Designing a marketing stack for your company is both parts exciting and overwhelming. There are hundreds of options and companies don’t have time to explore them all. You’re focused on growth and not becoming an expert in all the available vendors in your space.
In this short guide (emphasis on short), I will give you a playbook for how to build modular marketing stacks where you could easily swap out any given module e.g. email marketing.
I think flexibility is a fantastic goal because you will be forced to make changes to your stack at some point. As your company (and product) grows, you’ll have to swap out your existing tools. This swapping out process could be easy or really painful.
If you’re able to easily swap out tools then the process of choosing them also becomes less time-intensive. You can focus on choosing tools that are “good enough” for now without worrying about whether you will be able to scale with this specific vendor.
Looking at tools from the lens of “modularity” is different than overall functionality. I’ll show you how to think about data flows and how everything can fit together like a puzzle.
I used to focus on finding the perfect tools for companies. This meant lots of research and thinking through all the different scenarios that the company might go through. Despite all this, I was still seeing companies wanting to switch tools after 6 months or because they wanted to do new things.
I realized that it was impossible to predict everything a company will need or how it will change. There are so many variables that trying to manage them is a losing proposition. Instead, you should focus on making your marketing stack as modular as possible. This simply means that you can easily swap out elements of your stack as needed.
This trend of modularity is something that will become more important especially as companies move towards having a single source of truth.
Vendors would prefer for you to use their tools for a long time. That is normal and so is seeking vendor-lock in. Historically, vendors made it hard to move away from their tools.
This meant that it was difficult to export your data, transfer your settings or simply move over any work created within the tool itself (campaigns, reports, workflows).
Focusing on modularity aims to solve some of these problems. It won’t be perfect but thinking about this from day 1 will make changes easier in the long term.
You should be focusing on owning your data from day 1. This might not be useful right away but it will be necessary when you want to use machine learning to build your own models or when you want to analyze the entire customer journey from start to finish.
Data privacy is also putting pressure on companies to know exactly what data is being collected and where is it stored. Instead of living in a messy house where you can’t find anything, you can organize everything from day 1 and build upon it.
The solution to modularity comes from thinking in data flows. This means being clear as to how data flows between tools and what would happen if you want to swap any given tool in your stack. Does your data flow break? Are you stuck with a specific tool?
We start by capturing all the sources of data including websites, apps, other APIs, etc. All of this goes through some kind of central data abstraction & user model building and the flows into the 8 main categories of any stack. The user model portion is quite important and can make your life so much easier.
We have 3 major elements to play with so let’s dive into designing a stack, common tools and how to stress test your stack.
Designing your own stack isn’t rocket science. You’re working on mapping your business into a series of steps and then seeing how a given tool will support you along the way.
Based on this, here are the steps for starting to build our own stack. Remember that we want to have a visual representation of this stack similar to the image in the previous section and the images you will see in the next section.
1. Determine which data sources are important to your business right now. What are the websites, app or backend systems that you would like to connect?
2. How will you capture data in some kind of central way? This could involve using tools like Segment.com and mParticle (which we will cover below) or by building your own central data capture.
3. Which of the 8 categories are relevant to your business right now? You don’t have to use all 8 and you can always add other options later on.
Let’s look at popular tool options and how this list will seed your research efforts.
Don’t break your head when choosing tools. Every category has 4-5 major options that most companies use and that are production-ready. Here are some of the most popular options for your research process. Remember that we want to see how modular each tool is and how well it can play with our other choices.
This category is all about having a central place for capturing data and then sending it to other tools. This is an important step as we want data to flow through a central pipe to keep track of it.
This category is about knowing what marketing campaigns are converting and driving traffic.
This category is about knowing how your users are engaging with your products. The more complex your product e.g. web app, the more you will need these tools. Less complex products like ecommerce sites could get by with less. Also a great way to understand correlation vs causation.
This category is about being able to communicate with your users through different mediums including email, SMS, push, in-app and even physical mail.
This category is about capturing qualitative data about your users such as surveys, heatmaps and session recordings.
This category is about tracking the technical performance of your product such as page loading time and errors.
This category is about tracking sales data and deals.
This category is about running A/B tests on your website, product, and even backend systems.
Think about these options as your starting points in your research. You may discover other options that are better suited for your industry. You will also realize that some tools can tackle multiple areas which would simplify your stack.
