Every paid-media framework you've read was written for a market with CRM data, known margins and a marketing team that gets to decide. Georgian accounts rarely have any of the three. So the standard advice doesn't fail loudly. It fails slowly, one budget meeting at a time.
I went back through Georgian ad accounts I managed or audited and asked the same three questions of each: what was actually broken, what did it cost, and why hadn't anyone fixed it? Nine patterns kept coming back. They're ranked below by how much budget ran while each problem was in place.
1. The traffic was fine; the destination and the offer weren't
What it looks like. The ads get the blame while the real problem sits after the click: a site that takes minutes to load, a lead page with no form on it, a price nobody wants.
Why it sticks here. The site was built by the owner, their cousin, or an agency they're still friends with. Changing it is political before it's technical.
What to do instead.
- Triage before you test. A slow site or a missing form needs no experiment. Ninety seconds of looking finds it.
- Bring evidence the owner can't argue with. An owner who has never read a dashboard will watch thirty seconds of a real person failing to find the form on their own website.
- Pull the lever you can reach: the offer. You usually can't rebuild the site, but you can often change what's on it. At a home-and-decor retailer, one online night-shopping promotion produced a month that beat three of its five physical stores.
Post-click is often the real constraint, and diagnosing it before touching the ad account is right everywhere.
2. The account was profitable and the business wasn't
What it looks like. Spend judged by platform ROAS, not by the money left after product cost, delivery and tax.
Why it sticks here. Owners rarely share their margin. Cash sales never reach the tracking. With no CRM there's no lifetime value to weigh against.
What to do instead.
- Use MER: total revenue ÷ total ad spend. Platform ROAS depends on tracking. MER depends on the bank account.
- Ask for a bracket, not a number. "If you sell ₾100 of this, is what's left more or less than ₾30?" A bracket gives you a floor, and a floor is all you need.
At a large ticketing site, moving the target from platform ROAS to MER and profit-ROAS changed which campaigns survived, and the account turned.
MER and profit-first thinking are established practice. What's local is how you get the inputs.
3. Everyone plays the small market the same way
What it looks like. A playbook built for big markets, run in a market with two or three real competitors.
What to do instead.
- Buy share, not value. In a two- or three-player auction you aren't buying conversions efficiently. You're denying the click to your competitor. Value bidding optimises for the wrong objective. Run max clicks with a CPC cap at 3–5× your average CPC, and treat impression share above 90% as the KPI.
- Model the ceiling. It's real and it arrives fast. One B2B studio reached total demand in month two, and the budget kept running for months after. Estimate the searches that exist. At 90% impression share, more budget buys nothing.
- Sequence channels; don't spread them. ₾5,000 split across three channels gets none of them to the volume where they work. Take one channel to saturation, then open the next.
- Build the warm pool. If every campaign starts cold, nothing compounds.
- Treat the product feed as infrastructure. It's usually the real barrier to opening a second channel.
The auction mechanics. What breaks is every recommendation built on top of them.
4. Nobody said what winning meant
What it looks like. Budget allocated on instinct, requests with no traceable origin, and no conversion the business would stand behind.
Why it sticks here. The English answer is a stakeholder alignment workshop. Here there are no stakeholders: there's an owner with an opinion. And with no margin number, you can't set a ROAS target anyway.
What to do instead. Don't wait for alignment. Write the definition of a win yourself and get it ratified. Write it in the owner's units (orders, money at month end, calls answered), never in platform vocabulary. Get one sentence of agreement in writing before the money is spent.
The definition has to exist, be written down, and come before the spending.
5. Nobody knew which products actually made money
What it looks like. Thousands of products, no margin data, and ads promoting whatever the commercial team feels like pushing that week.
Why it sticks here. With a catalogue that broad, individual products never collect enough data, at $50K a month or ₾5K a month alike. And the standard method, bucketing by margin, is dead without margin.
What to do instead.
- Bucket by category, where the volume is. Start from the owner's gut ranking of bestsellers. They know their winners even when they know nothing else.
