysis that we have automated, for example, like cohort analysis. And then in terms of reporting, um, to answer your question, you know, very similar to other data teams, we have worked very hard over the past few years to streamline our reporting through clear dashboards that our stakeholders can use to self serve.
I think I mentioned in a previous question, we primarily use Tableau and Mode Analytics. for that, for those use cases, like fully automated. And like, you know, obviously have spent a lot of time in a stakeholder [00:16:00] education so they can be, um, empowered to use that. And then recently, uh, we've actually built some tools using generative AI that allows anyone in the company to.
Search for datasets and analysis and dashboards, because as you can imagine, there's like hundreds of them at any point in time across the company. Um, and previously we had to really rely on tribal knowledge to find those, but now we have these tools that like helps any user, like try to find that. And like, it's really helped with us and with the knowledge management across the company.
Thomas: Here we hear generative AI again, all the buzz. Uh, but you, you had me at causal inference. What are some recent data science initiatives around product led growth, uh, that you've led and really looking for, you know, those aha moments that your team's been able to unlock.
Anahita: Yeah, I'll give you a couple of examples.
We've spent a lot of time on and I'm particularly proud of. So the first one is, um, around, um, you know, we've like really spent a lot of time to understand our consumers this year, uh, and understand their habits [00:17:00] and their behaviors and how we can drive them more to our platform. And so one thing that we recently learned was that we are missing on an opportunity because we were neglecting our churned users.
And our growth strategy was primarily around new activations. Something that we learned, which was really interesting, was that a lot of the households in the U. S. have used Instacarts. And in fact, I've actually used the Instacart in the last year. So while activations continue to be of a lot of importance for us, we created a more balanced growth strategy around what we call resurrections, which is basically bringing these users who have previously tried Instacart back and helping them establish routines with our products.
Another example I would give you, which I'm particularly proud of, is the work that we have done to support our brand advertisers with understanding the effects of their advertising campaigns. You know, it's actually interesting. One of the things that I've loved about Instacart is like, we have transaction data.
We have like a, Very strong [00:18:00] sea of actually consumer data that we can help our brand advertisers really measure the effectiveness of their marketing campaigns. So if you think about it, you know, if you're an advertiser, like Pepsi, for example, I think we might be actually one of the only platforms, online platforms that we can provide that transaction data.
to these advertisers. So one thing that the data science team actually drove was we built, you know, tooling and methodologies that we can actually now measure the incrementality and the incremental value or incremental sales that you can get from each marketing campaign. And that really helped advertisers understand the true incremental return on ads.
Spend that they have been spending on our platform very directly, which is like, again, one of the only platforms that can drive that in the food category to them. So this was really powerful because it really helped build that trust with the advertiser that, you know, they can now trust our products a lot more, trust the budget that they spend on us.
Like they feel a lot better and it really strengthened our [00:19:00] partnership with them. And I use this example a lot because I love that. It's a great example of. How not only do you use like a very innovative techniques and the state of the art modeling in the industry to do that, but also it's something that drove really strong business results.
Vijay: Every enterprise talks about being data driven. You know, this is very common, but there's varying degrees of maturity. There's a whole spectrum of data maturity. How would you describe Instacart's maturity as a data driven company?
Anahita: Another thing that I love about this, I guess, the state that we are in is that we are in a very exciting middle ground, I would say, between having the maturity of tools and engineering that we can build great products on while we are still at the phase that we can, you know, still there are lots of insights.
to be unlocked. There's a lot of opportunities for what we call like zero to one problems to be solved or, you know, opportunities to have impactful strategic findings around which we can quickly and effectively build products, you know, as opposed to being in a, you know, in a more mature product, if you [00:20:00] will, that's You know, you're basically just optimizing around a mature product.
So it's been really a sweet spot, I would say, for the data scientists, where there's still a lot of opportunities for us to have impact, but then we have the tools and engineering needed and like I found the strong foundation of data that we can, uh, that we can use.
Vijay: You probably have lots of sources of data, and as a data team, you need access to all of the data in the full business context to make good business decisions. Can you describe Instacart's strategy for centralizing data in a warehouse and how This has enabled your team to quickly identify opportunities to build better products.
Anahita: Yeah, for sure. I mean, beyond everything I said, um, just now about how we leverage insights and analytics across the company to inform decision, we also use data very heavily to power our product experiences and to support our machine learning engineering as well. And so we need, to your point, we need a very strong data foundation and data [00:21:00] infrastructure to enable all of this.
