A martech masterclass, by Francesco Pittarello, former Director of Marketing Technology at Uber

Everyone is free to use other companies if they want! Most of us have actually tried to be either a driver or to do delivery. I think it’s eye opening, to be honest. Earlier in my career, I was already in food delivery, with Foodpanda within the rocket group, and I did deliveries. It is important to know how the business operates end to end and what the experience is for a driver or a rider, so that our people in marketing grasp the reality of the way customers experience our brand. It also helps to get in tune with what our demand looks like.

There are a few things I was doing in the beginning which I think are key to success. The first one was, staying very close to details, going deep into things, understanding how they work and how to innovate. When I was at foodpanda, we started testing automated ads from Smartly. It was an early Alpha and we built a feed to serve different promotions in different cities in India. And then my colleagues started getting interested and coming around with questions. At this point I started leading even though people were not reporting to me. From there someone noticed; I got a team and started growing it.

But in a nutshell, it’s really about staying close to the details. Then obviously when you grow into an organisation, you cannot stay close to everything. But having the capacity to zoom in and out, to be able to support teams in a jam session or a brainstorm, and still keep the focus on the big picture, that has been crucial

The second piece is the soft skills. Understanding people, spending a lot of time analysing characters, inclinations, communication style. I try to customise the way I interact with others depending on their personality, the place in the career, or what they are interested in. My goal is to get the best from everyone by adapting my communication style. I started with these two things and they are still my principles today.

Lastly, I learned from the people I worked with. Interestingly though, I would say that I learned the most from people who weren’t even part of marketing, tech or performance marketing, especially in the last few years at Uber. I’ve been engaging a lot with other departments – finance, general managers, operations – and that gave me a 360 degree view on how folks in completely different spaces interpret leadership. That’s where you start seeing that the traits are always the same. It doesn’t really matter if you’re doing tech or business or HR – you find common patterns, such as deep analytical skills, the capacity to simplify problems with frameworks, empathy, and communication skills.

Well, my role was basically divided into two parts. I managed a few of our paid global channels, including social programmatic, job boards, display and new channels. Basically the non search channels. I also managed two other teams. One was called Marketing Solutions, a tech team that does mostly feed management and BI, and the other, App store optimization and Downloads.

I was using this strategy where I identified every week the area or the pain point in the domains we were managing. If there were areas that are doing well, I didn’t really interact much apart from a one on one with the manager. If there were any other problems or we did not know what to do, or one of my direct reports needed guidance, then I dedicated more time to that specific area.

iOS15, let’s say the no IDFA world, that was the new big topic. How do we adapt attribution? How do we want to adapt our targets? How do we want to think about probabilistic stitching of LTV to campaigns? The other big topic was downloads and app store optimization. We were building a team around it, which meant investing a lot of work directing the team towards the right priorities.

The first thing we look at is what we need, and if it is available on the market at a fair price. If so, the decision is rather obvious and we can buy a service. But really that’s not my mental model for Martech – I rather think of ‘first mile’ and ‘last mile’. The build or buy decision doesn’t have to be an either or choice, you can do both at the same time. Meaning you can leverage a partner’s tooling and develop a layer on top of it.

That’s why Smartly is so great. They built a set of technologies you can expand on. For instance you can build automated ads, and scale it to any type of ads, and even build a team around it which can continuously iterate with your own data and tech.

The only case we do things in house is if nothing is on the market or the solutions are not customisable enough. But with the current explosion of SaaS businesses it doesn’t make sense to think of having everything in house anymore. The other way around is true too: you always need a part of internal logic and thinking on top of the tools you use.

We need to use media mix modelling and other models given the recent changes in the industry. But I am not sure how viable it is for the industry to go full blown into that type of measurement. At some point, you still need a way to measure your daily effectiveness, which probabilistic attribution is not doing yet.

In performance marketing, you invest until it doesn’t make economical sense to you. You need to know where that tipping point is every day. We lost a lot of direct feedback and data with SKAdNetwork and delays in the postback, that’s certain. But some type of direct measure will still be there to anchor your models, there’s no way around it if you want to be efficient.

