In partnership with
Blu Dot surpasses 2,000% ROAS with self-serve CTV ads
Home furniture brand Blu Dot blew up on CTV with help from Roku Ads Manager. Here’s how:
After a test campaign reached 211,000 households and achieved 1,010% ROAS, the brand went all in to promote its annual sales event. It removed age and income constraints to expand reach and shifted budget to custom audiences and retargeting, where intent was strongest.
The results speak for themselves. As Blu Dot increased their investment by 10x, ROAS jumped to 2,308% and more page-view conversions surpassed 50,000.
“For CTV campaigns, Roku has been a top performer,” said Claire Folkestad, Paid Media Strategist, Blu Dot. “Comping to our other platforms, we have seen really strong ROAS… and highly efficient CPMs, lower than any other CTV partner we've worked with.”
Using Roku Ads Manager, the campaign moved from a pilot to a permanent performance engine for the brand.
Every week I get a version of the same email. It usually starts with something like, "We tried AI and it did not really work for us." Then a paragraph or two of explanation, a screenshot of a half built workflow, and a question that translates roughly to, "What are we doing wrong?"
The answer is almost never what they expect. They are not doing AI wrong. They are doing the audit wrong. Or more accurately, they never did one in the first place.
Before you can use AI to grow a business, you have to know exactly where the business is bleeding. Not in a corporate sense. Not in some 40 page strategy deck. I mean the actual unglamorous spots where time, money, and attention are pouring out the side of the bucket while everyone is busy looking at the top.
This is the work I do in the AI Audit, and this is the work I am going to walk you through today. Five leaks I find in almost every small and mid sized business. If you tighten any one of them this week, you are already ahead of most of your competition.
Grab a coffee. This one is worth the time.
Leak One: The Inbox That Eats Your Mornings
Look at your last seven days of email and count how many messages got a response that was, in essence, the same response. Onboarding questions. Pricing questions. "Do you do this kind of project." "What is your turnaround time." "Can we get on a call."
Most owners spend somewhere between 60 and 90 minutes a day on this work. That is roughly 25 percent of a normal workweek going into a task that has almost zero variation in output. It is the textbook definition of a leak. High volume, low judgment, repeating shape.
The fix is not "use AI to write emails." That is the rookie answer and it is why most attempts fail. The fix is a routing layer. You build a small system, usually inside Make.com, that classifies every incoming message into one of about six buckets. Then for the buckets that are repetitive, you draft a response in your voice using a templated prompt and drop it in your inbox as a pending reply. You still hit send. You still review. But the writing is done.
I built one of these for a consulting client last quarter. We measured the time it took her to clear her inbox before and after. Before the system, average daily inbox time was 78 minutes. After, it was 19. That is roughly an hour a day, every day, that she now spends on actual client work. Multiply by 250 working days and tell me that is not real money.
The trigger for this leak is simple. If your inbox feels like a chore by 10 a.m., you have it.
A practical starting point looks like this. Open your sent folder and pull the last 50 messages you wrote. Read them quickly and tag each one as either "this required my judgment" or "I could have written this in my sleep." If more than 30 percent fall in the second bucket, you can automate that 30 percent without losing any quality. Most owners are shocked to find that number is closer to 60 percent in their first audit.
Leak Two: The Meeting That Makes No Decision
You know the one. Recurring on the calendar. Six to nine people. Forty five minutes. By the end, everyone has talked, no one has decided, and the action items live somewhere between a Slack message and somebody's notebook that no one will read again.
Here is the harsh part. The leak is not the meeting itself. It is the absence of structured capture. You walked out of the room with information and lost 80 percent of it within an hour.
This is where Fathom earns its keep. It records, transcribes, and summarizes the meeting, then pushes the action items wherever you want them. CRM, project board, email digest, all of it. The first time you run a week of meetings through it and look at what you actually said, you will be uncomfortable. Not because Fathom is wrong. Because you finally have proof of how much value gets discussed and never executed.
I will give you a specific use case. One client of mine, a small agency owner, has a Monday production meeting that historically dragged 60 to 75 minutes. We installed Fathom, set a strict 30 minute time box, and routed the AI summary into a shared doc that automatically tags owners. The meeting now runs 28 minutes. Nothing falls through the cracks. The team gets two hours of their week back collectively.
If you are in more than five meetings a week and you cannot remember what was decided in three of them, you have this leak.
Leak Three: The Content Calendar Built On Vibes
I am going to be careful here because content is one of those areas where everyone has an opinion and most of those opinions are wrong.
The leak is not that you are not posting enough. The leak is that you are posting without a system, which means every post requires a fresh round of decision making, brainstorming, and second guessing. The cognitive cost is enormous and the output is mediocre. Sound familiar?
Here is what works. You build a content engine in three layers.
The first layer is your pillar list. Three to five core topics you commit to for the next 90 days. Not 12, not 20. Three to five. This is the constraint that does the heavy lifting.