So our data infrastructure team, which we partner with very closely, they have been investing very heavily in building a world class data platform to support all of these activities. It's actually been a really great work that's recently been recognized by Snowflake as industry leading as well. So to answer your question, our data platform is built on a snowflake and data bricks.
And then some of the principles, um, that we have around what we want in a more fast data infrastructure is for our data to be secure, trustworthy, and, you know, ultimately easy to use. Um, and so given that, along with the foundational support that's needed for ingesting and transforming, you know, petabytes of data across hundreds of thousands of tables, The team has also made extensive investments in different areas, such as, you know, data governance or compliance or privacy and discoverability.
Vijay: During your tenure at Instacard, you've more than doubled [00:22:00] the size of your team, and you've hired many of your key leaders. What is your approach to hiring leadership
Anahita: teams? You know, I, I believe I actually have hired or promoted, uh, all of our care and data science directors. We've also hired and grown a very strong cohort of technical individual contributors, which is just as important, uh, on the team.
And so the framework that I use to hire these leaders for our organization is really around looking for. specific set of traits that I think is very important. So the first one is being a bar raiser in terms of technical and operational excellence. The second one, uh, I would say is humility and being a lifetime learner.
The third one is curiosity, intellectual honesty.
And then last, which I think is also very important is the ability to build a strong cross functional partnership while holding their ground on hard data, right? Because there are always [00:23:00] times in every data science leader's career that you have to deliver news to your cross functional partners, which may not be what they want to hear, or they may be, you know, like not supporting the strategy.
And so I think it's very important to be able to continue to build on the very strong foundation of partnership. Thank you. Be able to, like, keep your grounds with hard data, but then be an advisor to your cross functional partner on how to evolve from there.
Vijay: Yeah, I love your first item, bar raiser. You know, I think it's so important that every person who comes in and raises the bar of the team.
Great. How have you seen the team evolve as you've scaled in terms of the type of data scientists and skills you have built on the team?
Anahita: Look, I mean, when you're building a business and the ideal data scientist is what they call full stack data scientist. So you might have heard of the term generalist data scientist because the teams are small.
They need to be really nimble. So you need to hire people who have like really broad toolkits and they can quickly expand on the job. But then as the team grow, like, you know, for example in our case at Instacart, We [00:24:00] focus more on a specialization. Um, so for example, we've hired industry experts in marketing science or in advertising or in a statistics and causal inference.
And we've also tried to facilitate growth of these different team members based on their strengths and interests. So, for example, some data scientists want to operate at the intersection of data engineering and data science, you know, while others may more prefer to operate at the intersection of data science and machine learning engineering.
And we really try to lean into this diversity because in the aggregate, it means that we can answer more types of business questions with higher fidelity, and it creates clear growth paths for individuals with different interests on the team. You know, the advantage of generalist data scientists is that you reduce the number of handoffs.
And so it's a lot more efficient model to have. So again, you know, we've had a lot of that early on and we continue to have majority of our data science cohorts be generalist data scientists, but then leaning into these specializations have helped like really complement that as well. And so we [00:25:00] try to always balance that.
One challenge that we have actually recently faced, which is interesting to share is that creating clear job expectations when you have these different differences, I guess, as you said, for example, we found that in calibrations, we were comparing data scientists who have very different jobs and superpowers.
You know, all are equally important to the team, but then the contributions look different. And so it's hard to draw a fair comparison, um, you know, also the risks creating the false perception that we only regard a specific archetype of data scientists or we want to disincentivize a specialization. So to help solve for this, we recently personas.
to acknowledge for these different flavors of data science work at Instacart. So, you know, you can imagine complex system experts, strategists, problem solvers. technical domain experts, data architects, like all these different personas that can be in the data science team. And so this has really been [00:26:00] helpful framework for us to use because for individual contributors, and these personas have really served as a tool for them to have career conversations and professional development conversations with their managers.
And you know, like, of course, we see Still expect everyone in the organization to create a baseline expectation across all of these different skills that's needed. But then the personas have allowed us to cater development plans to individual interests and strengths. And then for managers, these personas actually have been an interesting tool for more deliberate org designs, uh, and then for more thoughtful comparisons during calibrations.
So
Vijay: in addition to hiring top talent, how have you approached leveling up your, uh, talent batch internally?
Anahita: You know, it's always been very important to me to grow and invest in the top talent within my team and only complemented when needed with the new industry talent. Of course, when I joined Instacart in the early days, you know, we were a small team and we were growing very fast, so we needed to hire at that time.
