LOADING CASE STUDY

Cleaning the Shelves:
Architecture for an FMCG Portfolio on Marketplaces

A multi-brand portfolio across home care, shaving & grooming and oral care looked “big” on paper, but behaved like a cluttered storage room in the algorithm. This is how we turned slow, confusing shelves into a 21-day rotation standard and pushed conversion from 0.03% to 1.2%.

01 // THE MESS

Not Underperforming Ads. Underperforming Portfolio.

This was not “one brand” work. It was a distribution portfolio: multiple global FMCG names in home care, grooming and oral care routed through marketplaces. On the outside it looked strong: wide set of SKUs, coverage across price tiers, presence in key categories.

Underneath, the behavior was wrong:

• low CTR on many items, even with acceptable pricing and assets;
• weak conversion and high bounce from product pages;
• rotation cycles that broke marketplace logic instead of playing with it.

"Traffic wasn’t falling into a funnel. It was spilling over a broken shelf."

When you plug a “retail-era” portfolio into a marketplace, you don’t get digital growth. You get amplified entropy. My job was not to “run better campaigns”; it was to rebuild the spine of the portfolio so traffic had something healthy to flow through.

02 // SEEING THE ORGANISM

ABC/XYZ as a Map of What Deserves to Live

ABC/XYZ for me is not an Excel ritual, it’s a way to see what kind of organism you’ve actually built. We plotted the portfolio along two dimensions:

ABC – contribution to revenue and margin;
XYZ – stability and predictability of demand.

FIG 1.0 // ABC / XYZ PORTFOLIO GRID
Low → High Revenue Contribution (A–C) Low → High Stability of Demand (X–Z)
C–X // Kill
Low impact, unstable. Even after fixing basics these SKUs refused to behave. Removing them decreased noise for algorithms and shoppers.
B–X // Fix or Bundle
Potential, but fragile. Here we tested: better content, pricing logic, and localization. Survivors became parts of profitable bundles.
A–X // Anchor
High contribution, volatile demand. Protected and supported via traffic and promo logic so they could mature into stable anchors.
C–Y // Diagnose
Mid-range behavior. Required classic “give them a chance” protocol: content repair, logistics and geo-check before final decision.
B–Y // Core
Solid but not explosive. Became operational backbone: predictable, easy to plan, good candidates for bundles and cross-sell.
A–Y // Growth Tier
Strong plus predictable. First in line for additional exposure, reviews, and shelf logic.
C–Z // Retire
Relics. No acceptable path to growth even with technical fixes. Keeping them hurt the whole organism.
B–Z // Quiet Winners
Not loud but efficient. Often under-marketed workhorses. We amplified them instead of inventing “new heroes”.
A–Z // Strategic Core
This is what people actually came for. These SKUs defined the economic reality of the portfolio and received the cleanest user journeys.

Before killing anything, we respected a simple rule: “Fix what you control first.” Content, price architecture, logistics, category mapping. Only then do you decide if the problem is demand itself.

03 // THE REPAIR PROTOCOL

Don’t Blame the Algorithm if the SKU Is Wrong

Each underperforming product went through a fixed sequence. This wasn’t about “feelings”; it was about a protocol that could be scaled to hundreds of items.

FIG 2.0 // REPAIR PROTOCOL
Technical Clean-Up
Audit content on all levels of the page, price ladders, logistics and geo-availability. Ensure the product is actually visible where demand exists.
Behavior Check
Re-evaluate signals: CTR, add-to-cart, bounce, checkout drop-offs. If behavior improves, keep and support. If not, the issue is real demand, not “bad luck”.
Cut or Rebuild
Truly weak SKUs get retired. Survivors are reassembled into stronger bundles so that one purchase moves multiple units and improves rotation.

We weren’t just chasing “best sellers”. We were protecting the marketplace’s perception of the portfolio as a whole. If you keep feeding the system weak offers, it learns to distrust you.

04 // FUNNEL & ROTATION

Making the Funnel Honest — and Fast

The second layer was analytics. In Power BI we built a simple, brutal thing: a view where each category and SKU could be traced from impression to sell-through with no room for excuses.

Early on, the conversion sat at ~0.03% — the number you get when traffic hits a wall of confusing, unstructured offers. After repair, we moved it to ~1.2% and hit a 21-day rotation standard that marketplaces treat as “golden”.

FIG 3.0 // FUNNEL & 21-DAY ROTATION
Impressions Top of Traffic
Clicks CTR Up After Repair
Add to Cart Bundles + Content
Purchases Conv. ≈ 1.2%
21 days
Average Sell-Through Cycle
This wasn’t an arbitrary KPI. Marketplaces implicitly reward portfolios that behave like this: stock moves fast, SKUs don’t rot, and algorithms learn that “when we send people here, things actually sell.”
The point wasn’t just growth. It was predictability — the ability to plan inventory, marketing and cashflow on a spine that doesn’t break.

Once the portfolio started behaving like a system instead of a pile of SKUs, advertising finally made sense: every impression had a higher probability of ending in real movement of goods, not a cosmetic lift in clicks.

// FINAL EXECUTION LOG STATUS: PORTFOLIO SPINE REPAIRED
Portfolio Disassembly COMPLETE

Full ABC/XYZ mapping of home care, shaving & grooming and oral care SKUs, showing which products deserved traffic, which deserved repair, and which had to be retired.

Repair Protocol DEPLOYED

A repeatable, SKU-level process: technical clean-up → behavior check → cut or rebuild into bundles. Scaled across the portfolio without losing control.

Funnel Telemetry INSTALLED

Power BI dashboards connecting impressions, clicks, add-to-carts and purchases with rotation speed, so everyone saw the same economic reality, not isolated metrics.

// IMPACT TELEMETRY DATA SOURCE: 6-MONTH WINDOW
STREAM: SALES
+40%
Portfolio Revenue Lift
Achieved not by dumping more media into the system, but by making each unit of traffic land on a portfolio that was capable of converting.
STREAM: CONVERSION
→ 1.2%
From ~0.03% Baseline
Content repair, SKU pruning and bundle logic turned a noisy shelf into a clean route from impression to purchase.
STREAM: ROTATION
21 days
Sell-Through Cycle
Hitting the marketplace “gold standard” for rotation, improving stock planning and strengthening the platform’s trust in the portfolio’s ability to move inventory.
← Return to Experience Logs