You’ll soon start to make choices across the categories and start building your stack. You want to make sure that you can “stress test” each element of your stack for modularity. Answer the following questions:
The last question is important because, in theory, you could connect anything. However, the effort to do this might not be worth it. You can also explore options like Zapier and Piesync as a way to connect elements in your stack.
Let’s now look at how stacks would look like for different industries. I also want to focus on the modularity piece and why certain tools tend to be a better overall fit.
In this stack, we are collecting data from a backend product database, the marketing website, a web app and pulling campaign data from Facebook & Google.
We’ll use Segment.com to centralize the data, normalize it and then send it downstream to out other tools.
In the categories, we’ll use Google Analytics for marketing attribution, Hubspot for Sales CRM and user communication, Mixpanel for user behavior, Hotjar for qualitative data, and Google Optimize for A/B testing. We’ll then store all the raw data within Amazon Redshift as our data warehouse and we can use Looker to visualize this data.
Also, note that we are sending the A/B testing data back into Segment so it can flow back into our other tools and we can keep track of which users saw which variations.
In our stress test, we could swap out Mixpanel for Amplitude with minimal disruption. We could also replace Hubspot with the Salesforce world and still continue on fine. You’ll notice that the more popular a tool is, the more integrations it has. Google Analytics integrates well with almost anything making it quite versatile.
It also pays to commit to a “world”. If you use Google products (Analytics, Optimize, Tag Manager, etc), you’ll know that they will integrate well with each other. The same goes for Salesforce products.
In this stack, we are collecting data from our marketing and ecommerce website, and pulling campaign data from Facebook and Google.
We’ll use Segment.com to centralize the data, normalize it and then send it downstream to out other tools.
In the categories, we are using Google Analytics for marketing attribution and user behavior, Klaviyo for user communication, Hotjar for qualitative data, Google Optimize for A/B testing and Amazon redshift with SQL for any complex analysis.
Ecommerce sites tend to be more simple and we can get away with less. It’s important to understand that we may have issues with volume, especially with Segment.com.
We are collecting data from a marketing website, mobile apps and Facebook & Google for campaign data.
We are also using Segment.com to centralize everything and build our user model.
On the tools themselves, we are using Google Analytics for web attribution, Appsflyer for mobile attribution, Iterable for user communication, Mixpanel for user behavior, UXcam for qualitative data and Amazon Redshift for storing the raw data.
We are also sending the attribution data from Appsflyer into Segment so it can flow back into other tools like Iterable and Mixpanel.
Mobile apps are tricky because they can be quite tedious to debug issues, especially around mobile attribution. Like in ecommerce, the volume will be a concern for Segment and something to keep in mind.
While most companies are fine collecting data and sending it to multiple tools, some industries don’t have that luxury. Products like banking and cryptocurrency have strict requirements on where data could be stored. This means that your marketing stack will need to be adapted to a new dimension: data privacy.
Besides modularity, you will need to consider data privacy. Having a central collection of data will be even more important for your company since that will make it easier to keep track of any data that is captured.
You will also need to confirm where data is being hosted. You may be ok with vendors hosting the data within specific geographical limitations e.g. European Union or you may need to actually own the database that owns the data.
These companies have to typically rely much more on open source solutions since that gives them that flexibility needed. You’ll find a few options in the next section but your options will be smaller.
The tools in this category will depend on how sensitive your company is. I focus on providing full open source solutions that are within your control but there are caveats since data will have to live in a cloud somewhere. You will need to trust Amazon or Google to some extent.
You will also notice that there are fewer options here. If you need to go open source, you will need to rely upon your engineers to build a lot of this functionality.
You will continue to optimize your stack even after your initial design. Your optimization efforts should revolve around the following 3 major areas.
Look for things that you can’t do. You may not be able to send SMS messages or you’re unable to properly attribute mobile app installs. Depending on how critical these gaps are, you can decide what needs to be added to your stack. If you’re running into errors, consider revisiting your data tracking plan for anything that was missed.
You should also look at how much time your team is spending on common tasks. I have seen teams spend a large amount of time building reports that could be automated with the right tools.
Finally, look at what is on your wishlist. This is where going to a conference can come in handy. You’ll see all kinds of things you could be doing such as using data enrichment to personalize communication or using machine learning to predict conversions. These are all potential additions to your stack.
Having the right marketing stack can make growth easier to achieve. Remember to focus on modularity since tools will change regardless of how much you research them.