1,300–2,000 RUB purchases from test

Promoting a ceramic artist via Instagram + site.digital (a Russian “website builder / landing platform” ecosystem)

This case happened before I, as the owner and founder, started building an actual team.

Back then I only had an assistant, Natasha, and a designer I knew — Varya.

There was no “us”.

Still, I’ll present it as “us”, and then the story will be told from me personally, as the lead on the project. A subtle hint that TODAY we can do it even better.
Because I’m more professional now, and a team lets you deliver results with more depth and stability.

We are The Quiet Orbit. A full-stack digital marketing agency.

We do both promotion and what should come before it: brand strategy and marketing strategy.

In our experience today:

Total projects — 87.
B2C — 71 of them.
B2B — 10 of them.
The rest are mixed.

We’ve already managed 3,683,498.40 roubles (RUB, Russian currency) in ad spend.

And now… a long time ago, in a galaxy far far away

**This is what I looked like when I wrote the case!**

This is me, Bogdan — the case is narrated from my perspective

About the project in this case:

Starting point:

Meet Tanya. The best way to explain what she does is the text we wrote for her website during the project.

How we met:

TikTok brought us together. The year before, I ran my TikTok as an internet-marketing channel for artists and personal brands. Then I added commentary on social and political topics. That’s how Tanya and I met — through values and the nature of her work. She’s a very typical representative of the kind of clients I work with.
And she’s also just a good person.

What assets we had:

An Instagram account with ~3,500 followers.

Her TikTok with almost 10K followers, but it’s not relevant here. She did everything there on her own.

What we had to finish:

Tanya came with a request: find customers via targeted ads in Instagram (Meta’s platform; in Russia access can be complicated, but the ad tools still existed at the time). Not for a long-term retainer. The initial plan was short cooperation, then she would come back a couple more times, and over time we’d move into ongoing work.

I looked at the actual mechanics of the request. And I decided we needed a proper landing with an FB Pixel (Meta Pixel — a tracking script). For those outside the bubble: it’s a small piece of code that tracks behaviour on the site and can match visitors to Facebook/Instagram profiles (when possible), then adds them into an audience list. Later you can run retargeting ads and/or build “similar audiences” — I call them LaL for short (Lookalike).

We debated between taplink (Taplink — a popular “link in bio” mini-site service in RU-market) versus a simple CMS website (a site builder).

We ended up choosing a real website. Because it’s both effective and long-term. Also, a website lets you verify the domain in Facebook Business Manager (Meta Business Manager), which matters for tracking, especially after iOS 14+ restrictions. The idea was: after domain verification, iOS 14+ user actions could be mapped into one of 8 “standard events” (Meta’s Aggregated Event Measurement concept).
At least that’s what Facebook said back then:

We also needed to polish Tanya’s Instagram “posadka” (посадка — the profile / landing setup you send traffic to), because people would check it anyway. Paid traffic, organic traffic — doesn’t matter. I wanted better conversion efficiency and fewer leaks.

One of the best decisions was building a website.
You’ll see why in a second…

Start of work:

First — a brain-storm

First, you need to crystallise the USP — UTP (УТП, “unique selling proposition” in Russian marketing slang).
That’s the base layer. Without it, it’s hard to create anything decent. Everything grows from it: copy, creatives, the angle you use to talk to your audience.
Even Hunt’s Ladder (Лестница Ханта — an awareness ladder). We’ll get back to it.

There are many methods. For example, 5W — and you answer “Why” questions: WHY your client will buy, won’t buy, etc.

In my practice I eventually had to reinvent a bicycle — my own version. The methods are fine, I’m not arguing, but they didn’t give me enough depth to later deduce specific solutions that express a promotion strategy.

That’s how I started interviewing people using Descartes’ Square (Логический квадрат Декарта — a structured logic framework built around four questions).

You see, it’s logic. And I love logic.
Descartes’ Square works when you set identities correctly: what “this” is, and what “will happen”.

“This” is a subject with a predicate. In Tanya’s case: author-made ceramics. Ceramics is the subject; “author-made” is the predicate.

At the same time, Tanya is also a subject, with her predicate being “ceramic artist”.

And “what happens” is both: Tanya continuing to be a ceramic artist, and people coming to Tanya for author-made ceramics.

You might not yet see why I’m making it this complicated, but you can sell a mug by simply putting it on a shelf in Auchan (Ашан — a hypermarket chain that used to be a normal everyday reference in Russia). You don’t even need a fancy “author corner”. There are rows of mugs hanging there in chorus. Simple: a person wants tea, they don’t have a mug — here, take a standard one. There are hundreds like it, sure, but do you want “the taxi sign” or do you want to get to the destination?
People come to Tanya for the “taxi sign” — when the decision to “get there” is already made.
And that’s abstraction, like most values in a human community.
The better I build this abstract cloud of meanings around Tanya’s work, the sharper and heavier my ad messaging becomes.