The second layer is your prompt library. For each pillar, you have a set of structured prompts that turn a single idea into a long form article, three short posts, an email, and a video script. You write these prompts once and reuse them weekly. I host mine in ChatGPT and Claude depending on the use case. Claude tends to win for longer thoughtful pieces. ChatGPT wins for quick punchy social.
The third layer is your scheduling stack. I use Buffer for distribution because it does the boring job of putting content live without requiring me to log into seven different platforms.
When all three layers are running, content goes from being a daily decision to a 90 minute weekly commitment. The leak closes. The output triples.
Leak Four: The Lead That Goes Cold In 11 Minutes
A study I keep coming back to found that the difference between contacting a new lead within five minutes versus 30 minutes was an order of magnitude in conversion rate. Not a few percentage points. Ten times.
Most small businesses respond to leads in hours. Some in days. By the time you reach out, the prospect has already booked with someone else, lost interest, or moved on entirely. That is a leak the size of a small swimming pool.
The fix is not complicated. You set up an automation in Make.com or Go High Level that triggers the moment a lead form is submitted. Within 60 seconds, the prospect gets a personalized acknowledgment with a calendar link. Inside of five minutes, an SMS goes out as a friendly reminder. Inside of 30 minutes, you have a follow up email queued up.
You do not need to do any of this manually. The system does it. You wake up to booked calls.
I have one client who took her response time from "sometime that day" to "under three minutes" using exactly this stack. Her lead to call ratio doubled in the first month. Doubled. From the same volume of leads.
If you can answer the question, "How long does it take from when someone fills out our form to when we contact them," and the honest answer is anything more than 10 minutes, you have this leak.
Leak Five: The Knowledge That Lives Only In One Head
This one is sneaky because it does not feel like a leak. It feels like normal operations. But it is the most expensive leak of all when it finally cracks open.
Every business has a person who knows things. How the billing system actually works. Why the third Tuesday of the month is always weird. Where the contracts live. What the password to the legacy system is. The names of the three vendors who actually deliver.
When that person takes a vacation, the business slows down. When that person leaves, the business breaks. When that person retires, you have a six month rebuild on your hands.
The fix is to externalize the knowledge into a searchable, AI queryable second brain. I use a combination of Clay for relationship and contact intelligence and a structured Notion or Drive setup for process documentation. Then I run weekly summarizations through Claude to keep everything organized.
The result is a business where any team member can ask, "How do we handle a refund on a recurring subscription," and get a complete answer in 15 seconds, not a Slack message to the one person who knows.
The fix is to externalize the knowledge into a searchable, AI queryable second brain. I use a combination of Clay for relationship and contact intelligence and a structured Notion or Drive setup for process documentation. Then I run weekly summarizations through Claude to keep everything organized.
The result is a business where any team member can ask, "How do we handle a refund on a recurring subscription," and get a complete answer in 15 seconds, not a Slack message to the one person who knows.
A practical starter move is to spend one Friday afternoon documenting the five processes that only live in one person's head. Record yourself doing each one using Fathom or just a screen recording, then drop the transcript into Claude with a prompt that says, "Turn this into a step by step SOP a new hire could follow on their first day." You will end the afternoon with five documents that protect the business. That is roughly four hours of work for an asset that pays compound interest forever.
The Order Of Operations
Here is a question I get asked all the time. "If I can only fix one of these this month, which one should I pick?" The honest answer depends on your business model, but I will give you a default heuristic that holds up well.
If you sell time, like a service business, fix the inbox first. Your day belongs to you, not your email.
If you sell products or run any kind of e commerce, fix the lead response loop first. Speed of contact is the single biggest lever you have on conversion.
If you are a solo creator or content business, fix the content engine first. The compounding effect of consistent output is too big to leave on the table.
If you have a team of five or more people, fix the meeting capture leak first. Misaligned execution is the silent killer in small teams and Fathom plugs it cleanly.
If you are about to grow, hire, or sell, fix the knowledge leak first. Future you will thank present you with embarrassing levels of gratitude.
You do not have to pick perfectly. You just have to pick. Inertia is the most expensive leak of all and it is not on the list because it is the foundation under all five.
Putting It Together
Five leaks. Inbox, meetings, content, leads, and knowledge. If you fix even two of these in the next 30 days, you will recover somewhere between 8 and 15 hours of personal time per week and likely add measurable revenue on the lead response side alone.
The mistake most owners make is trying to fix everything at once and ending up fixing nothing. Pick one leak. Spend a week on it. Build the system. Test it. Move to the next.
This is the framework I walk through in the AI Workflow Blueprint, which gives you the exact prompts, automations, and templates I use with my private clients to plug all five leaks. If you want it, just reply to this email with the word BLUEPRINT and I will get it to you.
If you want the full deep dive with implementation calls and a 90 day plan, the AI Business Accelerator is where that lives. Reply with ACCELERATOR for details.
The audit is the unsexy part. The wins it unlocks are not.
See you tomorrow on The Automation Edge.
Jordan Hale
The AI Newsroom
Are you tracking agent views on your docs?
AI agents already outnumber human visitors to your docs — now you can track them.