Case Studies

Results, not adjectives.

Anyone can show you a screenshot. Here is the harder thing: the full story of how a brand actually gets raised, the broken parts we fix first, the system we build, and the numbers that follow once the machine is real.

+330%
Peak ROI delivered
12
Verticals served
9
Ad platforms run
3.1x
Best blended MER
The full story · raising a brand

How a kitchen-table candle brand became a real one.

Names changed, numbers real. This is the arc we run on almost every brand we take, told the long way, so you can see exactly what changes and why it works.

01 · The brand

A product people loved, and nothing else.

She made candles at her kitchen table. The product was genuinely good, so good that a single video went viral one weekend and she sold out. Then silence. Sales came in spikes, from a lucky post or a holiday, and vanished in between. She could not plan inventory, could not hire, could not breathe. A great product with no system underneath it is not a business. It is a streak.

02 · The diagnosis

The account was flying half-blind.

Before touching a single ad, we looked at the wiring, because a beautiful campaign built on broken plumbing just spends money faster. A full technical audit turned up four separate signal problems stacked on top of each other, the kind that are common enough you have probably got at least one running right now:

Browser-only pixel

After iOS 14.5+ and Safari's Intelligent Tracking Prevention, the Meta and TikTok pixels sitting in the browser were missing a real chunk of purchases, especially anything on Safari, in-app browsers, or ad blockers. The platforms bid on whatever signal survives, and this account was showing them maybe two thirds of the real picture.

No server-side backup

There was no Conversions API, no server-side tagging container, nothing catching the events the browser dropped. When a third-party cookie gets blocked or a pixel fails to fire, a browser-only setup has no fallback. The sale still happens. The platform just never hears about it.

Undeduplicated events

Where a server-side event did exist, it was not deduplicated against the browser event using a shared event ID, so some purchases were counted twice across the two paths, quietly inflating reported ROAS above what was actually happening.

Every platform grading its own homework

Meta's dashboard, Google's dashboard and Klaviyo's dashboard were each reporting revenue with generous, overlapping attribution windows, and the three numbers added up to more than her actual monthly revenue. Not fraud, just no single source of truth to reconcile against.

On top of all that, one good video was being reused until people were sick of it, creative fatigue, the quiet killer of paid performance that no amount of tracking fixes can solve on its own.

03 · What we built

Measurement first, then the machine.

We rebuilt the measurement server-side: a first-party server captures the real order webhook straight from Shopify the moment it happens, then sends a single deduplicated event, using the same event ID as the browser pixel, to Meta Conversions API and Google Enhanced Conversions. No more double counting, no more guessing which sale belongs to which click. We also wired in Consent Mode v2 so EU and UK visitors who decline cookies still get modelled correctly instead of vanishing from the data entirely, and put everything behind one blended dashboard so no single platform's self-reported number gets to be the final word again.

On top of the plumbing we added a creative engine (a steady flow of fresh, native-feeling content so nothing goes stale), and a multi-layer acquisition system (cold demand, retargeting and an email and SMS flow working together), all watched by the team and Nexus on that same dashboard every day.

4xevents recovered
vs the old pixel
1.8 to 3.1blended MER
in one quarter
2xrepeat purchase
rate
04 · The turn

The algorithm finally found her buyers.

This is the moment that matters. Once the platforms could see real conversions again, the same budget started finding better customers. Cost to acquire dropped without spending more, because the system was no longer fishing blind. The email and SMS flows turned first orders into second and third ones, which raised how much each customer is worth, which let us afford to bid higher than competitors who were still flying blind.

05 · Where it landed

From a streak to a machine she can run.

Spikes became a forecastable line. She could finally plan inventory, hire her first two people, and scale into Q4 with confidence instead of hope. The brand did not just grow, it became predictable, which is the thing every founder actually wants and almost no one sells. That is what we mean when we say we raise a brand.

The case studies

Written up properly. Situation, diagnosis, fix, result.

Every engagement below is a real one, anonymised. Shops, a SaaS, apps we built and took to market ourselves, even a freight brokerage that needed software more than it needed ads. Where we had the client's actual dashboard on tape, we left the camera rolling, press play on the panel. Where we did not, the numbers are exactly as the founders reported them. No brand names, on purpose, and no rounding in our favour.

Google Ads dashboard for a UK brand showing spend, return on ad spend and CPC
CASE 01Footwear · Google Ads · UKDashboard on tape

£151K of spend, and finally a straight answer on the return.