The most important piece of work is going to be connecting probabilistic attribution and daily decisions. Some of the gaming players are already far ahead in the domain and start setting conversion values on SKAD based on predicted LTV and such. In my opinion, whoever solves this problem and integrates MMMs in their daily decisions will get a large competitive edge.

Well I know them for a long time now and I brought Smartly to all the companies I have been to. They took over and became global in only a few years. It really is amazing to see. The main thing I was asking them was to add more marketing channels and they are doing it now, so overall we are happy customers.

It truly is a great company and team. That something like this would come out of Finland, of all places, means there must be something in the air in Helsinki: Supercell is also there; Unity ads… It’s a great tech hub.

So on the ratio idea – it depends on the maturity of the company. When you start a company, there is demand and you try to capture it, which is the role of performance marketing. That’s actually why you start a business – you find a latent need in the market which you solve. Then at some point in your history you are going to start slowing down, because you have saturated your high demand, high intent users. That’s the point when brand becomes important. Thirty or fourty years out, you have businesses like Nike or Coca Cola where brand is pretty much the entire business.

That’s my mental model for the role of brand marketing at different stages of growth. The measurement side is the complicated one. The two best routes you can go to is either put everything on the same plane ; or use a completely different set of metrics.

The first way you have no Chinese wall between brand and performance like in most companies. You use a media mix modelling and attribute everything together. Obviously brand takes a longer time to generate revenue than performance, so it can be that your models over index in what drives short term value.

On the other hand you can separate your performance metrics from the brand ones. Brand is measured through top of mind awareness, consideration, Nielsen surveys, and this type of KPIs. You focus on getting these metrics up. That’s a cleaner way to separate things but it doesn’t tell you how much to invest in each bucket overall. You can’t put everything back to the same metric, dollar for dollar, so the end decision of how much to invest in Brand is still gut feeling based.

This is where marketing becomes hard. I have seen a trend of trying to overmeasure things but in the end, the big CMOs go with their gut and see what feels right. Is it important to be top of mind? Can we convince finance that it makes sense to invest for a few years? If you think back about Coca Cola and Nike, they sell the same things for generations but the associations with the brand are so powerful, you forget the drink is the same, or the pair of Nikes is only a pair of shoes. These are great examples of how you can go beyond simple brand measurement.

This separation is something that I’ve been witnessing myself. These two functions tend to go in different directions. First and foremost, I think sometimes the term CRM is misused or misunderstood. In many situations it gets conflated with email marketing or push notifications. I don’t see it this way – email marketing is not very different in nature than any other marketing channel apart from the fact that it is free. You can do targeting and audience splits, and even send emails to people who are not your users by buying a database. CRM is more about really how to target your customer base, how you talk to them and at what frequency.

In an ideal state it should be done across channels, including paid, email marketing and push notifications. It’s rarely done this way because everyone starts with emails, which is blended into CRM later in the life of companies. I imagine a CRM team to be more like a lifecycle strategy team that manages all the channels and uses them to talk to the consumer and look at cohorts, messaging, incrementality, all of that consistently and across all touch points.

Lifetime Value (LTV) is the big metric that CRM should go after. The way I see it is you can manage the cost of acquisition to an extent, but the CAC is actually outside of your hands. You can optimise it, but it depends on external factors. If you are a travel company and the CPC goes up like crazy on SEA you can’t do much. I guess you can get fifty data scientists and try to set the bids in the most granular way possible at the zipcode level or whatever, but the price is there. LTV is different. In theory there is no ceiling. You can launch more products, you can then add a subscription service: you can increase it over time.

Fundamentally the challenge in the larger organisation, and to be honest, probably also in startups, is that there are so many other functions that influence that metric: product, pricing, strategy, internationalisation… It’s very difficult to say that one department owns the LTV. In short, I think CRM should be cutting across all channels and it should probably be the department most connected to product, finance, and other departments. The metric should be LTV but it shouldn’t be exclusive to CRM.

I think the first piece is not thinking about email marketing push. Don’t start from there. Start from the customer. Spend your first three to six months thinking about customer segmentation, what customers do, how certain behaviours influence LTV. If you think: what type of frequency in the early days gives us a sense of long term LTV? What type of events correlate to reactivation?

Find metrics that do not depend on sending emails out. Email incrementality should be the metric of the email team, which should be sitting on the same side with the social team and the SEM team. Find metrics that everyone can go against. So then you align everyone around you. We have an incentive towards this impact.

