Today's guest essay is an in-depth walkthrough from the operators behind Obvi, the maker of top-quality supplements for women.

Ron Shah and Ash Melwani are DTC legends with a track record of success running an 8-figure brand and spending $100M+ on paid ads. And today, they're pulling back the curtain on how they structure their Meta ad account.

Let's dive in...

###

In this newsletter we're taking a step back from the "bite-sized" content and going a little bit more in-depth and in the weeds.

We get asked all the time - Should I use cost caps in my campaigns?

The answer is: It depends...

Because, the real real answer is that it doesn’t make sense for all brands to run.

If you have widespread attention like we do because of Walmart or a ton of organic traffic from a community with thousands of members - then Cost Caps is a viable option.

But…

If you don’t, then we wouldn’t recommended it to start with.

Instead, we would recommend you to start by running a strategy like the one we’ve outlined below. This is the exact strategy that we’ve run at Obvi for the longest time, and I believe it is one of the best account structures for newer brands.

It is also exactly what we're using for Coffee Over Cardio & Paw Rangers.

Let’s dive into it!

((Disclaimer: Everybody’s business, offer, and positioning is very different. So yes, this structure might not work for all brands. But it will work for most brands.))

First things first: Understanding Meta’s business objectives

One of the most important things when running ads is understanding the business objectives of the ad platforms and which direction they are “evolving” in.

For Meta, we’ve been moving more and more towards a simplified and automated campaign structure for a long time - where Meta wants to give more data and control to the machine learning.

So when building our campaigns, we want to make sure that we align ourselves with that goal. i.e. - build simple campaign structures that allows Meta’s ML to do it’s job.

Setting your columns up for a perfect birds eye view

Soft metrics:

• Reach
• Impressions
• Frequency
• Amount spent
• Outbound clicks
• Cost per outbound clicks
• Outbound click-through-rate

Conversion & Intra-Funnel Metrics:

• Number of add to carts
• Cost per add to cart
• Number of checkout initiated
• Cost per checkout initiated
• Number of website purchases
• NC-CPA (New Customer Acquisition cost) - See this in TripleWhale
• NC-ROAS (Return On Advertising Spend) - See this in TripleWhale

Creative analysis metrics

• Thumb-stop ratio (3-sec video-view / impressions)
• Hold-rate (75% video view / impressions)
• Drop-off rates for the following ranges
• ~0% to 25% video view-rate
• ~25% to 50% video view-rate
• ~50% to 75% video view-rate
• ~75% to 100% video view-rate

The Obvi™ Account Structure

Let’s dive into the “meat” of this newsletter

Our account and campaign structure.

We’ve tried our best to visualize it below, but in case you need further explanation - this is how we set it up

• Each product has its own unique testing and scaling campaign(s)
• ~Testing Campaign
• ~ASC Campaign For Video Winners
• ~ASC Campaign For Static Winners
• Within the testing campaign, we run tests in individual ad sets on ABO.
• ~Preferably, each ad set is it’s own angle
• ~Preferably, each ad set is also broken up by creative type (video/static)
• ~Note: If you don’t have enough creatives to test at a given time, you don’t need to break them up as complex as above.
• ~Note: We try to put around 4-6 ads in an ad set at a single time.
• When possible we prefer to run testing with DCT (Dynamic Creative Testing) and once we’ve found a winner - we take the post ID and plug it into the scaling campaigns

How we set up our Creative Tests

Whenever we run our creative tests, we always set it up using DCTs.

The reason why we do it this way is simple…

We believe Meta and their ML/AI is smarter than us

So we don’t want to analyze each variable and cross-variable combination ourselves and then make a decision

We’d rather give Meta all of our creatives and then let the machine determine which one is best (for example, by rewarding it the most spend)

That being said…

You don’t want to just stuff your DCT’s with a shit ton of creatives

Because then you’re defeating the purpose, because the data for each variable-combination becomes way too thin.

Instead, what you want to do is this:

• Turn on DCT under the ad set settings
• Under the main ad put in 3 videos/statics
• Put in 2 primary texts
• Put in 2 headlines
• Let it run (that’s 12 combinations in total - which is enough)

How to determine your starting budget

In order to optimally test, you want to have a budget that’s high enough to hit the number of conversions needed to get you out of the learning phase.

So let’s say that your AOV is $100.

And you need a 2x ROAS or a $50 CAC

In order to leave the learning phase you need 50 conversions in a 7 or 1 day period (depending on your attribution settings - for this example we’ll use 7)

Total Budget = 50 Conversions * $50 CAC = $2500

Now, this Total Budget is spread over a week - so you need to divide $2500 by 7 - and thereby you get daily budget.

Your starting budget per campaign, or per test should be approximately $357 per day.

Launch time / budget pacing

To make sure that the budget is spread out rather evenly during the day, We always start our ads at midnight.

Now, this may be “bro-science” - but it’s just the way we’ve always done it. And it has worked pretty alright for us tbh.

Testing period

Unlike what some gurus may tell you, and unlike what has been best practice in the past…

Launching new tests everyday or every other day is not the way to go

The reason why is simple

It simply creates way too much volatility in the ad account

Besides that, the ad accounts have become so fragile now - that even the smallest changes here or there will throw it into the learning phase again.

In fact if more than 20% of your total budget is in the learning phase you can guarantee that you’ll see volatility.