Imogen had a loyal UK following and a Google Ads account that was, generously, guessing. At that spend level, “we think it’s working” is an expensive sentence.

6.46×Conversion value / cost
£2.61Average cost per click
£973KSales tracked, window

Situation

Healthy demand and healthy spend, but conversion tracking patchy enough that nobody could state the real return with a straight face.

Diagnosis

Broken and partial conversion tags meant Smart Bidding was optimising toward noise, and budget was pooled on terms that only looked like they converted.

What we did

Rebuilt conversion tracking with Enhanced Conversions, then restructured the account around what was genuinely selling and cut what was quietly not.

Result

£6.46 back for every pound, CPC held at a lean £2.61. Imogen’s finance director asked to see the dashboard twice, the corporate equivalent of applause.

Google AdsEnhanced ConversionsAccount restructureSmart Bidding
Shopify overview dashboard showing returning customer rate and total orders
CASE 02Sleep & wellness · ShopifyDashboard on tape

A sleep brand that couldn’t rest on its own numbers.

Tasha built her brand on one idea: sleep is a skill, not an accident. The product converted. The account just couldn’t prove which channel was doing the work, so every scaling decision was a guess with money attached.

68.04%↑175%Returning customer rate
487↑185%Total orders, window
4×Events recovered vs old pixel

Situation

Strong product-market fit and a loyal early base, but a checkout that leaked and a Shopify account that could not attribute a single sale to the channel that earned it.

Diagnosis

A browser-only pixel gutted by iOS and Safari ITP, so Meta and Google were bidding on maybe two thirds of the real signal. No server-side fallback.

What we did

Rebuilt measurement server-side with deduplicated Conversions API events, then closed the funnel leaks so the algorithms could finally see real buyers.

Result

Returning-customer rate up 175% and orders up 185% in-window. Tasha’s face on the screen-share said the rest. Nobody expects a Tuesday to feel that good.

Server-side trackingMeta CAPIGoogle Enhanced ConversionsShopify CRORetention

Third agency in four years. First one that opened with “your tracking is broken” instead of a slide deck. Ninety days later the dashboard matches my bank account. That is the whole review.

Marcus · DTC apparel founder
Klaviyo business performance dashboard showing total and email-attributed revenue
CASE 03Supplements · KlaviyoDashboard on tape

The email account nobody had opened in two years.

Reuben had loyal buyers and a Klaviyo login collecting dust, three welcome emails from install day and then radio silence. The most profitable channel in e‑commerce was sitting switched off.

$998,191Total revenue, period
$321,807Email-attributed (32% of total)
84.95%Of email revenue from flows

Situation

A repeat-purchase product with real loyalty, and zero lifecycle marketing capturing it. Every new customer was being earned once and then left to drift.

Diagnosis

No welcome, no abandoned-cart, no post-purchase, no win-back. The list existed; nothing was talking to it. Revenue was being left on the table daily.

What we did

Built the full flow stack, welcome, cart, browse, post-purchase, win-back, on a real calendar instead of vibes, segmented by behaviour rather than blast-to-all.

Result

Email drove $321,807, just under a third of all revenue, and 85% of that came from flows that run themselves. Reuben says it’s the first time in two years he actually sleeps.

KlaviyoLifecycle flowsSegmentationEmail + SMSRetention
CASE 04B2B SaaS · inventory forecasting

The SaaS that was paying for signups from people who were never going to pay.

Anders built forecasting software that genuinely saves e‑commerce brands from over-ordering stock. His ads were “working”: signups every day. The bank account disagreed, because a free signup and a customer are two very different animals, and his whole ad stack was optimised for the wrong one.

23%from 9%Trial to paid
−41%Cost per qualified trial
+168%MRR, two quarters

Situation

Solid product, real retention once people paid, and an ad account that celebrated every free email address like it was revenue. Growth was flat while spend was not.

Diagnosis

Google was told a signup was the conversion, so it dutifully hunted tire-kickers. No CRM feedback loop meant the platform never learned which trials became money. The landing page listed features; nobody buys features at 11pm, they buy the end of a specific nightmare.

What we did

Wired offline conversion import from the CRM so qualified trials and closed-won deals flow back into Google as the real signal. Rewrote the page around the 3am stockout panic, not the feature list. Added a LinkedIn ABM layer for the 200 accounts that actually fit.