The brainstorm took 2 hours live. The method has its own quirks: how to interview, how to steer introspection, how to regulate frames, where to focus attention, how to catch the moment someone tries to dodge the second question by swapping it with the first.
You can learn it by reading a book on Focused Interviewing. I read it cover to cover.

I can write a separate article (or even run a webinar) about using the Square — if that’s what you’d want.

But the output was this mind map:

While we were unpacking her USP, I asked my assistant to dig into Facebook Ads Manager (Meta Ads Manager), because brand owners often try to run something themselves, and their “messy attempts” can reveal details you wouldn’t notice immediately.

Natasha (that’s her name) helped build a table where we could see click rankings across Tanya’s creatives and settings.
It’s better than nothing. A client can’t be wrong about everything. It’s smarter to lean on some real signals.
If we had nothing — we’d build from nothing. A/B tests were waiting for us anyway.

Natasha also drew Hunt’s Ladder (Лестница Ханта — awareness ladder showing how “warm” traffic is). The higher a person is on the ladder, the harder it is to pull them in with your USP — but it’s easier to find such people at scale 🙂

For example, the very first step is “No problem”. It’s the hardest group to sell to, because they don’t even recognise the task your product solves — assuming you correctly guessed the audience profile that might have that problem in the first place.

This is the ladder we got. It has descriptions, but it’s not a full audience avatar.

Choosing a CMS for the website

We based the choice on these points:
1. No dancing around a hosting setup with a million custom settings.
2. Convenient to build both a portfolio and a small ecosystem of pages.
3. Friendly for analytics tools.
4. Not priced like 700 roubles per month (RUB) like Tilda (Тильда — a popular Russian website builder; cheapest when paid yearly).

In the end we chose between Wix and Wfolio (Wfolio — a portfolio-centric site builder popular among photographers in RU). The second is more “for photographers”, the first is more universal.
But I had worked with Wfolio before and I knew how painless domain verification and pixel installation is there — plus Yandex Metrica (Яндекс.Метрика — Russian analytics) and Google Analytics setup.
I knew we wouldn’t get stuck on tracking; we only needed to understand the page structure and how to present information.
And since Wfolio was built for photographers, images look delicious there. Tanya had a sea of beautiful photos of her work.

Also the default design templates look strict. It creates a sense of product level.

I drew another mind map:

I didn’t implement this map fully :))) But it helped me avoid starting from zero.

We needed a “logo” and a “logotype” (in Russian practice these are often treated as separate: symbol vs wordmark). While I was configuring Wfolio, filling pages, and setting structure, the designer worked on variations.

Varya (the designer) is strong. She can hear needs when you articulate them properly. In the end we shipped these beauties.

First:

Second:

The copy came out easily

I just had to let myself breathe for a moment.
The best thing about brainstorms like the one above is that you don’t always need to stare at the map to write from it. Everything is already soldered into your head.

I wrote exactly five texts, two of them were rewrites of Tanya’s existing site copy.
Still, I double-checked against the map. Everything matched.
Now I’ll show how it looked in Ads Manager.
We were doing a split test, so images were the same. The first:

The second:

The third:

The fourth:

The fifth:

The sixth:

There were six texts too. I’ll show them as screenshots — how they were assembled in the ad account.

First:

Second:

Third:

Fourth:

Fifth:

Sixth:

Traffic behaves in a way where people will end up visiting the Instagram account anyway. So Natasha, after reviewing the profile, asked Tanya to refine her highlight covers (pinned Stories) — keep them as loud as the feed, with those screaming colours.
The contrast was funny. On the website: kitschy mugs and Tanya’s work in a strict layout. On Instagram: an explosion at a macaroni factory made of colours.

We liked that solution. It matched Tanya’s identity. She felt at home in those environments, and that matters — the creator shouldn’t feel nailed to the solutions and forced to fight internal resistance just to exist inside the strategy.

I also asked Natasha to analyse Tanya’s closest competitors’ content. We did competitor analysis earlier. Then we sent Tanya a small “content work” spreadsheet, so she could swim on her own with something more systematic under her feet.

So.
The website was ready. The final version was almost four pages. The “Artist & Ceramicist” page appeared at the end, already after traffic started.

And the “Portfolio” tab contained two more pages:

The Home page assembled everything about Tanya: selling copy, work types with examples. A “Shop” section was placed under “Portfolio”, and there was a reason: a visitor who finds it is, in our model, curious and motivated. Because the page isn’t front-and-centre — you need intent to move your cursor around. Hover over “Portfolio” and you see two new points of interest.

Will a client go there?
That’s why I connected Yandex Metrica (Яндекс.Метрика — Russian analytics) with Webvisor (Вебвизор — session recordings) to Wfolio.

