Transcript
Reva: You’re listening to Next in Consumer Products from Publicis Sapient. The podcast that shares insights on unlocking what’s next in digital and consumer products.
Reva: Consumer products companies have been trying to crack the case on data for quite some time now. The challenge can really be looked at in two ways: first, how CP firms acquire data and 2nd, how CP firms use data to drive growth. The second piece of this puzzle has proven to be the most challenging for many firms; it turns out leveraging these data into actionable meaningful insights to drive business decisions has started to become a key challenge for CP firms. Essentially, now that they have the data, how should they use it?
In this episode, we're going to discuss the critical components required to build an effective insights organization in CP. Joining me today are Daniel Liebermann, our Insights and Data Lead at Publicis Sapient and AJ Dalal, Data Strategy and Science Lead. I'm your host for the session, Reva Bhatia. Now let's dive in.
Daniel, I'm going to kick things off with a question for you. So insight-led organization is a heavy topic and can mean different things to different teams. Can you first unpack what an insight org even is and how it actually might work and dive into what some of the barriers you see CP firms are facing that make their efforts here quote unquote not work?
Daniel: Happy to dig into that. So an insights org I think most simply takes data and moves it into decision or insight, and that can be in like in almost any area. So it's kind of like a factory and that's, that's the value chain, and so I think most of these organizations function from the perspective or shift function from the perspective of being service organizations and having the perspective of what is it that their customers, their internal business customers, want. If their raw material is this data, there's a number of different things that they can produce with it, but just like any market they have to figure out what people want to buy and that could be a marketing organization that wants to know more about its consumers, that could be an HR organization, it could be a finance and operations organization… all parts of the business that can benefit from knowing more and having greater insights and taking better action. So insights organizations, especially if their centralized, whether they be called data science or analytics or insights, tend to have this mission of turning that data into action for their internal customers. You mentioned the second part of your question was some of the barriers that CP firms have had, I think traditionally compared to other industries CP firms have been behind for a number of reasons: they haven't always been the most data forward either in their products or their ways of working compared to financial services or retail, in particular, and then they certainly have many different world businesses. To take a step back to say they have within one business many different businesses many different brands which can create many different silos many different ways of working, and then of course there's the disintermediation from their customers in many cases, so they struggle sometimes to get that customer data at the level of, for example, a retailer and that's often made it harder for them to know exactly what it is that their customers want, who they are, and how to engage with them.
Reva: Interesting, so now that we've talked a little bit about what some of the challenges are, I'm going to kick it over to you, AJ, to share a little bit about what some of the key building blocks are of a successful insights organization.
AJ: The building blocks that a CP organization needs to focus on to develop a successful insights capability are data, logic, experiences and measurement. And data can mean a lot of things to a lot of organizations whether you're talking to an engineer, a data scientist and data analytics, and like Dan mentioned in the previous question a brand manager executive to every day operator. In its simplistic terms, data is around the combination of your first and third party data to create some type of profile, whether it's a product profile or a customer profile. Underpinning the data pillar is how you collect your data in a way that you can standardize across the organization so it generates scale, and what I mean by that is making sure that you think of all the different touchpoints your consumers have and your team have in thinking about a way to standardize or name the ways that you captured the data somewhat similar. And the reason for that is as the data is being captured and collected, integrated which means move through the organization and then put together and cleansed that whole process speeds up. And when it speeds up because you’ve standardized the first level of your layer, you generate what's one of the biggest outputs to me, which is confidence in the data pillar. And we've seen that whether it's in CPG or any other vertical is if you don't have confidence in your data the rest of the organization will suffer in adopting that data. The other three which I'll go relatively quickly in are logic, experiences, and measurement. And, to me around logic is how how do the engineers or data scientists create a model to resolved identities, and that is just saying if someone comes from a media website there they come from a third party cookie, if they come to the website you now have a first party data collection, if they come from an email another first party, but the data that's being collected are a bit different. So how do you resolve that and once you resolve those IDs that gives the basis for the engineers to to do more advanced types of modeling, like churn modeling, propensity model, predictive models, etc. Under experiences, I think there's two big pieces that I call out: one is making sure that an organization syndicates these audiences or these outputs that come from the logic layer into the appropriate channel. And a really great example I gave to a client last week was: you've got this list