Karen Nelson Digital

Home
United States
Anaheim, CA
Karen Nelson Digital

Facebook and Instagram Ads for Small Businesses. Helping you find more time. karennelsondigital.com

08/26/2026

🤖 We gave the exact same ad data to two different AI models.

They gave us two different interpretations.

In our tests, Claude tended to catastrophize. It focused on the wrong numbers, made questionable inferences, and then built worst-case scenarios around them.
ChatGPT was generally more measured but it wasn't perfect either..

The bigger issue wasn’t which model performed better.

Both models sometimes:
🫣 Focused on the wrong metrics
😬 Recommended Meta strategies that were severely out of date
😳 Made up information that wasn’t in the data

And they presented all of it confidently.🙈

AI can absolutely help you evaluate your ad performance. We use it ourselves.
But right now, you still need enough advertising knowledge to recognize when the analysis doesn’t make sense.

Otherwise, it’s very easy to mistake a confident answer for a correct one.
Use AI. Just double-check the data and the recommendations before you make decisions with your ad budget.

08/19/2026

A campaign’s average cost per result doesn’t always tell us whether the ads are working.

For example, a campaign might have an average cost per result of $15.20. But the daily results could look like this:

Day 1: $17
Day 2: $15
Day 3: $13
Day 4: $20
Day 5: $11

The $15.20 average doesn’t show us that costs were gradually improving before one unusually expensive day pulled the average back up. It also doesn’t show us that the cost dropped to $11 the following day.

That’s why we look at more than the average when we evaluate a client’s ads.

👉We look at the direction the numbers are moving.
👉We look for unusual days that may be distorting the average.
👉And we look at whether the campaign results are still "zigzagging" as Meta tests different combinations of audiences and creative.

How closely we watch those patterns depends on the campaign.

During a short, high-stakes launch, we may look at many numbers every day because there’s a limited window to respond.

With an evergreen campaign, we have more time. We don’t need to give every daily fluctuation the same amount of weight, so the longer-term averages become more useful.

After looking at thousands of days of ad results across different accounts, we know which patterns look normal, which may improve as the campaign stabilizes, and which ones need our attention.

The cost per result matters. But our job is to understand what the rest of the account is telling us too.

And that’s one reason it can be worth paying to have experienced eyes managing your ads.

You’re not just paying someone to read the numbers. You’re paying for the experience to interpret them and decide what, if anything, needs to change.

07/29/2026

It’s very easy to tinker with your ads too much.

You see a number you don’t love, so you adjust the audience. Then you change the budget. The next day, you wonder if you need new creative.

Before you know it, you’ve made 3 changes back to back without giving any of them enough time to work.

That doesn’t mean you should launch a campaign and ignore it. We monitor campaigns consistently and make changes when the data gives us a reason to.

But a campaign built well from the beginning shouldn’t need constant adjustments just to keep it running.

Sometimes the best thing you can do for your ads is leave them alone for a minute.

07/08/2026

A couple of weeks ago, I shared a split test we were running.

That test kept going, and we ended up with something that happens more often than people think:

We had two winners.

The video got more clicks.

But the image cost less to show, and it was still bringing in signups for a little less money.

That’s why I don’t usually judge ads by clicks alone.

Clicks matter for sure.
But I’m also looking at:

✅ How much it costs to show the ad
✅ How many people are clicking
✅ What it actually costs to get the signup or lead

In this case, if we only looked at clicks, we might think the video was the obvious winner.

But when we looked at the full picture, the image still deserved to stay in the campaign.

This is also why I don’t always turn off every ad except the “winner” right away.

Facebook and Instagram usually do better when they have some room to work. If an ad starts to slow down, it helps to have other good options already in the mix.

So the short version is:

💁♀️The video got more clicks.
💁♀️The image cost less to show.
💁♀️Both were worth keeping.

And that’s why we left both running.

07/01/2026

Claude is stressing me out.