In the same Metrica I set up “Goals” for clicks to Instagram and to those pages, so we could compare Facebook Ads numbers to Metrica and Analytics.
I also registered Tanya’s site in Yandex Webmaster (Яндекс.Вебмастер — search console) and Google Search Console, so it would be indexed faster.

I wrote descriptions (meta descriptions) for every or almost every page using Tanya’s business keywords, but in a human way. It helps indexing. It’s a quick-time SEO starter kit.
And it worked. Some days the site started getting visits from search.

For example, on 15 November there were these visits:

On the “Shop” page I created a “custom conversion” in Meta Events Manager. The site structure was: Home as a pseudo landing page, and the “Shop” was easy but still hidden behind intent. Analytics tools help track those visits, and the “custom conversion” lets you optimise traffic towards that page.

At the end of each target page we placed buttons to social networks for ordering a custom piece or buying something ready-made.

One:

Two:

And finally, I asked Tanya to activate her followers who had already bought from her and leave reviews on the site.

Not everyone responded, but some did 🙂

Starting traffic

What matters before launching an A/B test.
In Facebook, split testing (another name for A/B testing) lets you test up to 5 objects in parallel.
The more you test, the more precise you get, but a day is the minimum. If you pick audience size correctly, a day is enough for early direction.

Audience size per object, from experience, should have at least a 1,000-people reach forecast at launch.
Some things settle, some things inflate. The ad auction is unstable. Today there are 10 advertisers, tomorrow 100, the day after tomorrow 3 — impression/click prices move. But we need the test. For the test, you need reach volume.
In my calculations, 1,000 is a minimal threshold so you don’t run on a tiny sample where extremes are inevitable and meaningless.
Two clicks on 300 people is a negative extreme. CTR will be 0.6–0.7% and it tells you nothing. Three clicks on 100 people is a positive extreme. CTR will be 3% and it also tells you nothing, even if it looks pretty.

CTR is click-through rate — how well you managed to pull a potential client into clicking and looking. And maybe they actually need it.

1,000 is the minimum adequate threshold for regression to the mean.

Expecting no regression to the mean is a common beginner mistake. You launch traffic, see clicks at 1.5 roubles and celebrate. But that’s an extreme low price at the start, and an extreme “cheapness” signal. The campaign has just entered the auction…

Don’t hype the client up about cheap clicks yet. The result will regress to the mean, and over a longer distance you’ll get a more realistic cost per click.
That’s how the pie is baked.

The first split test was simple. Same audience size — insanely broad. Across all of Russia, restricted only by interest targeting. And not with AND logic, but OR. The goal was to test creative strength on a mass audience. Also: the wider the audience, the cheaper the traffic. Convenient for testing.

We started with these interests. Essentially we were aiming at steps one and two of Hunt’s Ladder.

If you’ve already picked your top text, then now we’ll see who wins.

First split test — the texts

The best performer was the most personalised text.
Two other texts were close behind.
So what’s the decision?
Use all three in the next creatives. Meta allows up to five alternative primary texts inside one ad.
Not different ads in one ad set, but one ad set — one ad — with these three texts.

And then I tested everything on that bundle.

Second split test — interests + a bit of regions

We removed “trash reactions”, like the Mari El Republic (a small Russian region), where we got a single click. We kept only those with 5–6+ clicks so it looked like a tendency. And Natasha added a couple more differentiating interests to reduce selection bias.

I figured we could test two ideas if we set the direction and constraints clearly.
We also removed age groups that clicked very weakly: 45–54, 55–64, and 65+ were cut. Today it looks obvious. But ceramics isn’t always only for 18–24 or 25–34.

There was no clear winner, but I noticed how hard Moscow and Saint Petersburg clicked compared to the rest of Russia.

Look:

So the third split test was purely by regions.

BY THE WAY, during the second split test Tanya said people started buying and ordering.

Third split test — regions

“38 Cities” was under 1,000,000 people in Meta’s estimate, while Moscow and Saint Petersburg were 1,000,000+ in Meta’s estimate.

Regions performed well. Moscow and SPb performed as expected from previous tests.

It might look obvious now, but I’ll warn you.
When you split traffic into these ad sets, the algorithm starts operating inside the boundaries you define.
And the region stats in Google Analytics were similar, but they also showed interest potential in places like Novosibirsk, which was second after Moscow and SPb.

I could realistically increase clicks in the “regions” ad set because the algorithm there wouldn’t try to hunt for Moscow/SPb clickers.

By separating Moscow and SPb, I got the best performance of interest in Tanya’s products from “the rest”. The algorithm inside that box didn’t get distracted by the top cities.

