Mastering the B2B Analytics Stack from Founding to Growth stage: A Comprehensive Guide
The evolution of the B2B analytics stack in relation to product growth.
Suppose you're working on a B2B SaaS product and looking to understand your customers and their needs better. In that case, various online resources can provide a good overview of the necessary instruments. However, many of these articles lack detailed mapping for each stage of the product development process. It's essential to consider the timing when using specific tools – ensuring you have the proper infrastructure and resources before implementing new tools is crucial.
Inspired by the six-year-old post (and still relevant) by DBT Labs' Founder Tristan Handy’s, “The Startup Founder’s Guide to Analytics”, I outlined my personal experience of the evolution of the data practice in the B2B SaaS organization that grew from 0 to a million ARR.
The structure of the guide is as follows: I've divided the narrative into three stages. These are the founding stage, the early stage, and the growth phase. In my definition, the founding stage is focused on finding the product-market fit, the early stage involves validating and proving the core business and meeting the GTM strategy, and the growth stage is a phase where the company tends to boost its momentum by acquiring additional resources and optimizing processes. Each stage has unique requirements in terms of product and operational stack. To elaborate on these stages, I'll be using the lens of tools and practices while highlighting the unique requirements in terms of product and operational stack of each stage.
Whether you're a product manager, data analyst, or business owner, this guide aims to provide valuable insights into the B2B analytics stack.
Founding Stage
The cost of insights is pretty low, and at this stage, it’s important to use qualitative data and be as close to your customers as possible. Here are some of the key ways you can explore customer needs and better navigate the product development stage to suit those needs:
Talk to your customers through in-person interviews or manual onboarding.
Conduct simple retention and churn analyses.
Validate your assumptions and try to mitigate risks.
Keep an eye on word of mouth about your product.
When you have your first customers, start picking up the operating systems to manage the customer relationship. For instance, you can simply create a Google Sheet or Airtable, nothing sophisticate. To succeed in B2B business, creating a profile of your ideal customer (ICP) is essential. Develop your tools in a manner that allows you to gain a clear understanding of the persona, which will fuel the feedback loop. In this stage as well, it's important to remember that the Go-To-Market (GTM) strategy can change rapidly and hence, it’s crucial to embrace these changes and pivot accordingly. As you gain more insight into your ideal customer profile and gather feedback from early customers, your understanding of your product and market will likely evolve.
An illustrative instance is the SpatialChat scenario. In this case, we initially utilized Close CRM and Zendesk as our CRM and help desk platforms, respectively. Through these platforms, our aim was to establish close connections with customers, primarily relying on a founder-led sales and support approach, i.e, the founders had to reply to every ticket in Zendesk.
Regarding data stack, it's not necessary to delve into complex tooling and it is also not advisable to make significant investments in data superiority.
Product analytics
Regarding product analytics, you can consider multiple tools like Amplitude, Mixpanel, June, Posthog, etc. These tools offer affordable plans for startups, making it easy for you to launch. If your product is collaborative, It's essential to ensure that the system you choose allows for group-based filtering so that you can perform analytics beyond individual user analysis.
There are a few templates on the web worth starting with. Don't follow the "log it all and figure it out later in the warehouse" methodology of logging telemetry. You're likely going to need to have more clever algorithms to isolate interesting signals from noise.
Regarding event tracking, it's crucial to prioritize tracking events from the backend rather than solely relying on client-side tracking. While client-side tracking can help capture user behavior and interactions, it can also be unreliable due to ad blockers, privacy settings, or other factors that may prevent data from being collected. In contrast, server-side tracking allows you to capture all relevant events regardless of whether they are blocked. You can also ensure that your analytics data is comprehensive and accurate, critical for making informed decisions about your product and business strategy.
Assuming you are not currently using paid marketing tools and managing your traffic distribution, Google Analytics may not be your most effective option. Instead, Fathom or Plausible can provide you with the same insight in a more secure and privacy-conscious manner.
Operational analytics
At this stage, setting up a data warehouse may not be worthwhile. If you're currently using MongoDB, MySQL, or Postgres as your production database, creating a replica of the database would be simpler. You can also query the data using self-hosted tools like Redash or Metabase.