But, you know, as we have been [00:27:00] in a more. Maybe in the last year that we haven't grown as much. It's been really important for us to like really grow from within. So some ways that I would say I foster this is through being very clear about my philosophy and performance culture within my org. You know, I really believe it helps.
motivate top performers. You know, I believe as a leader, it's very important that you're clear about your top talent and you invest in them proportionately. And it helps motivate them and retain them over time. And then, you know, propel them forward to where they want to grow in their, in their own careers.
I'd say another way to help level up talent is through consistent stream of feedback. You know, that has to be both positive and constructive. And I believe the best type is one that's. Direct and in real time, so that it gives individuals an opportunity to adjust quickly and learn. And then, you know, I would say lastly, I try to be very decisive and follow through with my decisions, both within my org and with the cross functional partners, so that it helps create that trust as well.
Thomas: Alright, so, Anita, you've [00:28:00] obviously now pulled together a world class team. Uh, Instacart. So let's switch gears a little bit here. Now that you have team in place, um, company has passed your hypergrowth stage at this point, and like all of us in industry now, we're all facing a vastly different macroeconomic environment.
You guys are calling it the year of efficiency ahead. How have you driven efficiency within your org during this downturn? I know you talked about JNAI, um, earlier in the show. Can you speak to some of that work? Yeah,
Anahita: you're right. I mean, it's definitely been an interesting year, and I think it's apparent that the market dynamics and business requirements have really shifted.
So as you mentioned, we've definitely found efficiency through improved tooling and technology. So, you know, we want, obviously, everyone on the team to be working on the projects that are highest impact to the business. But the reality is that any data science work fills a large number of operational tasks.
So we've tried to really build ourselves out of that part of the job through improved knowledge management, through AI that you talk about Thomas, and [00:29:00] then, um, better self serve dashboarding, which I touched on earlier in this conversation, you know, it's obviously, as I said, it's like also comes with an investment in a stakeholder education so that they can feel supported and equipped to handle more of their own ad hoc work.
And then I would say, in addition to tooling, the other thing we have done is like really thinking about Deliberate org design and people management that really help us find those efficiencies as well. So, you know, let me give you an example. About a year ago, we merged our data science team and our business operations teams into a single team.
You know, while each of these functions have a very unique role for the business, the distance between them Was creating duplications, right? And so we had, we had a lot of work that we're like kind of in this gray area between these two different functions. And so by bringing them together and really clarifying the mission of each team and like the roles and responsibilities between each of these teams, we really like help us find a lot of efficiencies.
there. And then even if we don't [00:30:00] necessarily do org changes and bring the teams together, it's been really important for us to work with these different functions. But there is our finance team, our data engineering team to really help clarify roles and responsibilities and make sure that we are removing any redundant work.
I'm sure you've heard a lot of tech companies talk about it. In this economic climate, there's a lot of overlap in what teams are doing, and it's not creating only inefficiencies, it's also creating, uh, unnecessary politics, um, right? And so, I would say we fortunately haven't gotten in that phase in hiring because we've always tried to stay lean, but we still pushed ourselves to find those spots where We can create more role clarity and we can reduce duplication.
And I'm sure you can imagine these conversations are hard, right? But they are really worth it. And I think they're very necessary in this environment. And then I'd say like, also, um, the other thing is like really investing in the team that we had. You know, it almost felt like overnight my priority [00:31:00] shifted.
It's from hiring and growing the team to really up leveling the team that we've had, which, you know, honestly was equally, if not more exciting, you know, we hired a lot of really talented people at various stages of professional growth. And by focusing on this, what we call talent density, uh, we've tried to create meaningful career paths for each individual on the team and increase the impact of them without increasing the absolute size of the team.
If I look back, I would say we've seen actually more I'm just more insights, more experiments, and more foundational investments in the past year compared to the year prior despite the fact that our headcount stayed flat, or even maybe very few folks down. Yeah,
Thomas: it's so admirable what you've built there at Instacart and the role and impact that data science is having on your team.
So I was just thinking, you know, this Instacart, you know, we can all relate. To what you've built, we've all navigated the pandemic here. And in the last few years, we're forced to seek [00:32:00] out healthy food options. And Instacart was an obvious one at our fingertips, uh, within our apps. So maybe you could leave our hungry listeners with one nugget of insight from data science that we can all think of the next time we're in the app.
Anahita: Yeah, for sure. You know, one thing that I spend a lot of time to think about is that how can we make the experience of buying grocery on Instacart not just more convenient, but also more efficient and more delightful and to inspire products for the strategy. We've spent a lot of time trying to understand patterns of shopping so we can build personalized experiences that balances this efficiency.