- No margin? Bucket by price. Shopping is a price-comparison surface. Show the cheapest products in each category and keep the expensive ones out of the placement. CTR goes up without anyone ever telling you the cost price.
The stars / opportunities / zombies logic. It just runs on categories and revenue rank instead of products and margin.
6. The sale happened where you couldn't see it
What it looks like. Physical stores, phone calls, a sales manager's memory. Marketing can neither prove nor disprove its contribution, so the blame war with sales never ends.
Why it sticks here. The standard answer is "connect your systems." There are no systems to connect, and often nobody is willing to buy one.
What to do instead.
- Ask. A short in-store questionnaire at one retailer showed roughly 20–25% of buyers came from the ads. It took days and cost nothing.
- Tag leads with sales. Agree a qualified-lead definition; the sales team tags every lead against it.
- Use the offer as an instrument. Run a promotion advertised in one place only. Everyone who redeems it came from there.
- Hold out a region or a week instead of tracking individual users.
Then the part that matters most: say the error band out loud. "Roughly a fifth of walk-ins come from us, give or take ten points, and here's how I got that" beats a dashboard number nobody believes.
Almost nothing. This is the cleanest break from the standard playbook.
7. The creative was written for nobody
What it looks like. With broad targeting, the creative is the targeting. At one real-estate developer, leads arrived believing they could move in next month, into a building two years from completion. The sales team saw it on day one. No dashboard ever did.
Why it sticks here. The standard way to build an ideal customer profile is to query the CRM. There's no CRM to query.
What to do instead. The sales team and the owner are the CRM. Build the profile by structured interview, not by query. Then check it against the sales team's lead-quality comments, which arrive the same day, rather than against sales outcomes, which arrive next quarter.
Creative is the targeting layer now, and concepts have to be built per segment. What's local is where the segment data comes from.
8. The creative was doing a different job than the campaign
What it looks like. A brand film running under a lead-gen objective. Hard-sell statics used for awareness. No way to score winners and losers, and no refresh rhythm.
Why it sticks here. A ₾3,000–15,000 budget can't fund parallel variants, there's often no content team, and the owner supplies the creative. At one developer, management's brand video won by default because nobody could prove it was wrong, and the budget produced no sales.
What to do instead.
- Judge on the signals that arrive fastest: hook rate, hold rate, CTR, cost per landing-page view, cost per add-to-cart. Use conversions as confirmation and veto, not as the first signal.
- Test direction, not variants. Don't A/B a headline on ₾5,000. Test whether the ad shows a price, whether a person speaks to camera, whether the call to action lands in the first three seconds.
- Refresh on frequency and CPM drift, not when CPL finally climbs. By then you've burned a month.
The creative principles themselves. What breaks is how often you can judge and how much you can test.
9. You saw it, said it once, and let it go
What it looks like. The problem was spotted and named, then dropped. Nothing changed, because nobody owned the fix. This is the pattern that lets the other eight survive.
Why it sticks here. Paid-media education in any language teaches tactics. It doesn't teach how to make a business act on what you found, and here the person you have to convince is often the founder, with personal money at stake.
What to do instead.
- Present a decision with a price tag, never a dashboard. "This is costing you about ₾4,000 a month" lands. A screenshot of Ads Manager doesn't.
- Put it in writing once, with a date. A dated one-paragraph note turns "I mentioned it" into evidence, and changes who owns the consequence.
- Use the owner's units. Orders, walk-ins, money at month end. Never CPL or ROAS.
- Know when to stop. Leaving an engagement you know is wasted is a legitimate professional move.
The order to fix them in
The list above is ranked by cost. That's not the order you fix things in. If you don't know where to start, use this order:
- Define winning (step 4). Everything else needs it.
- Map where the sale happens (step 6), so you know what you can and can't see.
- Measure against the business, not the platform (step 2).
- Find the products that make money (step 5).
- Prove the bottleneck is after the click (step 1). The ninety-second check for a slow site or a missing form can run on day one.
- Build the customer profile (step 7), then brief creative against it (step 8).
- Fit the strategy to the small market (step 3).
And step 9 runs through all of them: it's how each fix gets approved.