Every Wednesday, I answer your data questions — live

Data does not have a good middle ground. Either your data is small and simple, or it is not. Small is about volume — how many customers, how many rows. Simple is about distance — how many steps stand between a stranger and the outcome you want.

This depends on the complexity of your system — not the size of your company, or the stage of your funding.

The basic metrics are total sales and total visitors. Here is what matters most within those very basic metrics — and the information you need to make the best decisions.

• First session to first sale. How long does it take a stranger to become a customer?
• The two conversion rates in your cart. Add-to-cart (adding to checkout) and cart-to-purchase (checkout to actually buying).
• First-to-second order retention. How likely is someone to come back after their first purchase?
• Which product maximizes first-to-second order retention. We've driven enormous revenue improvement for clients by giving coupons on the single product that converts first-to-second order the best. Find it, give the coupon, optimize for repeat purchases.
• How many orders until someone becomes a lifelong customer. As people buy more, their likelihood of buying again increases. You want to know which products get them there.

Give email coupons, discounts, and reminders while a customer is in the early state. Then, after the third order, you can assume they'll keep buying without the discount. Stop paying for the behavior once the habit exists — and maximize your profit.

Goal: create purchasing habits.

You can usually get Google Analytics 4 (GA4) for free. The decision that matters most is the transformation layer — it determines whether your numbers are trustworthy, so it's the one to think about.

I clearly recommend Polyform here. It's a world-class tool with a team of experts acting as an external data team for you (see the George Ortiz customer story). I recommend it without hesitation — it's a tool I, and every founder I help, come to use of our own volition. Polyform manages this whole process end-to-end and gives you a BI layer and a transformation layer, built out by our team to your needs. It's also the best way to use AI, because it's one system with context across your entire data stack.

BI tools change as your data maturity changes. Keep this in mind and choose accordingly.

The pattern I saw constantly on data teams was someone saying: “They'll see this, then they'll wonder about that, so they'll click here, then notice this — and that's why I gave them fourteen charts.” Don't do that. Do the thinking yourself and give them the answer. Do the work for the user.

I hear this constantly. Looker is wrong. Tableau is wrong.

You have your data, and you have business questions you want answered. Your goal is to optimize how you slice and dice that data. That's the prework — the data preparation. There are three levels to work through. Identify and clarify each.

Think about GLAM as an org chart for your metrics.

To test a lever, watch it over time and check for statistical significance. Every month, compare people who had three meetings against people who didn't. If the outcome is consistently better with three meetings, you know it's a real lever. As you monitor, you want the rate to hold, the two lines to keep separating, and the share of people in the positive category to increase.

Each has a use case. Here's the framework for choosing the right automation.

The best way to visualize your data depends on your stage.