🚨Sirens blared the second I walked into Facebook HQ, got banned on the spot, took all my ad accounts down with me. Hell of a welcome.

We had one big TV spot to make and no way to test the four concepts first, so I ruled out the idea my own past tests showed this kind of buyer ignores and the look that is already flooding the feed. I picked the one people get right away.

About 90% of a creative request can be dropdowns. The more interesting question is what should never be on the form in the first place.

Someone suggested removing the benefits from our checkout because he figured shoppers already knew the product, I said no way. Checkout is when people hesitate before pulling out a credit card. We can test removing them later, but I'm keeping the reasons to buy there for now.

I hit my AI usage limit twice in one day because I kept using the biggest model to do everything, including running numbers. Now I choose the model for the job and keep checking because they keep improving. I'd rather spend a minute checking than be locked out until afternoon.

We spent heavily on organic traffic at one brand and barely made anything from it. So when more organic spend came up at another, I asked what revenue and profit we could actually see. The answer was basically zero or the tracking was wrong. Until we know which, why spend more?

Why my long customer survey and our review request are never going in the same message, even though putting them together sounds efficient.

Our email agency is still running the 90 day plan I wrote. In our own reporting, email is barely contributing and they haven't brought new ideas. I'm not against agencies. I've worked with great ones. But if this doesn't improve in a few months, I'd rather bring email in house.

We spent money testing three landing pages and found a clear winner. So why would I turn around and spend more money sending traffic to the pages that lost? I wanted the next campaign focused on the winning sales page. Meta will spend the budget either way and we'd already paid to find out which page performed best. The whole point was to make that testing money back before we started spending on the other pages again.

We were losing email revenue because shoppers had moved to a new checkout, but our abandoned cart emails were still built for the old one. Then another funnel launched for a different brand and while we were getting carts, it occurred to me that we might be making the same mistake again. So I asked whether the abandoned cart emails were even set up. We'd already seen what happens when the checkout changes and the emails don't follow.

One of my managers started doing his specialist’s work when the numbers fell behind. I wanted him to change one thing before making the next call.

One of our products was out of stock and kept selling, so I said there was obviously a communication breakdown. Somebody owned inventory and had sent a sheet the day before, but flagging low stock to the people running ads was still the next step. I told my media buyer we need to track stock too as we scale and give supply enough notice to reorder. More sales coming in faster matters when getting product back takes about two months, sometimes more.

We rebuilt a competitor's offer using the same bundle and price, but I wanted to test it without adding our own extra because our checkout was already different. If conversion changed, I wanted to know whether the basic offer was working before introducing another variable. Then someone questioned whether our page copy matched the competitor's. I thought it was original, but I needed to check. There's a difference between testing the structure of an offer that's working and copying the actual words and images. I wanted to do the first without doing the second.

We ran a TV campaign targeting an older audience and I thought it performed really well. Someone brought it up recently and said it never really crushed. I disagreed. The reason we shelved it was the customer acquisition cost relative to what people were spending with us at the time. But our average order value eventually went up to almost twice what it was back then. So now we're evaluating essentially the same acquisition cost against orders that are almost twice as big. I wanted to revisit TV but we'd already entered the slow season. Then it came up again for the holidays. The campaign didn't necessarily need to perform better. The economics around it had changed.

I asked my team when our email marketing last sold one product on its own. The answer was July, and that was not even the biggest problem in those emails.

When I first sat down with the team, I told them they were probably going to get tired of hearing me talk about contribution margin. I wanted us all looking at the same numbers, coming from one tracker so we weren't wasting time debating whose data was right. That's how I like to run our weekly meetings. Show me what's working, what's not and why. If we need to dig deeper, we have the dashboard for that. The numbers should drive the decisions not how attached we are to a campaign.

Months of design work is a lot to commit to an idea nobody has reacted to yet. I wanted an early read on it. So the idea gets rendered, ads go up and people land on a test page while nothing is built yet. What comes back is some type of CPA, nothing precise. From there I can estimate how much money might be left for development at the ad budget I would use.

I gave the team a whole warning about native ad networks before realizing they were talking about a picture with a long caption in the feed. The comparison a month later was pretty clear: video was costing about three times as much as those image ads. I want more of them going to new people, as long as the sales support it.

When an outside team is only working with you temporarily, the creative they make is only part of the opportunity. I want to know what they are doing better or faster while they are still here.

A funnel builder told me it looked like we had done a ton of work on our funnels, which was fair because at the height of Q4 we had four to six tests running at once. We only stopped for Black Friday week. We had enough traffic to get data quickly, and when he asked what you can possibly split test with no traffic, I agreed with him. Any time we had a headline that worked, I wanted to see if we could beat it, so we put another one up.