I used to think of it as my go-to for ex*****on, tasks, and systems work, while ChatGPT/Codex felt better for writing and reflection. But lately, Claude has been missing instructions, forgetting things we worked hard to set up, and making simple tasks feel excruciatingly tedious.

I’m finding myself going back to Codex/ChatGPT because it feels easier to use and more reliable.

Anyone else noticing this?

06/24/2026

When you boost a post, you’re generally running what we call an engagement campaign.

That means you’re paying Meta to find people who are likely to react, comment, share, or watch your content.

That isn’t necessarily bad. Boosting can be a reasonable place to start if you’re new to ads or don’t have someone managing more complicated campaigns for you.

But as your ad strategy grows, you’ll usually want to put more of your budget toward campaigns designed to generate sales, leads, applications, booked calls, or registrations.

We still use engagement campaigns when they have a clear job to do. They just receive the smallest share of the budget because we prioritize the campaigns most directly connected to bringing money into the business.

The boost button can be a good place to start. It’s also something many businesses eventually outgrow.

06/17/2026

Your fancy words are probably hurting your marketing.

Take the phrase “emergent curriculum.”

Early childhood educators know what it means. But many parents looking for a preschool probably don’t.

Instead, tell parents that your teachers notice what their children are interested in and use those interests to plan activities that make them excited to learn.

That’s what they care about.

You absolutely want your marketing to show that you know what you’re doing. But insider language can make you sound disconnected from the people you’re trying to reach.

People aren’t going to stop scrolling to look up a phrase they don’t understand. They’re just going to keep scrolling.

Take a look at something you’ve written recently. Is there a phrase your customers have to translate before they understand why it matters?

Might be worth an edit.

06/10/2026

We’re running a split test for a client right now where the video is getting more clicks…but the image is producing the less expensive signups.

A couple of years ago, I would have expected the video to win. Video often got cheaper delivery and stronger click-through rates, so it made sense to lean on it.

But in this particular auction, the image is cheaper for Meta to deliver. So even though fewer people are clicking, the cost per lead is better.

That’s why I don’t love blanket advice like:

“Always use video.”
“Static images are dead.”
“Meta only wants Reels.”

Sometimes video wins because people click.

Sometimes an image wins because it costs less to show.

The better answer is usually: use both.

Give Meta different formats, different creative, and enough room to find the least expensive path to the result you actually want.

05/28/2026

I built my first AI employee this week.

Her name is Callie, and she helps me prepare for and follow up after sales calls.

And yes, I know I’m personifying my AI a little bit.

For me, naming the tool helps me remember what it does, how I’ve instructed it, and how I need to prompt it.

Also, Callie helps me with calls, so the name works. And it’s more fun.

Before a sales call, Callie reviews the information I have about the prospective client, looks at their website, and helps me create a prep document.

After the call, she reads the transcript and grades me on a scale of 1 to 10 in a few different categories based on the sales process I’m trying to practice.

Yikes. 😳

She tells me where I got off track, what I did well, and what I should work on next time.

She also helps me think through whether the prospect seems like a good fit for the kind of work we do and drafts the follow-up email.

I still make the decisions. I still personalize the follow-up. I still bring the human judgment.

But Callie helps me save time, stay more consistent, and show up more prepared.

The biggest benefit, though, is what happens during the call.

Because I’m not using half my brain to remember whether I covered everything I was supposed to cover, I can pay better attention to the person in front of me.

I can listen better, ask better questions, and offer more helpful insight in the moment.

That is my hope for AI: that it makes our humanness matter more, not less.

Address

800 W Katella Ave
Anaheim, CA
92802-3415

Website

http://karennelsondigital.com/

Alerts

Be the first to know and let us send you an email when Karen Nelson Digital posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Contact The Business

Send a message to Karen Nelson Digital:

Want your business to be the top-listed Advertising & Marketing Company in Anaheim?