For example, you can choose a few key value metrics representing the value the product creates for users and a few product metrics to see how users interact with the product. Please keep in mind that your metrics reflect your early strategy. They help answer questions like, "Is the strategy working?" There's no need to pick up the metric if you don't know how it's embedded in the strategy. I agree with Tristan Handy here – you can hypothetically measure a dozen things since you know the details of your business and this information can help you make pretty good decisions relying on non-quantitative data.
Early stage
As you understand the market and your ICP, and the product has proved its mettle to B2B customers, you can focus on building a strong Go-To-Market (GTM) strategy to help you secure funding and grow your business. Investing in specialized operating systems such as CRM, Help Desk, Data enrichment software, Subscription management, Appointment scheduling, and Marketing Automation may make more sense as your business grows. Additionally, you may need to hire a dedicated person responsible for ensuring functional performance.
The tools will likely be more useful based on the strategy – whether you should do outbound, inbound, or partnerships and whether you will have a sales-driven or marketing-driven organization.
In a sales-oriented funnel, the B2B marketing team generates sign-ups and leads to fill the top of the funnel. Additionally, they are responsible for nurturing and qualifying leads into marketing qualified leads (MQLs).
In the product-led funnel, the marketing team still focuses on driving top-of-funnel leads. However, since potential customers can immediately access the free product, their usage becomes the primary way to educate and evaluate them and nurture leads. Instead of simply showing or telling potential users about the product, they experience it firsthand, which is a more robust and direct approach. Therefore, the critical factors for generating product-qualified leads (PQLs) are quick activation (i.e. getting to the "aha moment") and meaningful product usage.
In B2B SaaS, the correct operating system selection is nearly as crucial as setting up the data warehouse or product analytics tool.
At this point, avoid any significant analytical upgrades from the previous stage. Probably, you don't have the resources to hire a dedicated data team at the moment, so it's essential to keep the cost of maintaining the data infrastructure low. It's better to focus on the data quality and nail the decision-making based on the observable data.
Operational data
You can rely on the built-in reporting capabilities of your various SaaS products to run your business. Most CRMs have pre-built reporting tools; it could be better but can cover 90% of the cases.
If you're not satisfied with reporting in your CRM, switch to another CRM. I am against exporting the data to Excel or Google Sheets, digging into the data, and building sales reports there.
It's normal to migrate from one operating system to another during your company's growth. At SpatialChat, we switched from Close CRM to Hubspot to transfer data consistently.
As the B2B SaaS enterprise grows, so does the number of tools available to help companies manage their data. In recent years, one trend has been the rise of no-code data analytics and data transformation platforms. These platforms aim to ease the process of data discovery and integrations between various SaaS tools by eliminating the need for complex coding and technical expertise. No-code data analytics platforms offer a range of benefits for businesses at all stages. They can help early-stage startups get off the ground quickly by providing easy-to-use interfaces that allow them to analyze their data immediately. In my practice, I have experimented with tools such as Equals, Canvas, and Baremetrics and would recommend them to early-stage product managers and founders.
Regarding communication and marketing tools, focus on selecting the ones easily integrated with your existing operational stack using low-code solutions. This will help minimize the need for engineers to create workflows and triggers.
Product analytics
Early-stage startups must prioritize efficient data management to ensure long-term success. Consider switching from an SDK to a data ingestion solution at that stage, like Segment. This will give you greater flexibility when working with event data and allow you to stream it to other applications easily. Another benefit of using Segment is its flexibility. As startups grow and evolve, they may need to add or change new event-logging tools. With Segment, this process is simple thanks to its extensive library of pre-built integrations with hundreds of popular tools. Also, I suggest considering Jitsu and Rudderstack as an alternative to Segment. They are both open-source platforms that offer more affordable pricing options.
These tools have the advantage of being suitable for both early-stage companies and growth enterprises. They offer a range of functionality that can be scaled according to your analytical needs. These tools can be transformed into a comprehensive CDP or data ingestion layer, but I will share more below.
Although I assume your team is still small and may need a dedicated data analyst, assigning someone responsible for tracking plans is essential. This person should ensure that the data collected can answer critical business questions, avoid unnecessary rework, and empower teams to use the data effectively with a shared understanding and language. This best practice can increase trust in your product data and make it more useful across the organization.
Also, at that stage, consider using collaborative event management systems, such as Avo, to reduce the overload of data quality. For data to be trustworthy, it needs to be accurate, reliable, up-to-date, consistent, continuous, complete and attributed to (and representative of) the cohort, population, or problem being studied. The collaborative schema will make a big difference in the consistency and consumption of the data, which ultimately provides greater trust.