So to prevent that, here’s what we recommend you to do:

• Create a process for yourself
• Launch all your tests on the same day every single week
• Let them run for at least 7 days before you touch anything
• There are exceptions but evaluate those on a case by case basis
• ~Example: Extreme cases like extremely high CPC, or spending hundreds/thousands without a single conversion.
• Evaluate after 7 days, and use the data to inform your next tests

Attribution Settings

People have a ton of different opinions when it comes to attribution…

And we won’t write an entire thesis on why we think our way of doing it, is better than other peoples way of doing it.

Besides that, we’d also say that the way you tackle attribution very much depends on the product you’re selling and the customer journey for that product

You can’t reasonably compare a CPG product like ours that cost $40 with a high-end furniture product that costs $4000.

Anyway, this is what we do

• No view-based attribution
• ~You’re asking for Meta to over-report if you use any sort of view-based attribution.
• ~The only time we would ever run this is if you’re JUST starting a new brand and your only marketing channel is Meta.
• For testing purposes we run 7-day click
• ~Lower budget, will take a longer window to get 50 conversions
• For scaling, we run 1-day click
• ~Higher budget, more of a chance to hit 50 conversions in a single day

Audiences

We only recommend two types of audiences: Broad and Proxima AI lookalikes.

If you are spending less than $100k, then just go broad.

If you are spending more than $100k a month, then it definitely makes sense to add something like Proxima to your stack.

Their audiences are currently outperforming our broad campaigns and giving us 20-30% cheaper customer acquisition costs.

But… the caveat here is that you need to have a solid level of existing data and monthly spend level for it to perform for you

As for exclusions we typically only exclude our customers.

• Create a pixel based customer list (Max lookback is 180 Days)
• Create a full customer list in Klaviyo/Sendlane and sync back into your ad account

The Obvi™ Testing Methodology

Here are our notes on how to run tests and figure out when to scale or not.

First things first: Always check your blended NC-CPA, NC-ROAS in addition to in platform metrics. Are things at a healthy level and hitting targets?

If yes, then there’s 2 things we look for: Did our test leave learning in the first 7 days? Can it handle higher spend.

1. The first one is obvious, are you getting sales at your target CPA and are you getting at least 50 conversions or close to it in a 7 day period? If yes, great that’s one criteria down.
2. The second one is putting your test under stress test. Sure your new ad is performing at a smaller budget but can it handle a larger budget? This is where we will scale up our testing ad set by 19% every couple days if things look good. Not only will this give you more spend behind something that is working but it will also stress test this asset. If things hold, then it’s ready for scale. If things break, scale back down and see if things level out.
   ‍

If no, then the creative isn’t hitting. But don’t give up there, can it be saved?‍

Video Assets‍

Analyze which part of the creative is suffering

• Look at the Thumbstop Rate
• ~If this is less than ideal can you improve the first 3 seconds of the video?
• ~If this is more than ideal does the hook correlate with the rest of the video or is lacking cohesiveness
  ‍
• Look at the Hold Rate
• ~Is there a big drop off? Make changes to the body of the video, faster cuts, call outs, etc.
• ~Does hold rate look good? Is there enough cohesiveness between ad and landing page.

Static Assets

• Analyze CTR and Outbound CPC
• There isn’t much you can improve here other than making sure you’re using eye catching headlines, call outs, USPs, etc.
• So typically if a static doesn’t work, we will just rework it entirely.

All in all - Iterate creatives if you show some signs of promise but aren’t quite there yet and try to run a new test.

If performance is still poor: Kill it

How We Set Up Scaling Campaigns

There are 2 ways we scale up our daily ad budgets.

The first way is to create a separate campaign at a higher budget with your winning ads. Let’s get into this set up.

Advantage+ Shopping Campaigns

By Meta’s definition: Advantage+ shopping campaigns is part of Meta's Advantage+ products, which leverage machine learning to help you reach valuable audiences with less set up time and greater efficiency. Advantage+ shopping campaigns is designed to be the most efficient solution for performance-focused advertisers looking to drive online sales.

Great now that we know what it is, how do we use it?

ASC campaign’s basically stops the advertiser from playing around with too many variables.

It forces the advertiser to go broad and focus on the creatives.

Here is what your setup should look like:

Budget

• Our rule of thumb for this is if we are testing with an ad set at $500 daily budget, we will start our scaling campaign budget at 5x that, so $2500.
• There isn’t a set rule for this, if you want to go lower or higher give it a test and see what works best for your ad account. We also believe this depends on the history of spend levels. If your account is prime for spending a lot, then you can push the budget a bit higher. If your account hasn’t spend that much, then start lower and inch your way up.

Ads

• As mentioned above we typically have an ASC campaign per product/funnel, per creative style.
• We typically like to have 6-10 winners from tests to start up an ASC campaign.
• Take the post ids from your DCT tests and drop them into an ASC campaign.

Scaling Budgets

• Things look great? Our rule of thumb is bump budgets by max 19% at a time.
• ~Do this once every couple days for stability.
• ~Do this every day if you want to be aggressive.

What To Do With Your Testing Campaigns After Moving To Scaling

The biggest question we get it is should I turn off my tests once I move them to scaling? Well the answer is, is it performing well? If yes, then don’t touch it. In fact, this is another area where you can scale up your budget like we mentioned before.

If a test is still performing after you moved it into scale, scale it where it is. Ride the wave and enjoy the profits!

Once it stops performing, then naturally you will want to turn it off.

What’s Next?

This is where most people get complacent. Things are working? Budgets are scaling? Don’t get comfortable! To keep things going, you need to fuel the fire so continue to build out more creatives weekly, explore new angles, explore new creative concepts, templates, etc. Never stop testing.

Thanks for reading along

As always, thanks for reading.
‍

All the best,
Ron & Ash

Subscribe to Ron & Ash's newsletter for weekly eCom tips.