Result

Trial-to-paid went 9% to 23%, cost per qualified trial fell 41%, and MRR grew 168% in two quarters. Anders now opens the ads dashboard voluntarily, which he describes as new behaviour.

Offline conversion importGoogle AdsLinkedIn ABMLanding page rebuildCRM integration

Server-side events, deduplicated, fed back into the platforms. Blended MER went 1.8 to 3.1 in one quarter. I can produce a receipt for every number in that sentence, which is exactly the point of hiring them.

Priya · CMO, supplements
CASE 05Mobile app · fitness · iOS + Android

The fitness app that was famous for being downloaded and deleted.

Mateo’s coaching app had glowing reviews from the people who stayed. The problem was the other 89%: installed on Sunday night with the best intentions, hit a paywall before they’d done a single workout, gone by Tuesday. Cheap installs, expensive subscribers.

−57%Cost per trial start
26%from 11%Install to trial
<90 daysSubscriber payback

Situation

A genuinely good program, a 4.8-star rating, and paid campaigns judged on cost per install, the vanity metric of app marketing. Installs were up and to the right. Subscribers were a rumour.

Diagnosis

Campaigns optimised for installs, not trials. SKAdNetwork was misconfigured, so iOS data arrived as mush and the algorithms flew blind on half the market. And the paywall greeted users at second zero, before the app had proven it was worth a coffee a month.

What we did

Remapped events so trial start, not install, is what every platform optimises toward, fixed SKAN so iOS finally reported honestly, moved the paywall to after the first completed workout, and took over Apple Search Ads on high-intent terms competitors were ignoring.

Result

Cost per trial down 57%, install-to-trial up 11% to 26%, and each subscriber pays back in under 90 days. Mateo’s words on the call: “so we were advertising the door, not the gym.” Exactly that.

Apple Search AdsSKAdNetworkEvent mappingPaywall placementMeta app campaigns
CASE 06Custom software · logistics · 40-person team

The company that ran on five tools, nine spreadsheets and one very tired woman named in every escalation.

Rhea’s freight brokerage was profitable and quietly drowning. Every quote meant swivel-chairing between a CRM, two carrier portals, email and a rate sheet last updated “recently.” The tool she needed did not exist. So we built it.

31 hrsManual work removed, weekly
−92%Quoting error rate
12 minQuote turnaround, from 4 hours

Situation

Growing demand, good margins, and an operation held together by copy-paste. Every new customer made the problem 3% worse. Hiring more people to do the copy-pasting was the only plan on the table.

Diagnosis

This was never a marketing problem. The bottleneck was quote turnaround: four hours on average, during which the customer had usually already said yes to someone faster. The data existed; it just lived in five places that refused to speak.

What we did

Built a custom quoting engine that pulls carrier rates, margins and history into one screen, then put an AI agent on the inbox that reads inbound requests and drafts the quote before a human even opens the email. Humans approve; they no longer assemble.

Result

Quotes out in 12 minutes, errors down 92%, and 31 hours a week of manual work simply gone. Win rate followed speed, which it always does. Rhea took her first two-week holiday in six years and the system did not notice.

Custom buildAI agentSystems integrationOps automationDashboards
CASE 07Carbiu · app development · go-to-market

Carbiu: the app we built, shipped, and priced until free users started paying.

Carbiu came to us as a validated idea and a waiting list, with no engineering team and no launch plan. We do not just market apps, we build them, so we did both: the product, the free tier, and the road to market, designed as one system instead of three handoffs.

28%Free to paid conversion
~7×The freemium industry average
iOS + AndroidBuilt and shipped by Kavalsia

Situation

Real demand, a waiting list that kept growing, and nothing to install. Every month without a product was a month the audience cooled. Speed mattered as much as quality.

Diagnosis

Most freemium apps fail before launch, at the whiteboard: they give away too much, so the free tier becomes charity instead of marketing. We defined the activation moment and the upgrade wall before designing a single screen.

What we did

Built the app for iOS and Android, shaped onboarding so users hit the value moment in the first session, gated the features people actually pay for, then ran the go-to-market: store optimisation, launch campaigns and lifecycle messaging that nudges at the right moment, not on a timer.

Result

28% of free users upgraded to paid. For context, freemium products typically convert 2 to 5%, and the celebrated ones touch 10. When the product, the pricing and the marketing are designed together, the funnel stops leaking between departments.

