The Startup Founder’s Guide to Analytics: You need analytics.

I’m very confident of that, because today, everyone needs analytics. Not just product, not just marketing, not just finance… sales, fulfillment, everyone at a startup needs analytics today. Analytics powers every decision, from the strategic to the tactical, from the board room to your line level employees.

This post is about how to create the analytics competency at your organization. It’s not about what metrics to track (there are plenty of good posts about that), it’s about how to actually get your business to produce them. As it turns out, the implementation question — How do I build a business that produces actionable data?—is much harder to answer.

And the answer is changing fast. The analytics ecosystem is moving very quickly, and the options you have at your disposal have changed significantly in the past 24 months. This post reflects recommendations and experience with the data technology of 2017.

FOUNDING STAGE: (0 to 10 employees)

At this stage, you have no resources and no time. There are a million things you could be measuring, but you’re so close to the details of your business that you’re actually able to make fairly good instinctual decisions. The one thing you need to make sure you are measuring is your product, because it’s your product metrics that will help you iterate quickly in this critical phase. Everything else can take a back seat.

WHAT TO DO

Install Google Analytics on your website via Google Tag Manager. The data won’t be perfect without more work but it’s not the right time to worry about that.

If you are an ecommerce business, you really need to make sure that your Google Analytics ecommerce data is good. GA can do a decent job of tracking your ecommerce business all the way from visitor to purchase, so spend the time to make sure it’s right.

If you build software of any type, you need real event tracking. I don’t care what tool you use — Mixpanel and Heap are very similar and they’re both good. At this point I wouldn’t think too hard about what you’re tracking: just use Mixpanel’s autotrack or Heap’s default installation. If you realize you need a datapoint, you’ll find it’s already there. This approach does not scale well, but for now, it’ll do.

Your financial reporting should be done in Quickbooks. Your forecasting should be done in Excel. If you’re a subscription business, use Baremetrics for your subscription metrics. If you’re an ecommerce business, use your shopping cart platform to measure GMV. Don’t get fancy.

If you’re not technical, you may need an engineer to help out with GA and event tracking. This entire exercise shouldn’t take more than an hour or two, including reading the docs. It’s worth it to take the time out of building for this.

WHAT NOT TO DO

Everything that is not one of the things above. Do not let someone sell you a data warehouse, a BI platform, a big consulting project, or…yeah, you get it. Stay focused. When you make a commitment to analytics, there is an ongoing cost. Data changes. Business logic changes. Once you start down this road, you can’t really put the project on pause. Wait to make this investment until later.

There will be many questions that you just can’t answer yet. That’s fine (for now).

VERY EARLY STAGE: (10 to 20 employees)

You’re growing your team a bit. These people need data to do their jobs. They may or may not be data experts, and you need to make sure that they’re doing the basic things right.

WHAT TO DO

You’ve probably hired a marketing person. Make sure they own GA. Hold them accountable for making sure the data is clean. They need to UTM track every damn link they create. They need to make sure your subdomains aren’t double-tracking. Your marketing person may say that they’re “not a GA person”. Don’t listen. There is enough information on the web about GA that if they’re smart and motivated they can learn it and figure it out. If they can’t figure it out, fire them and find someone else (seriously).

If you have a sales person or two and use a CRM, use the built-in reporting. Make sure that your people know how to use it. You need to be able to know basic things like rep productivity and conversion rates by stage. Salesforce can do this stuff out of the box. Don’t export data to Excel, build the reports in the (terrible) report builder. Even if it’s painful, this will save you tons of time in the coming months.

You probably have a couple of people in customer success. Most help desk systems don’t have great reporting, so choose KPIs that you can measure easily within the interface.

Make sure you track NPS. Use Wootric or Delighted.

WHAT NOT TO DO

It’s still too early for a data warehouse and for SQL-based analytics—it just takes too much time. You need to spend all of your time doing, not analyzing, and the most straightforward way to do that is to use the built-in reporting capabilities of the various SaaS products you’re using to run your business. You also shouldn’t hire a full-time analyst yet. There are more important things to spend your limited funds on at this point.

EARLY STAGE: (20 to 50 employees)

This is where things get interesting, and where the changes in the past two years really start to become apparent. Once you’ve raised your A round and have 20+ employees you start to have new options.

These options are all driven by one thing: analytics tech is getting better, fast. Previously this type of infrastructure was reserved for much larger companies. Its benefits? More reliable metrics, more flexibility, and a better platform for future growth.

This is the hardest and most critical phase: promising if you do it right, but pain-inducing if you do it wrong.

WHAT TO DO

Set up your data infrastructure. This means choosing a data warehouse, an ETL tool, and a BI tool. For data warehouses, look into Snowflake and Redshift (I prefer to work with Snowflake given the choice). For ETL tools look into Stitch and Fivetran. For BI look into Mode and Looker. There are many, many products in this space; these six are the ones that we come back to time and again with our clients.

Hire a strong analytics lead. Down the road, you’re going to need an entire team of analytics professionals: engineers, analysts, data scientists… But for now, you can only afford (at most) a single headcount. You need to find that special person who will be able to provide value on day 1, but who will also be able to hire the team around them as you grow. This person is hard to find—invest the time to find them. Often these folks have backgrounds in consulting or finance, and they frequently have MBAs. While this person should be able to roll up their sleeves and get their hands dirty, focus on hiring someone that can think about data, and about your business, strategically: they’re going to be the most important piece of your analytics puzzle for years to come.

Consider hiring a consultant. While it’s great that you’ve found your analytics lead, that person isn’t going to have the expertise required to put together all of the components of your tech stack or the experience to solve all of the different analytics problems you’ll face across your business. Mistakes made at this critical stage have serious costs in both time and money as you grow, so it’s important to lay a solid foundation. To do this, more startups today are choosing to work with consultants to help them get set up, and then building a team around that infrastructure.

WHAT NOT TO DO

Unless machine learning is a core part of your product, don’t hire a data scientist yet. You need a generalist, not a specialist, to build your analytics team.

For the love of all that is holy, do not build your own ETL pipelines. This will waste so many hours of engineering time. Buy off-the-shelf from Stitch or Fivetran.

Don’t use any other BI tool than the two I mentioned above. You will pay for this down the road, hard.

Don’t try to “get away with” using a more traditional database like Postgres as your data warehouse. It’s not that much cheaper and it’ll be a real time suck to switch later when you max it out. Postgres does not scale well as a data warehouse.

MID-STAGE: (50 to 150 employees)

This stage is potentially the most challenging. You still have a relatively small team and few resources, but you’re being asked to deliver increasingly sophisticated and diverse analytics to the business, and your work can directly impact the success or failure of the company as a whole. No pressure.

It’s important to make forward progress here while making sure that you continue to lay the groundwork for future phases of your growth. The decisions you make in this phase can cause you to charge straight into a brick wall if you don’t think hard about the future.

WHAT TO DO