Polar Analytics

We looked at Black Friday retention across 33 top brands. Almost $5B in GMV. 30 of 33 saw Black Friday customers come back less. Retention was 29% lower on average. And almost nobody adjusts their CAC target for this. A Black Friday customer carries two discounts: - the offer you run - the lower retention of the customers you acquire Some brands saw little decline. Some did better. So use your own numbers: 1. Pull last year's BFCM first-time buyers (Nov 28 - Dec 1) and their repeat rate at 90 and 180 days 2. Compare them to customers acquired the rest of the year 3. Set a separate BFCM CAC ceiling.

Was on a call with the CEO of a ~$100M brand running in 6 countries. His Head of Growth asked for 2 more data analysts. CEO: "No. Let's speak to Polar first." What was actually going on: - daily reporting ate the equivalent of 3 FTEs - simple questions took half a day, sometimes a whole weekend - teams waited 2-4 weeks for answers - weekly stock planning = 10 people in a room. 50% data, 50% gut The playbook we're exploring with them: 1. one warehouse for all company data 2. a governed front door so everyone pulls numbers they trust 3. connect AI to that data, train agents on the company's rules 4. start with the analyses eating the most hours 5. teams ask in plain English instead of waiting for a report 6. humans still own the final call Next gen of companies won't scale by hiring more people to answer questions. They'll let the people closest to the business answer their own. Smaller teams. Bigger revenue. More products.

Custom attribution models in Polar, built by prompting Claude. Most tools leave you with two bad options: → rigid templates (first/last/linear/U-shaped) you understand but can't bend → black boxes that "adapt" and you can't explain to finance Budget still depends on a model that doesn't fit your journey or that you can't defend. Workflow: 1. Connect Claude to Polar via MCP 2. See Polar Pixel journeys: channels, touches, time to purchase 3. Describe how credit should work. Prompt the model. Valabasas built custom models with us. After using them with Ask Polar, Head of Ecommerce David Lawson told me: "I've gotten insights that every other model and view and platform missed." Built one live in under 5 minutes below.

@polar_analytics just got 10x more valuable for consumer brands on @OpenAI. Anyone who worked in data knows that making decisions comes from handling the details: - calculating acquisition costs net of $0 orders and gift to influencers - looking at conversion rate net of bot traffics - reviewing COGS by elegantly jumping from Shopify to NetSuite Connectors are great but all of this logic belong in a data warehouse. This is exactly where Polar fits in: making it easy to ingest, model and store all of brands data inside Snowflake, then safely exposing it to AI via a Semantic Layer MCP. Now using the Polar MCP, this new @ChatGPT data agent can create accurate visualizations, help answer questions, and overall democratize data for everyone. Very excited to see this vision of giving any business operator their own team of data scientists continuing to progress.

Stress-test your Black Friday plan in 48 hours with Claude. Every brand I talk to is already deep in offers, budgets, inventory, and calendars. There's still time to fix the assumptions. Claude on your live data compresses 2 months of analysis into one focused session. I pulled the 10 BFCM questions that matter into one guide, including: → What can we realistically expect to sell? → Which products and offers to run → The full holiday calendar, not just the weekend → Where the marginal dollar goes (incrementality, not platform ROAS) → What breaks if demand beats the plan → How BFCM buyers become repeat Each question: a ready prompt + an example output. Plus a calendar from now through Cyber Monday and after. Brands that win Q4 lock it in September and spend October stress-testing it. Reply "BFCM PLANNING" and I'll send the guide.

Everyone I speak to thinks they're behind on AI. Nobody realizes how early we are. This week: a founder at $1M/year, the CIO of a $1B brand, and the director of a Shopify platinum agency. Same question every time: "How is everyone building all these agents?" I call it AI imposter syndrome. A lot of what you see online is not running in production. That "agentic OS with a built-in feedback loop"? Often a Claude chat with a good prompt. Meanwhile smart operators won't start because they think they're already 12 months behind. You're not. So I'm doing something I've never done: Live AI 101 for ecommerce operators + agencies. Next Tuesday. 1 hour. The 20% that gets you 80%: - a virtual assistant for day-to-day - one better decision per week - one recurring task off your plate Model-agnostic. Claude, Codex, Grok-doesn't matter. Live only. No recording. Max 15. I answer every question myself. Reply if you want in.