App developmentGo-to-marketOnboarding designPaywall strategyASOLifecycle messaging

I make candles at my kitchen table. The first month we hit a number I used to only dream about, I actually cried. They treated my tiny brand like it mattered, and they explained everything so I understood it. My husband knows what a conversion API is now. Against his will.

Elena · home fragrance, solo founder
More on the record

Same discipline, numbers reported by the founders.

Clothing rail in a minimal fashion boutique
+330%ROI
Fashion · DTC apparel

The spreadsheet that finally agreed with the bank.

The problem

Naomi had the aesthetic nailed and her numbers scattered across four dashboards that never agreed with each other, so no decision felt safe.

What changed

Rebuilt tracking for clean data, then ran Google, Meta and TikTok as one engine. Reported ROAS and real revenue finally told the same story, “genuinely emotional, for a spreadsheet.”

Server-side trackingGoogleMetaTikTokStackAdapt

I run my own numbers and I tried to catch them out. Could not do it. Every claim reconciled to the cent. They are the only agency I have kept past ninety days, and I am not an easy client.

Dan · operator, multi-brand
Dog waiting at a pet supply store counter
+315%ROI
Pet supplies · e-commerce

The website that was quietly losing him customers.

The problem

Owen’s store was held together “by tape and hope.” The checkout was clunky enough that people added to cart and then simply left.

What changed

Rebuilt the storefront and layered a five-platform paid mix on top. Conversion rate and new-customer flow lifted, and the store stopped leaking traffic it already paid for.

Shopify CROGoogleTaboolaMetaPinterest
Makeup and cosmetics arranged flat on a table
+245%ROI
Beauty · cosmetics

Popular with everyone who already knew her.

The problem

Farrah’s brand was loved by people who already found it. Reaching everyone else was the wall she kept hitting.

What changed

CRO work plus Google, TikTok and Meta reached a younger buyer, fed by a steady drip of fresh creative so nothing fatigued. Purchase rate climbed, audience widened.

CROGoogleTikTokMetaCreative engine
Ceramic homeware arranged on a shelf
+213%ROI
Home goods · e-commerce

Marketing spread so thin it was basically confetti.

The problem

Good products, budget scattered across every platform at once, so nothing had enough weight to actually work.

What changed

Tightened into one Google, Meta and Taboola mix over a Shopify tune-up, with retention lifting lifetime value behind the scenes. “I used to dread the ad-spend line. Now I read it for fun.”

GoogleMetaShopifyTaboolaEmail + SMS

Onboarding took nine days and they answered Slack faster than my own team, which was mildly embarrassing for my own team. They refused to celebrate signups until they were qualified trials, and our board deck now has one number on it, the right one. Docking half a point because they deleted my favourite dashboard. It was lying to me, apparently.

Lena · co-founder, B2B SaaS
Minimal living room interior with soft light
+118%ROI
Home decor · e-commerce

The smallest number here, the one she trusts most.

The problem

Ingrid’s brand was the slow, built-to-last type, so “blast ads and hope” was never going to fit the way she sells.

What changed

Rebuilt the site and treated email and SMS as a real channel, not an afterthought. Click-through rose, bounce fell, and one-time buyers slowly became a base, earned rather than bought.

Web rebuildEmail + SMSRetention
Shopify analytics reel across nineteen case studies
The full reel · 19 dashboards on tape

Rather see all of them back to back?

One scroll, nineteen accounts

Real Shopify dashboards, unedited, stitched end to end. The frame you land on alone holds $1,128,559 in sales, a 69.98% returning-customer rate and a $509.41 average order value, up 31%. Same playbook, run nineteen times. Press play on the banner.

Every other agency promised installs. These people asked what a subscriber is worth on day ninety and rebuilt everything backwards from that answer. Payback went under three months. I have stopped checking my phone at dinner, my girlfriend says thank you.

Jules · founder, fitness app
Why it works every time

It is a method, not a miracle.

01

See the truth

Rebuild measurement so every real sale is captured and fed back. You cannot optimise what you cannot see.

02

Feed the machine

A steady flow of native creative, the real lever behind paid performance at scale.

03

Compound it

Cold demand, retargeting and retention working as one, so each part lifts the others.

04

Make it predictable

Scale on the numbers that matter until spikes become a line you can forecast.

Your brand next

Let us write the next case study together.

Start with a free audit of where you stand, then a call to map the system that raises your brand.

Book a strategy call → Get a free audit →