of individuals you want to email so you send the email out, but how many times does an organization send that same list—obviously clean it up a bit—but also send it to your media channel, and that to me is one of the key elements of an advanced organization within the CPG space of really capitalizing on your data. Because you what you're doing there is you're sharing with media what you're doing in email, so if someone responds to an email or doesn't respond the media channel can be told to see whether they need to do any type of retargeting or suppression or look-alike or anything like that. And lastly under measurement, there's obviously a bunch of dashboards, attribution, lifetime value, all that type of stuff, and the only thing that I would say under measurement that I think is very important with this pillar is thinking beyond just the simple KPI of engagement rate, conversion rate and measure that against something that can drive insight—which is a conversion rate versus an expectation, a conversion rate versus momentum, so you can actually be able to see whether you need to increase spend, decrease spend, change the optimization, etc. And if a company focuses on all these four starting with data leading all the way through to measurement, to me that equates an insight organization of your future.
Reva: Awesome, AJ. And you just did a great job at describing what you know the Golden panacea would be in terms of end state for CP firms to strive against, and I know, Daniel, on the previous question, you started to unpack what some of the challenges CP firms are facing that make their efforts quote unquote not work as I said. But you know I'd love for you to dive into what some of the key challenges or key business questions are for CP firms that they need to address to really get started on this journey. I'm conscious that many already have efforts underway, but I'd love to unpack a little bit more what some of the starting points might look like for CP firms to build what we're describing as an insights- led org of the future.
Daniel: Yeah, no, happy to answer that and I'm going to tackle that from the organizational side 'cause I think AJ just did a great job talking about the technology and data side, which is the necessary fundamental core, right? You have to have the data, you have to have standardization and governance in the data, you have to have the tools and technology to do that. I think if you're looking at setting up an insights org though a couple things that you may not immediately think about when you're looking at technology licenses and data collection are back to this earlier point I had of what are the insights that you plan to deliver and also how are you going to manage that. So there's some really fundamental organization and operations questions most primarily on decentralize d,o you decentralize, do you matrix that staff and where do they live in the organization? The right answer really depends on how your organization works and how you're going to intake requests for insights—meaning do you need deep subject matter expertise such that you want to have your insight staff sit within their embedded business units and support your team, or would you rather have more capability focus in which case you’d have a straight insights organization where your insights analytics data science staff was set. I think those questions about where exactly does this belong in the organization are key and they are answered by figuring out what those key questions are. So certainly there's primary ones around how to market, how to do media, how to do owned marketing CRM email, etc., but then there's also these additional insights that you would hope to get around your product and around your processes, and those may hit the marketing organization, but they may hit other parts of the organization as well. And so types of questions you want to ask are: Who are these groups within the company that could value from knowing more about consumers or other data that we have? What are the business problems that they're trying to solve with that data? And then I think, most importantly, if you think about this customer service mindset is, how do they want to be engaged with? Do they want to have a tableau dashboard that they can check on their iPad once a week and it's all hands off? That's one way to do it. Do they want to be in the trenches with you and roll up their sleeves and understand exactly what how you're generating insights and what other insights they could generate? Are they forward looking, are they risk adverse? All these things, this people side element, is so important 'cause what I found at least having done this in the past is usually you can get the data right at some point and usually you can get the technology right, but if you fumble defining the problem that you're going to solve and if you fumble the adoption of it, and that tends to be actually a person/people management and design of how you're going to tackle the problem question, then it’s content to fail. One other thing just real quick that I think is important is quality control and management, and I know not sexy at all and definitely very boring, but you tend to find, especially as you're moving quickly to stand up these organizations without any previous precedent, that there's actually a lot of risks, reputational risk, and just the risk of making the wrong decisions if you're not controlling for the quality of the analysis, the quality of the data that's coming in, how you define your KPIs or other measures, even how you handle that delivery—if you have internal customers, does insights organization, making sure that you are updating the dashboard before they have a Monday meeting or that you're actually pushing the results of your machine learning or decision model in advance of when they have to make their decision. And a lot of those things are very separate from the core of what it means to be a good data scientist or analyst or insights expert. I think for any organization looking to get started I would focus on those two areas primarily.