And here comes the key chapter of the Odyssey…

Fourth split test — Lookalikes vs interest targeting

The site and its pixel-enabled pages accumulated visitors for LaL (Lookalike).
Lookalikes are audiences that are a few percent similar to those who visited the site.

The traffic got much narrower. So LaL came out more expensive than the very wide interest targeting across Moscow, SPb, and the whole country. But… it was worth it 🙂

I also built a Lookalike from those who messaged Tanya in DMs. Beyond friends and acquaintances, there were plenty of clients there too.

So we ended up with three LaL ad sets:
DMs, the “Shop” page, and the Home page (meaning: those who at least clicked and visited).

Funny outcome: LaL reached ~2.5× more people.

And it produced 2.6× more clicks. The link CTR difference was 0.80% — a strong trend towards a full percentage point. That’s meaningful.

Yet at forecast stage the system said: with your LaL, for 1,243 roubles 11 kopeks (RUB) you’ll reach about 1,100 people.

And for 543 roubles 11 kopeks (RUB) you’ll reach 1,100 people from this insanely wide interest targeting across 40 Russian cities, including Moscow and SPb.

But the LaL segment reached the optimistic threshold of 3.5K people for a price that was “promised” to only buy you 1,100 :))
Nice deal 🙂

Tanya asked the pot to stop cooking. We got 7 orders and 7 purchases.

Enough was sold out, enough was custom-ordered. Tanya said we already recouped the budget. The ad budget was 10,000 roubles (RUB).

Tanya asked to pour some traffic into the account so follower count grows for the future.
By the way, followers were growing during all A/B tests anyway.
Because people can click not only the link, but the account itself — to stare at it.

Even here: “link clicks” means website, and “clicks (all)” — subtract website clicks and you’ll get how many went to the account:

And it matters to say: throughout the tests I optimised for the “Shop page visit” conversion.
To hit two birds with one stone.

And I monitored how the “back to Instagram from the site” metric changed:

To make my traffic ideas sharper, closer to the end I started a Google Doc where, before each launch, I wrote my assumptions and then compared them to results — tracking how my judgement shifts.

This is to reduce cognitive bias. The brain can rewrite your judgement in hindsight, and then it feels like “I always thought that”. Hygiene here is a real part of the work: you adjust ideas and assumptions that you bake into a launch.

So, it seemed simple: switch optimisation from “Conversions” to “Traffic”, and in the link field paste the Instagram profile link. Yes, it opens the web version of Instagram, and Meta doesn’t recommend growing followers that way. Because the link opens in the phone browser, and the mystery is: is the person logged in to Instagram in that browser or not? Will the browser hand off into the app correctly or not?

My eyes popped out when the follower cost came out at 45 roubles.

So I launched like before, only with LaL. I poured traffic into the site again with “Conversion” optimisation.

Why did that happen? I think for two reasons:

Different optimisation goals lead the algorithm differently.
Landing on the site might impress people, and then they follow Tanya on Instagram.

I ran it for three days, here’s what I recorded in the journal:

First entry:

Second entry:

Third entry:

The entry dated 07.12.2021 says it all.

Stretching one strategy onto another goal isn’t a great idea.
Still, we got purchases again 🙂
And Tanya’s products cost on average 1,300–2,000 roubles (RUB).
So two buyers — and 4K roubles of traffic costs are back to zero.

These were the final LaL-to-site numbers at the end of the traffic work:

Totals in numbers:

— Reach: 16,235, Impressions: 18,953

— Total website clicks — 306,

— Average / best link CTR — 1.61% / 2.05%

— Total clicks: 923

— Average / best CTR (all clicks) — 4.87% / 6.14%

Tanya said I “paid off”, which means at least 12–13 customers came through during the ads.

From that, the lead cost is 769.23–833.33 roubles (RUB). But since my services were repaid already during the campaign, the real CPL was lower, because more leads happened.

And we remember LTV — a person who bought once often comes back for the second and third order.

What I left her with

Why did I test everything so carefully?
So that she would come back to me later. Before ending this iteration, I recorded a video where I showed how to duplicate the ad set with that bundled creative — same images and the three best texts. The LaLs are already built in. The only thing left is to adjust the bid, depending on how much money you want to spend.

And I scheduled a launch after the holidays so it would start on its own and she’d get more clients and more purchases.

Some content strategy notes, and a dedicated page on the site so people who are interested could read about her like a blog — case posts about her work. So they could get to know her as an artist:

I named it “Artist & Ceramicist”.

Natasha also drafted a few alt texts for images for better indexing of the Home page. Wfolio has a simple, clear guide for this.

And one of the most important things — the site will get organic traffic from search and curious people from social networks anyway. For example from the Instagram bio link.

Which means the Pixel will “catch” them, and later you can both retarget them with ads and build new Lookalike audiences.
At that time Google Analytics showed 4–5 people per day landing on her site.