It's difficult, costly, and often impossible to change event-based data retrospectively - and that assumes you know there's a specific issue in the first place. Planning the data you want to track well is essential to avoid duplicating dev effort. These platforms provide a centralized platform for managing your event schema, saving time, and ensuring consistency across your various analytics tools. You can define your event schema once and automatically generate code snippets to instrument your app or website with tracking events. This makes adding new events faster and less error-prone, as you don't have to manually update each analytics tool every time you add a new event. It's essential to ensure that even when the product undergoes a significant redesign, the event tracking can leverage the original event names and properties logic.
Workflows
Disregard what tools you choose, automation is your friend. Well-defined workflow automation improves communication among internal teams.
Maintaining good business operability and proper communication among the internal teams is vital. We temporarily solved the connection of all our back-office tools with workflow automation tools like Zapier, Integromat, and n8n to build the trigger-based data flows.
For reference sake, this is what our automation "create a new lead in close" looked like during this stage
I would not suggest staying long with the no-code workflow tools. With the increasing complexity and lack of engineering practices, you risk sticking with a spaghetti code that makes it hard to maintain. 🍝
Frame your metrics wisely
At this stage, all our metrics were split into three portions:
Product metrics answer questions about the product itself. They are used to understand how the changes made have affected user behavior. I recommend investing time in researching your product’s activation metric. Make sure you ask customer-facing employees what they think activation should be. Support reps, sales reps, and others spend all day talking to your users and trying to get them to find some value, so their insights are helpful for this early list.
Growth metrics answer questions about the business built around the product. To succeed in the competitive market of SaaS products, the key is to master the art of monetization. The ability to effectively generate revenue from your product sets apart the winners. It's not just about having a great product but also about finding the right pricing strategy that aligns with your target audience and their willingness to pay. Therefore, it's crucial to understand how the product acquires, activates, monetizes, and retains the account. The one who nails the monetization in SaaS products wins the market.
Functional metrics help your teammates understand how well their team performs. Those metrics answer the question about sales pipeline coverage, the number of qualified leads, or the NPS from the customer care team.
Note: Usage creates revenue, but revenue does not create usage. As a result, the most critical metrics in creating growth are not your revenue metrics, so make sure all KPIs and everything needed structurally are covered by tracking. In this, I’d be a bit more holistic and proactive. Even if there are no people to look at it, at least cover the primary customer journey, funnels, etc.
At this stage also, I would advise against investing in a data warehouse. Doing so would divert attention away from the product and require additional capacity for maintenance. Hiring a good analyst for an early-stage product is notoriously difficult unless you build an analytical tool.
Growth stage
As your customer base grows, so will the amount of data generated. Expanding your data stack is crucial to effectively managing this influx of information. This can be achieved by investing in more sophisticated analytics tools such as data warehouses, ETL pipelines, and experimentation platforms. Utilizing these tools allows you to collect and analyze vast amounts of data from various sources, such as user events, CRM systems, and marketing channels.
While this article will primarily focus on the PLG stack, it's worth noting that the B2B SaaS space is currently converging towards a hybrid model. But without data, you are flying blind.
Data warehouse
Upon further analysis, the data inputs may be vast and inconsistently collected, leading to a fragmented data landscape. To address this issue, recommend implementing a data warehouse tool. Some popular options are Google BigQuery, Amazon Redshift, Snowflake, and Clickhouse.
At the growth stage, it's crucial to recognize that the data warehouse becomes the main operation engine powering all other operational systems. A data warehouse provides a single source of truth for your business by consolidating data from various sources, including product usage, customer interactions, and marketing campaigns.
At this stage, it's important to remember that all operational systems rely on the data stored in the data warehouse. Your goal should be to implement a consistent read-write-read loop across all applications in the organization. This will ensure that all teams can access accurate and up-to-date information when making critical business decisions.
You may need to invest in an ETL (Extract, Transform, Load) pipeline to automate data movement between your various systems and the data warehouse. With an ETL pipeline, you can ensure your data is clean, consistent, and accurate across all applications. Additionally, you can use reverse-ETL tools like Hightouch or Census to push data from your warehouse back into your operational systems. I'll talk about that later in the issue.