Reva: Cool. Excellent, so it's interesting Daniel that you described the phenomenon where defining the problem is often not what you know a lot of companies are doing before they start collecting data and it's created this problem where many firms have been like data junkies. They've just been collecting data left and right and they don't really know what they need it for and what it's intending to solve, and so I think perhaps taking a step back it’d be great for you guys, we've gone through some of the academic examples of what folks should be doing, but it’d be great to talk a little bit about how you've seen this come to life. So, I'm going to ask each of you to just dive in briefly into some of the best examples of insights orgs that you've seen and how you've seen it deliver value to clients and CP firm specifically.
Daniel: I'll let AJ go first, but just real quick you remind me of a good anecdote from my past consulting experience, but I had a client to your point about attractions of like the tiny object and maybe the cool being the enemy of the state sometimes, I had a client who was really excited to be doing natural language; I said what is it that you're going to analyze, what are you going to use it for? And they said: word. I was like that's kind of like telling me you have an incredible new capability which with is math and or statistics and I said: what are you going to use, what are you going to analyze with that? And they said: numbers. So it's such a good point that it really is about the impact and not so much the what did you do, did I use AI, did I use ML, am I in the cloud, am I using a serverless architecture? All important matters less than the actual outcome.
Reva: Yeah, it's not about checking off a box and doing these things just because it's what all the cool kids are doing.
AJ: Perfect. I’ll give an example. So without naming naming a client, but in the CPG space, so I was on this client for approximately 4 years earlier on in my career with Sapient and we help them stand up their architecture, we talked about the pillars around data logic experiences in measurement—so Fast forward into the application of all that type of stuff there was a project that we did where we married a lot of this quantitative data with a lot of qualitative data, so the research team came in and created a whole bunch of personas and by the time they were done there was about 14 of them. What we did with confidence was we worked with them to identify the behaviors that these personas would do and then we showed the numbers matched to them, so now these personas came to life. The second thing that we did that rallied from the executive board all the way through to the working teams is we married those two together to represent how customers navigated through their entire digital ecosystem, whether it was media, whether it was on the website, whether it was in email or any type of offline data experience that they would have such as an event or if they were they were partnering with a retailer, etc. The value out of it outside of the numbers in the data was the operational efficiencies that came out of it, so with this dashboard that we created we understood leakage points throughout the behavioral driven personas and as we understood where they were leaking we could then work with the user experience team and the technology team to then say what were they experiencing during this time and where are you focused your efforts on to reprioritize their next quarter of focus. So the CEO CMO CIO was able to shift engineers in there and their technologists to have the team focus on middle of the funnel say instead of bottom of the funnel. They were able to change direction and creative or events or etc., and the outcome of it was, without again sharing too much, was a 2X lift over the course of a year in terms of sales Attribution. So it was the catalyst, in my opinion, that was part of the catalyst. There's alot of things that they were doing right that helped drive that organization and by the time that I was complete their analytics team or data team that had qual quant folks grew from 10 to just under 100 by the time me and my team were done after those four years.