Understanding what the right levers are and what the right signals affect LTV is key. Because if you get to know that you can realign the organisation against, let’s say, a distillation of a few events that you proxy to. Because that’s the other problem that I noticed, it’s very easy to get lost with CRM. You go into all of these different cohorts and hundreds of different tests, but at the end of the day, you should simplify to around two or three things that really matter in the lifecycle and go after them.

Obviously this is a massive challenge, not only for Uber, but for every multimodal company. I’m sure that also in the ecommerce space and others it is the same. I haven’t seen someone really cracking this, because you don’t start multimodal. You start with a single business model and add layers onto it.

If I wanted to start from a blank slate, and control it from the beginning, the way to think about it is around what I would call a baseline LTV and then an incremental LTV. The baseline LTV is the minimum that you can assume with a certain degree of certainty, say a low mean average percentage error on a net user that comes to your product. If for instance, someone orders at Amazon, you know on average the person is going to order X, and you start based on that. In the past there was also a lot of work around being super precise with that LTV prediction. I’m not sure how much sense it makes in a world with no user level data. So you probably need to think about clusters and brackets of LTV.

And then you can go after incremental LTV. You calculate: “if I launch this other product, how does my LTV increase on the baseline versus this other product?”. This way, you can still function in terms of acquisition with your baseline LTV. And if you have a set of people working on the second product, you can realign them towards the incremental value this product generates.

If you see that a big chunk of users order from your second product after doing it first on your main service, you can start ingesting that into your baseline LTV. It is obviously not hyper scientific, but I don’t think there is a way to make this perfect. I see this more as a mental model to redirect people to the correct metrics.

This approach is great but it has a potential drawback. In case you are basing your bids on search directly on your forecasted LTV, you might be off target if the LTV was not measured accurately. If you overestimated LTV, you are overspending until you realise it. I would rather under spend until I am really clear on the value generated by a product.

I learned the hard way that it’s always more work than you think it is. You always think that you can take what you’re doing on a channel, go somewhere else and scale it, but it never works that way. Launching a new channel requires dedicated focus, patience. If you look at connected TV, LinkedIn, Podcasts it can be that you look at totally new environments. It can be TikTok, but also vertical video formats given the growth of Reels and the development in the area in the past years.

You want to ask yourself, aside from the prospective side of the channel: is the tech there? Is the reporting there? Do you find performance ? Twitter is a fantastic example of a channel that is still waiting to build a decent Direct Response (DR) product despite a good audience, time spent on the platform, and signals being available. This is an example of a channel with high promise but that in general doesn’t deliver for DR.

Then you have to see how you integrate it in your stack – do you add it to your attribution; do you do custom incrementality measurement; etc.

Obviously this is another huge challenge. It starts with your tech stack. If you build it correctly so it can spin creatives rapidly out of templates, using localisation at scale within these templates and so on, then you save a lot of time. This is how we try to do everything. The second layer is on local teams and cultural relevance. For evergreen performance marketing campaigns, my experience is that it doesn’t require going country by country and being so nuanced. You can run similar concepts.

If you are truly a global player however you have these different cultures, with Japan for instance or the Middle East, which are used to different styles, colours and references. Here you have to be more subtle and understand the culture, which will force you to have dedicated people on the topic.

The channel mix is similar but the funnel is much more complicated and diverse when you are trying to attract earners (internal name for drivers). In some markets it’s a simple B2B thing, you acquire someone looking for a side hustle. In others you acquire professional drivers, so the messaging and the funnel changes completely. It is not as scalable as with customers. Since you also get less new users, you can’t easily predict an LTV as we can for riders. The funnel is also way longer.

What is in common – they both are price sensitive. Consumers react well to discounts; earners react well to earning opportunities. The pricing component is always prevalent.

n our industry the main reason people would leave is pricing – so the way we act against this is two fold. First we try to make sure we are top of mind for customers and clients. We want to maintain high brand awareness. And then we compete on price.

In previous industries where I worked, the analysis could go deeper and more granular since you could find more reasons why people left. Maybe it was customer support or the frequency of usage of the tool and you can create campaigns around this theme.