Meta Ads Mastery

I plan, launch and optimize Meta Ads accounts for multiple clients at Limi Creatives in Kathmandu, and from December 2025 to June 2026 I ran paid ads and SEO remotely for Minnesota-based clients on contract with SoloiTech. That matters for one reason: I have shipped this work for American businesses, without an agency layer between the founder and the person touching the account.

When you message me about a spike in cost per result, I am the one who opens Ads Manager. Not a strategist who relays it to a media buyer who replies on Friday.

What I change in the first two weeks of a Meta account

The first fortnight is diagnosis and demolition, not new ideas. Almost every account I inherit has the same four problems, and adding campaigns on top just spreads the budget thinner.

I start with the dataset, not the campaigns. I check whether the pixel and the Conversions API are both firing, whether they are deduplicating properly on a shared event ID, and what the event match quality looks like for the purchase or lead event. An account sending browser-only events with no server-side backup is reporting a fraction of what it earns, and the bidding algorithm is learning from that same incomplete picture.

Then I map the account against its actual conversion volume. Meta wants roughly fifty optimization events per ad set per week before delivery stabilizes. An account producing a handful of conversions a week, split across six ad sets, will never get any of them out of the learning phase.

I audit the aggregated event measurement priority order, because the eight-event limit per domain is still there and still misconfigured in most accounts I open. If "ViewContent" outranks your purchase event, you have told Meta at the domain level to optimize toward the wrong outcome.

Last, I pull the creative report and sort by spend. Usually one or two creatives are carrying almost the whole budget while the rest were never switched off. That first efficiency win costs nothing to take.

Why I collapse ad sets instead of adding more

The instinct when results dip is to build more: another audience, another lookalike, another interest stack. On a small budget this is the single most reliable way to make performance worse.

Every ad set you add splits the same conversion signal into smaller pieces, and each piece re-enters the learning phase on its own. You pay for six separate education processes instead of one. I consolidate hard: one prospecting campaign with a small number of broad or Advantage+ audiences, and one retargeting campaign with a genuinely warm audience behind it.

Detailed interest targeting is not what it was. Meta's delivery system finds buyers from creative signals and conversion data far better than from a stack of interests someone guessed at in a planning document. I keep exclusions, geography, age and language, because those are constraints rather than guesses. I rarely build twelve-interest audiences, because I have not seen them beat broad at the account sizes I work with.

Campaign budget optimization goes on almost every time. Cost caps come later, once there is enough conversion history for a cap to mean anything, because a cost cap on a cold account usually just stops delivery and someone concludes Meta is broken.

Creative is the targeting layer now, so testing has to be a system

If audiences no longer differentiate accounts, creative does. That means creative testing cannot be occasional. It has to run on a schedule with rules decided before the test starts.

I test at the concept level first, not the detail level. Concept means the angle: problem-led versus outcome-led, founder on camera versus product demo, social proof versus offer-forward. Testing three thumbnails of one video answers a question nobody asked. Testing three different reasons to care answers the one that matters.

Only after a concept wins do I test variants inside it: hook rewrite for the first three seconds, static versus motion, caption length, call-to-action framing. I work with the content and design teams at Limi Creatives on this, so ad creative and the landing page it points at get built as one thing rather than handed over separately.

I decide the kill threshold before launch. A creative gets a defined budget and a defined window, and if it has not produced a result by then it goes off, regardless of how much anyone liked it in the review call. Judging creative by three-second video views or engagement rate is how accounts end up full of popular ads that sell nothing.

What does not work: boosting organic posts and calling it a campaign, running the same creative for four months because month one was profitable, and refreshing creative on a calendar without checking frequency and first-time impression ratio. Fatigue is measurable, not a monthly ritual.

The tracking layer most Meta accounts are quietly missing

Meta will happily report conversions it cannot substantiate, and it will just as happily miss conversions it genuinely drove. Both errors push budget in the wrong direction, which is why I treat the landing page and tracking side as part of the same job rather than a separate project.

Practically, that means Conversions API set up server-side rather than through a one-click integration, event deduplication verified in the test events tool, and a real look at the attribution setting on each campaign. The seven-day click and one-day view default is not neutral. For a considered purchase with a long deliberation, it undercounts. For an impulse product, view-through credit can flatter the account badly.

For lead-gen clients I push for offline conversion upload, so Meta learns from qualified leads and closed deals rather than form fills. A form is not a customer. An account optimized toward form volume will deliver cheaper and worse leads every month until someone changes what it is aiming at.