I would resist using the term CDP (Customer data platform) here as it has been used and misused by a variety of software vendors in different contexts. Instead, I will explain how to establish a composable structure that strengthens the process of collecting, analyzing, and acting on user data. This structure will enable businesses to manage customer data and derive valuable insights from it effectively. By implementing this approach, businesses can enhance their understanding of their customers and improve their overall customer experience. Data modeling is crucial for creating a conceptual representation of your business's data, ensuring all team members understand how the data is structured. The ultimate objective is to better understand the customer by implementing a process called Customer Data Automation, which entails designing a comprehensive data model that captures changes in customer data over time. This ensures that all team members understand how the data is structured. While data modeling techniques were initially created to solve different problems, they can still help analyze data structure, quality, and dependencies. At this stage, I would recommend dividing data into sales, operation, revenue, marketing, and finance marts. This distinction naturally reflects the number of stakeholders and their demand for better data visibility.
It's easy to get data messed up at this point, so invest in data quality. Defining metrics is an exact endeavor, and semantic layers should allow us to focus on business complexity. This distinction naturally reflected the number of stakeholders and their demand for better data visibility.
Product analytics
If you have followed the advice provided in the previous stage, you should now have product analytics tools that are ready to be scaled up.
Data warehouses can be a valuable tool for enhancing event data. Instead of sending events directly from a product to analytical systems, it is recommended to gradually transition to routing them through a data warehouse. This approach allows for greater flexibility in enriching events with additional properties associated with entities from other operational systems like the help desk or CRM. Additionally, data warehouses offer the advantage of synthetic events, which can improve the accuracy and reliability of event data. Synthetic events are model data points that look like the events in your warehouse but are sourced from your systems of record, like CRM or other tools.
In today's data-driven world, relying on a single analytics tool for all your marketing data needs can be limiting. While Google Analytics is a powerful tool, it may not be the best fit for B2B SaaS companies that require deeper integrations and a lower scale of users. Moreover, Google's incubated strategy and looming vendor lock can cause concern. With Google forcing all users onto GA4 and leveraging their market share to drive broad monetization of their cloud infrastructure and advanced analytics products, it's essential to consider alternative solutions.
One such solution is event-streaming platforms like Segment, RudderStack or Jitsu, which are especially beneficial for companies with marketing activities. By consolidating raw traffic and behavioral data and all other customer data into a single place, you gain complete ownership of your data and can perform much deeper analyses. This also means you can quickly move between and onboard different analytics tools for different teams, as each tool has access to all your historical data.
In contrast, Google Analytics is a notorious silo for behavioral data. Even though you can technically export the data through ETL jobs or, with GA4, to Google products, it still requires a ton of modeling to be usable. And that's before you even try to figure out how to join user identities with other kinds of customer data from other sources.
Constant experimentation is crucial for understanding the statistical performance of a feature and answering why it's performing as such. However, product and engineering teams typically build feature experiments in the backend, making outsourcing difficult to a third party like a CRO agency. As a result, we don't hear about feature experimentation and server-side testing as often as we do for web experimentation. I suggest exploring Eppo, Kameleoon, and other server-side experimentations platforms,
Business Intelligence
At this point, I don't believe using enterprise-level business intelligence tools such as Tableau or Power BI is necessary. Instead, there are several open-source solutions available, including the tool Metabase. But before picking up the BI tool, don't confuse operational dashboards for business management with analytical tools for searching for new knowledge and forming and testing hypotheses. Consider the classic analogy. The dashboard of a car tells you "what's going on now," but it doesn't tell you where to go. You can build 20 speedometers into the car, but they must still tell you the direction. So two tasks are mixed - operational analytics (I want to know how fast I'm going) and search analytics insights (where and how to go). That is, expectations from the dashboards are exaggerated - the speedometer allows you not to grab a penalty or not to get out of turn, but it will not show you the way.
Both vendors and BI managers, consciously or not, are playing "how much money will the company make/save when employees see new ideas in the dashboard. But in reality, the money is saved on automating report creation, reducing response time to problems, and less often on insights and new ideas. That's why I think there are often inflated expectations from BI systems and, ultimately, disappointment when they don't match reality. Many teams view the goal of data initiatives as tracking metrics though the real goal is to analyze those metrics; those two things are very different. The latter is how we make information actionable.
