Most Lebanese Stores Cannot Afford Meta Ads Yet

A merchant tells me Meta does not work for their store. They spent four hundred dollars over a month, got a handful of orders, and stopped. The conclusion sounds reasonable. It is almost always wrong, and not because Meta is better than they think. It is wrong because nothing was tested. At that budget the campaign never left the learning phase, which means the result they are describing is not a result. It is a coin toss they paid for.

This is the part of paid media that gets skipped in every budget conversation I have in Beirut. There is a floor. Below it the platform cannot optimise, the numbers you see are noise, and the honest answer to "did it work" is that you do not know yet. Working out where that floor sits for your own store takes about ten minutes, and it is the single most useful ten minutes you can spend before putting money into Meta.

The floor is a documented number

Meta's own help centre describes the learning phase as the period in which the delivery system is still learning how to deliver an ad set, and states that an ad set generally exits it after roughly 50 optimisation events in a week. That figure is Meta's, not mine. It is also per ad set, not per campaign, and it resets when you make a significant edit.

Read that sentence as a merchant rather than as a media buyer and it says something uncomfortable. If your optimisation event is a purchase, you need about fifty purchases a week, from one ad set, before the system is delivering on anything other than guesswork. Not fifty in total. Fifty a week, repeatedly.

So the question is not what you want to spend. It is what fifty purchases a week costs you, because that number is set by your cost per purchase and nothing else.

The arithmetic, with the assumptions showing

These are modelled figures on stated assumptions, not measurements. Put your own numbers in, because the shape of the answer matters more than mine.

Take a store with a $40 average order value and a 40 percent contribution margin, so $16 of gross profit per order before any advertising. Say the cost per purchase on Meta comes in at $8, which is optimistic for a cold audience in a small market. Fifty purchases a week at $8 is $400 a week. That is roughly $1,700 a month, for one ad set, to reach the point where the delivery system has learned anything.

If the cost per purchase is $12 instead, which is more common early on, the same floor is $600 a week, or about $2,600 a month. Still one ad set. The merchant who spent $400 across a whole month was running at roughly a sixth of the minimum, and they were almost certainly running three or four ad sets at once, which divides the signal again.

This is why I am blunt about it. At $400 a month across four ad sets you are not buying a test. You are buying four separate coin tosses and then making a strategy decision on the result.

Cash on delivery raises the floor again

Everything above assumes a purchase is a purchase. In Lebanon it is not. When most of your orders are cash on delivery, the purchase event fires when someone completes checkout, and you find out whether it was real when the courier knocks.

Meta counts the checkout. Your bank counts the handover. If a fifth of cash on delivery orders come back, then fifty optimisation events a week represents about forty actual sales, and your true cost per completed sale is roughly a quarter higher than the dashboard shows. The floor you need to clear to learn anything is unchanged, but the gross profit funding it has shrunk.

That refusal rate is the number almost nobody in this market measures properly, and it is the one that decides whether any of this is affordable. My colleague Mohamad has written separately on why efficiency ratios have to be computed on cash collected rather than orders placed, and the same correction applies to every figure in this article.

What to do when you are below the floor

Most stores reading this are below it. That is not a reason to stop advertising. It is a reason to stop pretending the campaign is a purchase campaign.

The usual move is to optimise for a cheaper event that happens further up the funnel, so the ad set can actually accumulate fifty a week. Add to cart, initiate checkout, a lead, a WhatsApp conversation started. The signal is weaker and the system is learning toward something that is not revenue, which is a real cost, but it is a genuine trade rather than an illusion. You are deciding to learn something imprecise instead of learning nothing.

The second move is to stop splitting. One ad set, one audience, one market. Every ad set you add divides the events you can generate, and a merchant at this budget running separate ad sets for Beirut and the rest of the country has halved an amount that was already too small.

The third one is the one nobody wants to hear. If your margin cannot fund the floor even at the cheaper event, Meta is not your channel yet. WhatsApp broadcasts to people who already bought, organic content, and a referral offer cost almost nothing and do not have a minimum spend. I would rather a merchant spend six months building a list and then advertise properly than spend six months burning $400 a month to learn nothing.

How to tell when you have crossed it

Two things change when a campaign has actually left the learning phase. The cost per result stops swinging wildly week to week, and Ads Manager stops labelling the ad set as learning. The second is easy to check and the first is the one that matters, because a stable cost per result is what you are buying in the first place. Until you have it, you cannot compare creatives, you cannot judge an audience, and you cannot answer whether Meta works for you.

The merchant at the start of this article did not find out that Meta does not work. They found out that $400 is not enough money to find out anything, which is a different and much more useful conclusion, and considerably cheaper to reach with a calculator than with a credit card.

The longer version

I teach the full budget and breakeven method, including how to derive your own breakeven ROAS from your margin rather than copying a benchmark, in the Meta Ads budgets and ROAS guide for Lebanese e-commerce on qlev, and in more depth in the Digital Marketing and Content Strategy course. This article is the argument. That one is the method.

Sources

• Meta Business Help Center, About the learning phase, for the description of the learning phase and the approximately 50 optimisation events per ad set per week threshold, and for the learning phase resetting on a significant edit.
• Meta Business Help Center, Purchases ROAS (return on ad spend), for how Meta defines and attributes purchase conversion value, which is the figure the dashboard reports before any cash on delivery correction.

The $40 average order value, the 40 percent contribution margin, the $8 and $12 costs per purchase and the 20 percent refusal rate in this article are modelled illustrations on the assumptions stated beside them. They are not measurements of any particular store. Checked against Meta's published documentation on 25 October 2026; the learning phase threshold is platform behaviour and is worth re-checking before relying on it a year from now.