When cost per lead spikes and the algorithm is not the reason

Clients almost always report this as a Meta problem. It usually is not. Before I touch bids, I check five things in order.

1. Did tracking break? A theme update, a consent banner or a redesigned thank-you page can silently stop events, and the platform reads that as your ads failing.
2. Did a competitor enter the auction, or did the season change? Auction pressure is real and it is not a strategy failure.
3. Is frequency climbing while unique reach is flat? That is fatigue, and no bid adjustment fixes it. New creative does.
4. Did the offer or the landing page change? A price rise or a longer form moves cost per lead without anyone touching the ad account.
5. Did someone edit an ad set and reset learning? Large budget changes and edits to targeting or optimization goal restart the learning phase.

Only after those five come out clean do I accept that it is a delivery and bidding problem. I would rather tell you the ads are fine and your form is broken than bill you for optimization that was never the issue. If you want the wider argument about whether the channel still pays for itself, I wrote about whether Meta Ads are still worth the spend separately.

Businesses I tell to skip Meta Ads

Meta is a demand generation channel. It interrupts people who were not looking for you. That works well for visually demonstrable products, offers with an obvious hook, local services with a wide catchment, and anything with an impulse or emotional trigger.

It works badly for high-consideration B2B with a tiny addressable market, for products people only buy at a specific moment of need, and for businesses whose entire funnel is one page with no follow-up. If someone only searches for your category when their boiler fails, search campaigns for people already looking to buy will beat paid social every time, and I will say so before you spend anything.

I will also tell you if your budget is too small to learn anything. Below a certain monthly spend, the account cannot generate enough conversion events for the system to optimize, and you are paying for noise. Ads are also not a substitute for demand you have not built, which is why I usually pair them with organic search as the long-term counterweight.

What working with someone in Kathmandu actually looks like

I am in Nepal. My clients are overseas. In practice that has been an advantage rather than a friction point, and I would rather explain why than pretend distance is irrelevant.

Overnight for you is my working day, so account checks, budget pacing and creative uploads happen while your business is closed. I hold a live overlap window for calls, and I write down what changed, why, and what I expect it to do. My background is computer science at the Institute of Science and Technology, Tribhuvan University, and I spent part of 2025 and 2026 teaching it to class 11 students, which is decent training for explaining an attribution window to someone with no wish to become an expert in one.

You keep ownership of everything: Business Manager, pixel, ad account, creative files. I get added as a partner, not an owner. If we stop working together you lose nothing but me.

You can see campaigns I've run, and if you are still deciding where the first dollar should go, how the two channels trade off is the honest version of that comparison. Otherwise, send me your account and the number you need it to hit, and let's talk through your ad account.

Frequently asked questions

How much should I spend on Meta Ads to get useful results?

Enough that the account produces steady conversion events, because Meta's delivery system needs roughly fifty optimization events per ad set per week to stabilize. Below that, results swing wildly and you are paying for noise rather than learning. If your budget cannot support that on your main conversion, optimize for an earlier event or start with search instead.

Why did my Meta cost per lead suddenly go up?

Check tracking, auction pressure, creative fatigue, offer or landing page changes, and recent ad set edits before blaming the algorithm. Broken events after a site update are the single most common cause I find. Rising frequency with flat reach means fatigue, which only new creative fixes. Large budget or targeting edits reset the learning phase and cost per lead climbs temporarily.

Do I still need detailed interest targeting on Facebook Ads?

Usually not. Meta's delivery system now finds buyers from creative signals and conversion history more reliably than from stacked interests. I keep hard constraints such as geography, age, language and exclusions, and let broad or Advantage+ audiences do the rest. Layering many interests on a small budget mostly fragments your conversion signal across ad sets that never exit learning.

What is the Conversions API and do I actually need it?

It sends conversion events to Meta from your server rather than only from the visitor's browser, so events survive ad blockers, tracking prevention and consent restrictions. Paired with the pixel and deduplicated on a shared event ID, it improves both reporting accuracy and the data your bidding learns from. For any account spending real money, yes, you need it.

Can you run Meta Ads for a US business from Nepal?

Yes. I ran paid ads and SEO remotely for Minnesota-based clients on contract with SoloiTech between December 2025 and June 2026. Your overnight is my working day, so optimization and uploads happen while you are closed. I keep a live overlap window for calls and document every account change in writing.

Should I run Meta Ads or Google Ads first?

If people already search for what you sell, start with Google, because capturing existing demand is cheaper than creating it. If nobody knows your category exists, or the product sells visually, start with Meta. Businesses with an urgent trigger, such as emergency repairs, should almost always fund search before paid social.