Reva: It's hard to not want to invest in an organization that's helping you bring in two X your values so that that makes sense. I'm curious you walk through this example and particularly the upfront work of doing that persona build. Was that something that was considered to be a point in time effort that they would revisit on an ongoing basis? Like, was there some level of strategy behind how often they would need to scrub their qual in order to continue to make sure that they are using data and insights as meaningfully as they as they can and should?
AJ: It's a great question. Yes, upfront there is definitely a strategy in terms of the life cycle of this research for the specific client it was every six months that they would revisit the qual and they would… the words that we were using we would conduct just a sniffer test, so it wouldn't be a full blown qual analysis or research with kind of customer interviews and accept product testing and all that type of stuff. It would just be in areas that were not yielding the results that we wanted or that we were expecting so we would just go back to that group and just under try to understand it a bit more. So things that were performing very very well we left alone, things that look like they needed to be optimized or we looked at it's what we did in that occured on average every six months.
Reva: Awesome. Thanks, AJ. Alright, Daniel, dazzle us with your example.
Daniel: Yes, I actually have two examples that come to mind so both sort of past experiences of mine. The first one was for an organization that was just starting its journey into having sort of an insights organization and had different pockets of insights capability and lots of questions and was was struggling to pull that all together, so actually what they did was build a central platform, sort of an analytics clearinghouse insights clearinghouse if you will, where different people across the organization from any function could propose insights questions basically and groups of analysts across the company would actually solve them for the people asking questions. And so this is really cool because it actually was able to load balance without changing the organization the requests that were coming from other parts of the business that were then triaged prioritized and then matched with these analysts who otherwise wouldn't have found these questions to build insights. The analysts also had profiles and so they started to develop their own reputation, so you would have like your star analysts within the organization, right, who answered the hardest questions and in the shortest amount of time and in these programming languages and use these Python packages and did things that no other analysts did. And in a sense there was the gamification of the process which got even more work out of the, this distributed insights team than they would have otherwise. There's also this really cool metadata effect which is you could measure what people were asking, so to the earlier point around like what is the demand for insights in the organization, you actually had a record here of which businesses are asking, what questions are they asking, how did questions look similar, what datasets are needed across these, and suddenly you built a basis for understanding of what the organization actually needs to solve for. And you can do that from basically like a prototype, right, sort of a test and learn to figure out what you have to solve.
The other example that comes to mind, this is actually working with a client in on the IT side of the organization, that a very large and potentially very uninteresting database, however,they had seen that they actually had lots of value in the data so I had worked with them to start thinking of this as less of a IT cost line item and more of something that can actually generate value within the organization and to start thinking of that database as a product and themselves as a product manager and the data that they have in there is something that they could provide as a service, and then they actually matured from just owning this data source to becoming sort of the insights broker for the organization. They offered some self service capabilities so they allowed people to access it directly through an intermediate link where people could self explore within the organization, they started to build dashboards on top of it that they used to serve different business units, and then for more complex problems they actually developed their own team that would go and sit with different business units and collaboratively solve big challenges that were dependent on this data. So they went from a very unsexy place within the organization to a much more interesting place at a time for them to where they were potentially facing budget cuts and needed a way to further justify their value.
Reva: Cool. Awesome. Thank you, both. It's great to see how these types of efforts can be brought to life. So, a light-hearted closer. What’s the most interesting piece of insight you have observed from one of your clients that wouldn't have been discovered if not for a great insights org? I am going to open it up to either of you to take this one.