Right? So we can see an exploration at the right time for the right user. Some of these patterns are, I guess, I'm sure you can imagine it's pretty intuitive, right? So like, if you think about it, um, consumers that are, um, you know, ordering in the morning, for example, are like more focused on like an efficient, more focused on efficiency.
They want to quickly create their shopping list and make an order, whereas there is more ordering of salesmen. Snacks or ice creams, like later in [00:33:00] the evening, for example, or, you know, consumers buy select group of items together. You know, if they buy hamburger patties, they're likely to also want buns and pickles.
Um, these are the type of insights that really help us build a very personalized and intuitive product experience and the type of work that we do in partnership with our product partners to help them really build these personalized customer journey features for our consumers.
Thomas: Yeah, that's really fascinating.
I know that I should never order anything when I'm very hungry.
Anahita: For sure, yes.
Thomas: Well, thank you so much for joining us today. This has been a very fascinating conversation and especially delighted to see. Again, the role that data science is playing at Instacart, of course, a company that is very well respected in the industry.
You guys have risen to prominence during COVID and post COVID and kudos to everything that you and the team have built over there.
Anahita: Well, thank you so much for having me, Tomas and Vijay. Also, very, [00:34:00] uh, you know, it was a pleasure talking to both of you. And thank you for doing this. I think it's very important to really, you know, up level the industry understanding of what the data team can do, you know, not just within a company, but it's...
to the extent that we can do
Vijay: this thank you, Anahita. This was very insightful, and it's going to be very valuable for our listeners. Anybody that's a leader in the data science space, I'm sure will benefit hugely from listening to this episode. Thank you so much.
Thomas: All right, wow, what a, what a fascinating conversation we had here today with Anahita, which obviously a great leader and really helping to promote.
The, uh, discipline of data science and the importance of that. So Vijay, what were some of your key takeaways out of today's episode?
Vijay: Yeah, first of all, I think fantastic episode. Uh, clearly, uh, she's a thought leader. She's built a fantastic organization, you know, great product, great brand, done some phenomenal [00:35:00] work, you know?
So I think that she's built a very data savvy organization. Uh, I think the thing that struck me is how they think about themselves as a data led, right? You know, we're not. Product led, sales led, or data led first, right? So, and this idea that data engineering has a seat at the table very early in all the planning process, you know, as they're thinking about product roadmaps and things like that, which is, to me, a sign of a very mature data savvy organization.
Thomas: Yeah, and I had probably two takeaways from the conversation. There's probably many, many more I can think of, but you know, as she was talking, there was one recent quote that I heard recently that really resonated, and it's that the data teams are actually growth teams, right? And with these virtual teams that they've created around, you know, the different sides of the marketplace, but representation from product teams and growth teams, design teams, and the data team, you know, really sets themselves Kind of around this unified alignment around growth.
And she had that great example around resurrections. It's the first [00:36:00] time I actually heard that, right? In the PLG world, we have the pirate metrics frameworks around acquisition, activation, retention, so on and so forth. But we have another R, or another synonym for one of the R's around resurrection. I thought that was, uh.
Interesting.
Vijay: You know, the virtual team is interesting. You know, virtual teams are hard in large organizations, you know, so it's also a sign of organizational maturity, right? In putting together virtual teams, because virtual teams are the most effective ones. You know, a lot of these things, particularly in a complex.
A, you know, four sided marketplace type environment that they operate in. You need people from different functions to come together and make decisions. So virtual teams are critical and, and then to pull that off, you know, at scale, uh, routinely is something that speaks to their organizational maturity.
Thomas: Absolutely. And actually going back to the data led point, the only way this is possible is the data strategy that they have in place, where they've centralized all of their data [00:37:00] into their cloud data warehouse. She talked about Snowflake, recognizing the efforts they've made. She also mentioned Databricks.
And then she had this fabulous example of the data science team combining the transactional data with the behavioral data to actually measure the incremental value of their marketing programs. Right. And it's just like when all that data is there and the data team can have access to all of it, bring together all the relevant context in their analysis, that's extremely powerful.
And I think something for data leaders to take away from this conversation. And that
Vijay: probably reflects in the thing she was talking about of how they want to provide personalized. Delightful experiences. And to provide a personalized experience in, in your app, you really need data from many different sources, right?
Including your, you know, marketing and your partners and your advertisers and, and then so many sources. So probably is, um, related to what you're saying that unless you have all of that [00:38:00] context in the data warehouse, you cannot achieve that level of, uh, personalization and delight in the product. All right.
That
Thomas: concludes today's show. Thank you for joining us and feel free to reach out to Vijay or I on LinkedIn or Twitter with any questions. Until next time, goodbye.