Jupyter Notebooks are often the tools of choice for exploration, but they suffer from notable shortcomings. Sharing is cumbersome, reproducing or rerunning analyses is brittle, and the technical barriers to adoption are needlessly high. You may try using Hex to combat such disadvantages.
Data Activation and Delivery
To elaborate on my previous point, the core goal of analytics is to balance the democratization of access to information with driving business impact for the organization. The self-service culture has led to an increasing number of tools meant to make things easier using AI, NLP, and other technologies. However, the problem is not the tool itself but rather the management of the generated content and user decisions based on that data. Data-driven approaches should not be applied at all company levels but only in the right places - at the proper management level and where the data may contain insights and immediate benefits, at least in theory.
Contextualizing the data to build productized experiences for data consumers
The rapid adoption of cloud data warehouses has given rise to Reverse ETL—a new paradigm in data integration that enables you to activate data already in the data warehouse by syncing it to third-party tools. I want to elaborate on the point from the previous section, where I advised to replace in-build reporting in operational systems. Dashboards are not enough to make data actionable by business users. Every operation system can be transformed to the micro-BI using reverse ETL.
Time and resources in a world of quarterly OKRs and fast-moving competitive markets will always constrain teams. Don't force people to look into the dashboards and write SQL queries; instead, contextualize their experience with data.
From dashboards to activation PLG companies can leverage their bottoms-up tailwinds to drive a top-down sales pipeline. We pushed the current data to Hubspot to enrich the work process for the sales team. Why? Having data in Hubspot gives sales reps a better understanding of the conversations they need to have with prospective customers and which accounts are most ready for an enterprise sale. For the operational folks, it helps to spend less time on integration work. Increased capabilities of the data function unlock deeper accountability as specific teams can now be responsible for input metrics that lead to better scoring instead of a generalized, shared responsibility of output metrics. But when you pipe data into downstream applications, remember that it's easy to overwhelm your team. A better approach is to think about the end goal.
Along with Hubspot, you can also rely on A PLG CRMm, Product-Led Growth Customer Relationship Management like Correlated, Endgame, Pocus, Calixa, Breadcrumbs, HeadsUp, or Variance, Breyta. This type of CRM focuses on the customer's product journey rather than the sales journey. One of the main advantages of using a PLG CRM is that it can help identify more qualified leads, such as Product Qualified Leads (PQLs), who have interacted with your product but have not yet made a purchase or had any contact with your sales team.
SDRs can use product data to personalize outreach and improve their conversion rate
CS teams can use product data to proactively reach out to customers who might be ready to expand and convert to a higher contract value.
AM can have a better picture of the client usage and thus craft and personalize their outreach.
Data as a productivity step
Notifications enable a healthy combination of tech-touch and high-touch communications. When it comes to B2B business, there's always a conflict between the means of communication. In the end, you need to help your teammates reach their goals, so we developed a convenient notification engine in Slack. In that case, the account manager can decide whether they need to reach out to the customer, add the task for themselves, or ask them to review the product on G2.
If you're looking for a reliable tool to monitor business events and respond to incidents, I highly recommend checking out Avenue. This platform offers a comprehensive solution for alerting flow, allowing you to stay on top of critical issues and respond quickly and effectively. With Avenue, you can set up custom alerts based on specific criteria, such as error rates, latency, or user behavior. You can also automate incident response workflows, so your team can quickly investigate and resolve issues as they arise.
Hiring
Be mindful when growing your analytical force.
Honestly, I believe there is no option other than initially hiring a full-stack data person. Small data teams must be creative, take on multiple roles, and focus on quick results.
I'd suggest investing more in sharing and transferring knowledge across the company at the beginning rather than hiring aggressively. Save some time to clear the technical debt at least once every two cycles.
If you take anything from all of this, remember:
Data needs to be seen as a strategic lever for GTM strategy.
Value process over tooling. For instance, PostgreSQL is as fine as DWH.
Prioritize the DWH-native operation systems for your data stack.
Business problems are extraordinarily complicated, and analytical recommendations are mostly educated guesses.
At every stage of the company, analysts should think about maximizing the value of observable data.
After our two-year journey, operational analytics should be the backbone of your GTM strategy, disregarding the stage. After all, the more stable your foundation, the faster you can innovate.
We haven't addressed the challenges related to scaling data organization, but that's certainly a topic for future discussion.