AJ: So, the example I'll give is outside of CPG and it's in retail and it's for one of the larger retailers in the world so we ingested over 2 trillion rows of data and that data the organization had a lot of confidence in the way it was collected, and teams work very collaboratively front of house, back of the house, and the insights that came out of it I thought the one that was very fascinating was: if someone, actually there's one of many, but imagine this retailer also sells probably 50,000 products, so if someone comes to their ecommerce website and they went to the camping gear section and they added something to cart they knew that this family was planning on going camping. So they would do a recommendation rule on a product to recommend, and typical companies will either do one of two things: they'll recommend other complimentary camping products, so if you bought a tent do you want some oil for the fire, do you want some food or marshmallows, right, or other camping products like a sweater or stuff like that, but what they would recommend was batteries. So they figured if they needed they were going camping they would need batteries, so they would take him to the battery section and they would sure enough would buy batteries. And then the next product that they would recommend would be games and they ran him through this journey and the output of it was twofold: one was a 3X lift in the number of products that they would add to cart and purchase and 100% lift in the average order value because the products that they were recommending were based off of that two years of data of two trillion plus records. Now here's where this client can also monetize it: so if you take them to the battery section you really have two big players in the world, you have Energizer and you have Duracell. And at this level retailers sell ad space, right, they monetize their side as well, so now you have a really strong data-driven monetization opportunity by negotiating deals with your suppliers such as Duracell or Energizer and saying: hey if you'd like to advertise on our website, place your banner, we're going to charge you a higher CPM because of these results that the back office data can be we can attribute sales towards. So not only did it benefit the retailer from a sales perspective and I'm sure customer satisfaction perspective, but it also helped them monetize their website from their suppliers.
Reva: Cool. That was a great example. Thank you.
AJ: I could share another one.
Reva: Keep them coming.
AJ: So, and I think this is not with the CPG client, but I think it would resonate really well for CPGs. So you kicked it off by saying they're requiring third party data and I'm sure these CPGs are looking into DSPs / DMPs, creating second party data sets, they’re probably exploring other outside sources for enrichment and all that type of stuff. And they're doing all of this because they want to spend their media dollars a lot more efficiently and get a broader reach. For a client of ours in another industry which I won't really share the industry just 'cause they’re are a client that I still work with, this client is relatively small compared to their competitors, their competitive set, and they wanted to win in media, but they couldn't spend as much. They didn't have the technology that their competitors have nor did they have the team size, so we needed to be efficient, smart, innovative. Going back to that whole data logic experience as a measurement, we in our first six months we spent a lot of time organizing their data correctly, made the organization very confident in the way that they collect data in the way that it's syndicated through the organization, and we worked heavily around process and governance—all that sound sounds very exciting and also boring at the same time, I know. Where it's exciting is, what we identified by analyzing the data was, we could maintain our market share in the big Metro cities by reducing specific dollars, so without kind of getting into much nerdiness but we ran what's called the double decay model—that's one of the models we ran which kind of calculated how much money we're spending against the additional share voice that we get and was it really returning the ROI and we figured out where that kind of that diminishing return was occurring, and we reallocated those that money in small markets where the competition level was small. So where you have populations of one hundred, 150,000, two hundred thousand, 50,000 and the results of that was a cost of acquisition decline of over 30% and an increase in sales in small markets by 500%. So what they did then as the next piece was, right, it's the next piece was it sounds great that the reduced cost of acquisition by 30% but typical marketers don't want to go to the boss and say I've reduced cost acquisition by 30% 'cause that means next year they’re gonna get a reduction in their budgets. So what we did is we continued to expand and again very similar to what we did at the Metro level is figuring out that double decay model and diminishing return, we started figuring out how do we maximize the outputs. So long story short, we were able to maintain market share versus their fierce competitors who are four to 10 times the size while also dominating in the small market without having to spend another dollar, without having to hire another person, all we did was make sure that our data quality was correct and everyone had confidence in it. We all united on the ideas that we had and we figured which ones had the best opportunity, and we failed forward. And I think that's the big piece too is we made sure we got to market at a very specific timeframe, tested, constantly tested, observed, tested until we got the results that I was that I was sharing.
Reva: Cool, I like it. Awesome. Well, that was it. Thank you guys for joining.
AJ: No problem.
Daniel: Thank you.
Reva: Thanks for tuning into Next in Consumer Products. Be sure to subscribe so you don’t miss a beat on the future of digital in the CP industry.
