## 1. Executive Summary

The core claim is that a normal ecommerce operation is structurally more complex than traditional direct marketing the classic VSL funnel world. Direct marketing optimizes one offer, one funnel and one traffic stream, read at a daily or weekly grain. Ecommerce forces you to optimize a catalogue of thousands of products at once, against hourly demand swings, a seasonal curve that inverts twice a year, and a marketplace negotiation layer.

The illustration is Black Friday at Zapos. The whole marketing team arrives at the office at 4am. Certain hours of the day generate 1 million in sales, and some hours have climbed to 2 million. At that scale a day is far too coarse a unit, so the Google bid is optimized per hour: when the hour closes, you download the data, put it into a spreadsheet (the process was manual), check the approximate ROAS, and adjust the bids then wait for the next hour and do it again.

The intensity is not a one-off stunt. On that same Black Friday, Adidas was selling while Nike was booming, which means raising one bid and lowering another in the same breath. Thousands of products each need their own judgment. Seasonality inverts the curve: sneakers and sandals in summer, then boots climb and sandals fall once winter arrives. And companies negotiate to get into ZOS because it is a large marketplace essentially an Amazon of shoes. Huge product volume, giant complexity.

## 2. Alpha / Sauce - what actually moves the needle, and exactly how to run it

- **Optimize on the hour, not on the day.** When single hours carry 1M 2M in sales, a daily close throws the money away. Close the hour, act, move to the next.
- **Download the data the moment the hour ends.** The edge came from pulling the numbers immediately into a spreadsheet not from a dashboard.
- **Judge each hour by approximate ROAS.** Read roughly where the return is sitting; speed of reaction beats decimal precision.
- **Move bids per product, every hour.** The rule is mechanical: a product is not hitting ROAS lower its bid; a product is booming raise it.
- **Let brand-level demand tell you which way to push.** Adidas selling while Nike booms forces opposite bid moves in the same hour.
- **Accept that the loop is manual at this scale.** Thousands of products plus hourly granularity equals a hands-on job the intensity is the point, not a defect.
- **Re-baseline when the season flips.** Summer is sneakers and sandals; as winter arrives, boots rise and sandals fall an inverse curve demanding fresh bids.
- **Treat the marketplace as a growth surface, not just a channel.** Companies negotiate to enter ZOS because it is a big marketplace; an Amazon of shoes means volume you can win or lose.

## 3. Methods & Setup - Step-by-Step

1. Staff the peak window in person, with the entire marketing team on site; setup: a 4am office start for Black Friday.
2. Identify which hours actually carry the load, since some hours generate 1M in sales and some have reached 2M; setup: an hourly sales target per daypart.
3. Export the hour's data as soon as it closes and load it somewhere you can read it; setup: a spreadsheet tab dedicated to the closed hour.
4. Compute the approximate ROAS for that closed hour, reading the trend rather than waiting for perfect numbers; setup: a single "approximate ROAS" column.
5. Adjust Google bids product by product raise what is booming, cut what is not returning; setup: a bid-change note logged against each product.
6. Wait for the next hour to close, then repeat the whole read-adjust cycle; setup: an hourly alarm pinned to the hour boundary.
7. Re-tune for seasonality as the curve inverts, shifting budget from sandals toward boots when winter starts; setup: a seasonal curve per category.
8. Run marketplace entry negotiations in parallel with the bid loop; setup: a negotiation track alongside hourly optimization.

## 4. Tools & Stack - what each one is for here

| Tool | Role | In this episode |
| --- | --- | --- |
| Google Ads bids | Primary lever | Optimized hourly; raised or lowered per product on Black Friday |
| Spreadsheet | Hourly analysis surface | Data downloaded each hour to read the approximate ROAS |
| Zapos office and team | Execution unit | Marketing team on site from 4am on Black Friday |
| ZOS marketplace | Sales channel and negotiation layer | A large marketplace an Amazon of shoes that brands negotiate to enter |
| Product catalogue | Complexity driver | Thousands of products, each needing its own bid |
| Seasonal demand pattern | Planning input | Inverse curve: sandals fall, boots climb |

## 5. Actionable Checklist

- [ ] Schedule on-site coverage for your highest-revenue day, starting before dawn (4am).
- [ ] Identify which hours of that day can carry 1M 2M in sales.
- [ ] Set up an hourly data download into a spreadsheet manual is acceptable.
- [ ] Define the ROAS threshold that decides raise vs. lower for each product.
- [ ] Adjust Google bids individually, per product, every single hour.
- [ ] Log every bid change against the hour it belongs to.
- [ ] Map your catalogue by season and pre-build the inverse curve (sandals vs. boots).
- [ ] Spot which brands or products move independently (Adidas versus Nike) and split their bids.
- [ ] Count your product volume and staff the peak window expecting manual work.
- [ ] Open a negotiation track for marketplace entry requests if you run a large marketplace.
- [ ] After the peak day, document the full hourly loop for the next peak.