Daniel Popa

Growth Partner and Marketing Consultant I help ambitious startups scale profitably through paid ads, conversion rate optimisation, and data-driven growth.

[Daniel Popa @danielpopa_me]

Free PPC Audit for startups & SMEs.

If you’re running ads and not sure whats working, I will audit your account and send you 3 fixes to improve ROAS + CPL.
#PPC #startup #startuplife #StartupGrowth #ROAS

[Daniel Popa @danielpopa_me]

Keyword expansion used to take me hours. Not because it's hard.

Because before any thinking, you have to export the search term report, format it, strip out the rubbish, cross-reference what's converting, work out what's next to it, and find what's been burning budget.

Now I ask @AdPlugMCP three questions:

What search terms are already converting?
What's close to them that we're not targeting yet?
What's been spending without converting?

Live account data. No export, no formatting, no headache.

That's what AI is for. Not the judgement, I keep that. The hours of admin standing between me and the judgement.

[Daniel Popa @danielpopa_me]

Every week someone declares the PPC job dead.

AI does the bidding. AI writes the ads. AI runs the account. Pack up, specialists.

12 years and $100M+ in spend later, here's my less exciting version.

AI in PPC is real. I use it every day. It builds my Monday report before my coffee goes cold and audits accounts at a scale I never could by hand.

That's enablement. Not replacement.

What it can't do:
- decide which client goal matters this quarter
- spot a "great" CPA that's hiding broken tracking
- see why a campaign that looks like it's losing is doing exactly its job

The people posting "PPC is over" are usually selling the tool. The people quietly using it are still the ones deciding what to do with it.

[Daniel Popa @danielpopa_me]

I haven't opened Looker in months.

My Google Ads dashboard is one sentence to Claude now. Spend, conversions and CPA by campaign, the daily trend, and the search terms burning money.

Google Ads connected through @AdPlugMCP. Demo account in the video.

[Daniel Popa @danielpopa_me]

One of my LinkedIn posts went semi-viral. I'm still not sure what that means.

What I do know: for the first time I got AI comments. You know the ones.

"The interesting part here is…"
"What you're really pointing at is…"
"The real shift is…"

Behind each one is a real person who's been told to "engage in your niche", pasting the post into a tool without reading it.

Like speed-reading the menu at a restaurant you don't want to be in, ordering something random, and leaving before the food arrives.

Save your AI credits and write something real. It'll do more, promise.

[Daniel Popa @danielpopa_me]

Every founder should be posting.

Some posts get 200 views. Some get 20k. Most land somewhere in between.

You'll get replies from people who didn't read it. You'll get DMs selling you a fix for a problem you don't have.

There's no formula. If there was, everyone would have found it and the feeds would be even more unbearable.

You don't post to go viral. You post because the people who need to find you can't if you're quiet.

Being quiet is very comfortable and completely useless.

[Daniel Popa @danielpopa_me]

The quickest way to make Claude worse at your ad accounts is to connect everything to it.

More tools to choose from means more wrong picks. Ad accounts first (mine go through @AdPlugMCP), then analytics and the CRM, all read-only to start.

[Daniel Popa @danielpopa_me]

I'm convinced the LinkedIn algorithm has mood swings.

Personal story: 1,700 impressions.
Detailed industry insight: 400.
Post I thought was average: my second best ever.
Post I was proud of: went out on a Friday, seen by nobody.

A month of posting every day and I still had no idea what it wanted. So I kept posting. Pretty sure that's the only strategy that works.

The post where I moaned about all this then did 26k impressions. Still none the wiser.

[Daniel Popa @danielpopa_me]

Microsoft Ads is always the account people check last. That's why its leak runs for weeks.

I ask ChatGPT which partner sites spent with zero conversions (Microsoft Ads plugged in through @AdPlugMCP). It waits for a yes before excluding anything.

[Daniel Popa @danielpopa_me]

Campaign pacing problems used to be something I'd catch on a random Thursday. Three days after the budget started doing something weird.

Way too late. Hands up.

Now @AdPlugMCP flags it the same day. If a campaign is on track to blow its monthly budget by the 18th, or spend half of it, I know before the client does.

Nobody writes case studies about catching a pacing issue early. But boring problems are exactly where AI earns its keep.

It isn't doing anything clever. It's checking the thing I used to have to remember to check, on top of everything else I had to remember.

[Daniel Popa @danielpopa_me]

75% more revenue on 24% less ad spend.

The fix was a conversion I made up.

The campaign was starved. Smart Bidding needs enough conversions in a rolling 30 days to learn anything, and this B2B account was nowhere near. So it guessed. Expensively.

The usual advice is loosen the target or go back to manual. Both mean agreeing to run a worse account.

So I went earlier in the funnel:

Found the steps that had volume
Worked out which ones predicted a decent lead and which predicted a tyre kicker
Built one conversion out of them, good signals weighted heavier

Suddenly it had enough to learn from, and it was learning from the right sort of person.

It took me embarrassingly long to stop blaming the algorithm and look at what I was feeding it.

The algorithm was never the problem. The diet was.

[Daniel Popa @danielpopa_me]

Cheap leads are the most expensive thing in a B2B ad account.

Put cost per lead next to cost per deal in Claude (Google Ads and HubSpot through @AdPlugMCP) and there's always one campaign with cheap leads and no deals.

It's usually someone's favourite.

[Daniel Popa @danielpopa_me]

ChatGPT, Claude, Gemini or Perplexity. Which is best for marketers?

Wrong question. The model isn't what's holding you back.

A great model working from a two-week-old CSV is still working from two-week-old data. Flawless reasoning, stale answers.

In paid media, the account data sits behind three tabs and a download button. You pull it once a week if you're disciplined.

Connect the same model to live account data and the questions change completely.

Not "summarise last week's report."

"What changed in the account in the last 48 hours, and is it worth acting on?"

Fix the data access problem before the model debate. It's the whole reason I built @AdPlugMCP.

[Daniel Popa @danielpopa_me]

The best accounts now aren't the ones with the most campaigns.

They're the ones with enough signal in each campaign for the algorithm to do its job.

[Daniel Popa @danielpopa_me]

The Google Ads account structure that made you look like an expert in 2018 is killing your performance in 2026.

[Daniel Popa @danielpopa_me]

So the account looks disciplined and performs badly.

The fix is consolidation. Merge the splits that only made sense under manual bidding, keep the ones that reflect real business differences:

Different margins
Different conversion goals
Different landing pages

[Daniel Popa @danielpopa_me]

You can target directors on LinkedIn and still pay to reach their juniors. Audience Expansion is usually why.

Claude puts who LinkedIn reached next to who became deals in HubSpot, with both connected through @AdPlugMCP.

[Daniel Popa @danielpopa_me]

All five run through @AdPlugMCP in Claude or ChatGPT, and the prompts are free to copy: https://t.co/12nX55Kyox

[Daniel Popa @danielpopa_me]

Search Console is the most useful free tool Google makes, and the hardest one to get a straight answer out of.

Since I plugged it into Claude, I just ask. Five questions worth starting with:

[Daniel Popa @danielpopa_me]

5. "Which search terms am I paying for that I already rank top 3 for?"

Needs Google Ads connected too. Maybe sit down first.