A D2C women's fashion brand with strong product-market fit but a Google Ads account built for vanity metrics. Here's how we rebuilt it around actual margin — and what happened to their bank balance when we did.
Likely is a women's fashion brand selling occasion dresses, tops, and coordinated sets. Strong product, strong aesthetic, a loyal customer base building organically. But their paid acquisition was a different story.
When we first looked at the account, it appeared to be performing well. The Google Ads dashboard showed a blended ROAS above 4. Campaigns were active. Budget was being spent. On paper, everything was working.
The problem: that number was a lie — not because it was fabricated, but because it was incomplete. It counted revenue. It didn't count what it cost to deliver that revenue.
Dashboard ROAS: 4.1× — looked profitable. But when we mapped actual fulfilment costs, return rates (which in fashion run significantly higher than other categories), payment gateway fees, and ad spend against gross revenue, the real picture was very different. The business was growing its top line while quietly compressing its bottom line.
The first thing we did wasn't optimisation. It was research. Before changing a single bid or headline, we needed to understand who was actually buying, why, and what search behaviour looked like in their market.
We identified two distinct buyer profiles — occasion shoppers (event-driven, high intent, shorter purchase cycle) and wardrobe-builders (browsing, aspirational, longer cycle). Each needed a different campaign type and creative brief.
We mapped the actual search terms converting to purchases vs those generating traffic. A significant portion of budget was going to informational and comparison queries that weren't converting — they were just racking up impressions.
We ran a full competitor ad audit. Brands in the same space were bidding heavily on generic fashion terms — "women's dresses," "occasion wear" — with low intent and high competition. We identified gaps: specific occasion types and dress styles with high purchase intent and lower CPC.
Fashion returns are brutal on ROAS. We analysed return rates by product category and found that certain dress styles were driving high AOV but also high returns, effectively making them negative-margin acquisition vehicles.
The previous agency was optimising for revenue volume. We needed to optimise for revenue quality — transactions that stayed revenue after returns were processed, at a cost that left actual margin intact.
Once we understood the audience and the margin profile, we rebuilt the campaign structure from scratch. The old account had one broad Performance Max campaign doing everything — brand, non-brand, shopping, display, all mixed together. This gave Google's algorithm conflicting signals and made attribution impossible.
We separated brand keywords into their own Search campaign with manual CPC. This stopped Performance Max from cannibalising brand traffic (which converts at a much higher rate) and inflating the apparent ROAS of the prospecting campaigns. Brand ROAS and non-brand ROAS were now measured separately.
We built a Search campaign targeting high-intent, occasion-specific queries: wedding guest dress, cocktail dress Australia, formal occasion wear, and similar. Phrase and exact match only. Extensive negative keyword list built from 3 months of search term data to exclude broad, non-converting traffic.
We launched PMax only after the Search campaign had generated 30 days of clean conversion data. This gave the algorithm real purchase signals to optimise against — not guesses. Asset groups were segmented by product category and audience intent, not dumped into one catch-all group.
We worked with the client to tag products by margin tier. High-margin products got higher target ROAS targets and more budget. Low-margin or high-return products were excluded or bid down. The shopping campaign now optimised for profit, not revenue volume.
The existing tracking was firing duplicate conversions. Purchase events were being counted twice in some sessions, which meant the account's historical ROAS data was structurally overstated. We rebuilt GA4 and Google Ads conversion tracking, cross-verified with backend order data, and recalibrated bidding targets based on accurate numbers.
This is the part most agencies skip — because it's uncomfortable. A 4× ROAS sounds great until you map every cost that sits between gross revenue and actual cash in the bank. For an e-commerce fashion brand, that list is long.
Every number below represents a real cost category that a dashboard ROAS figure ignores entirely.
| Cost Category | What It Is | Impact on Margin |
|---|---|---|
| Ad Spend | What Google charges per click | Directly reduces margin |
| Return Rate (~18–22%) | Fashion returns are significantly higher than other categories. Revenue is reversed but fulfilment costs are not. | Reduces effective revenue by 18–22% |
| Outbound Shipping | Cost to fulfil each order, whether subsidised or not | Fixed cost per transaction |
| Return Shipping + Processing | Reverse logistics, inspection, restocking or write-off | Often hidden in ops costs |
| Payment Gateway Fees | Typically 1.5–2.5% of transaction value | Scales directly with revenue |
| COGS (Product Cost) | Cost of manufacturing or wholesale | The largest margin lever |
| True Blended ROAS Target | ROAS needed to actually be profitable after all of the above | For this brand: minimum 2.8× to break even on paid |
Once we mapped the full cost stack, we recalculated what ROAS the brand actually needed to be profitable — not just to show positive numbers on a dashboard. The answer was a minimum of 2.8× true blended ROAS. The previous campaigns were hitting 4.1× on the dashboard but barely 1.9× in true terms after returns and fulfilment costs.
We reset all bidding targets based on the true margin model. Some campaigns that looked profitable were paused. Some that looked marginal were actually fine once returns were excluded. Budget was reallocated toward high-margin product categories and high-intent search terms. The dashboard ROAS dropped slightly. Actual profit increased significantly.
After 90 days of running the rebuilt account, with optimisation cycles every week and a monthly strategy review against actual business metrics, here's where the brand landed.
The dashboard ROAS dropped from 4.1× to 3.4× in the first 30 days after rebuild. This is expected — we removed campaigns that were generating cheap, high-return purchases that looked good on paper. The client understood this because we explained the true margin model upfront. By day 60, revenue had recovered. By day 90, it had grown — at a margin that was actually reflected in their bank balance.
Every e-commerce brand has a version of this problem. The platforms are incentivised to show you the best possible attribution. Your agency is often incentivised to show you growth in the metrics that are easiest to grow. Neither of them has direct visibility into your fulfilment costs, your return rates, or your true unit economics.
The only way to know if paid acquisition is actually working is to build a margin model first — before you set a single ROAS target, before you scale a single campaign. Everything else is just optimising for the wrong number faster.
This is why every engagement we take on starts with a tracking and margin audit before we touch a single campaign.
We'll review your account, map your true margin model, and tell you honestly what your paid acquisition is actually returning — before you pay us anything.
Get a Free Audit →