Case Study: A Consumer Electronics Reseller Cleaned Up 15,000 Messy Attributes in 72 Hours

A consumer electronics reseller had onboarded six suppliers over 14 months, growing their catalog to 4,200 products covering laptops, monitors, peripherals, networking hardware, and audio equipment. After the sixth supplier integration, their head of catalog identified the problem that had been compounding since the second supplier: product attribute cleanup had never happened. The catalog contained 15,000 attribute instances with duplicate names, inconsistent value formats, and broken faceted navigation that was causing serious usability damage on their Shopify storefront.

The Challenge

When a consumer electronics catalog grows through multiple supplier integrations without a centralization step, the attribute data is the first casualty. Each supplier uses their own naming conventions, their own value formats, and their own assumptions about which specifications matter. Import those feeds directly into a storefront without a normalization layer, and the resulting attribute set is not a catalog — it is an aggregation of six catalogs that have never spoken to each other.

The reseller’s specific situation, documented by an internal audit:

Attribute name duplication: The same physical property had been imported under different names from different suppliers. RAM capacity appeared in the catalog as “RAM,” “Memory,” “RAM Memory,” “Installed RAM,” and “Total Memory” — all referring to the same specification, all existing as separate attribute fields. Storage capacity appeared as “Storage,” “HDD,” “SSD Capacity,” “Drive Size,” and “Hard Drive.” Across all attribute types, the audit identified 94 cases of the same property represented under two or more different attribute names.

Value format inconsistency: Even where attribute names were the same, value formats differed. Screen resolution was stored as “1920x1080,” “1920 x 1080,” “1920 × 1080,” “Full HD,” “FHD,” and “1080p” — six different representations of the same resolution. Processor speed appeared as “2.4GHz,” “2.4 GHz,” “2400MHz,” and “2.4” (no unit). USB port counts as “2,” “2x USB,” “USB x2,” and “Two.”

Missing attribute values: 31% of products had one or more key attributes that were blank — not because the specification was unavailable, but because the supplier had put the value in a different field or column name that had not been mapped on import.

The practical consequences showed up directly in the storefront. The faceted navigation on Shopify — the filter sidebar that lets customers narrow by RAM, storage, screen size, processor, and connectivity — was effectively broken. Filtering by “16GB RAM” returned 34 products. But the actual number of products with 16GB RAM in the catalog was 127; the other 93 had the same specification stored under “Memory,” “Installed RAM,” or “RAM Memory.” A customer filtering for 16GB laptops was seeing 73% of the relevant products excluded from results because of attribute name inconsistency.

The SEO impact was secondary but present: structured data markup for product pages was pulling from malformed attribute fields, and Google’s product knowledge panel was surfacing incorrect or incomplete specifications for several high-traffic products.

[SCREENSHOT: Shopify faceted navigation sidebar showing the attribute filter chaos — six separate “RAM” filter options each with different labels and partial product counts, instead of a single consolidated filter]

What They Tried First

The catalog head’s first attempt at product attribute cleanup was manual: open each duplicate attribute, identify which products used it, reassign those products to the correct canonical attribute, then delete the duplicate. This is the correct process in theory. In practice, working through 94 duplicate attribute cases while also managing four attributes per product that might need reassignment produced a combinatorially complex manual operation that consumed 18 hours before the team had addressed 11 of the 94 duplicate pairs — and introduced three new inconsistencies through manual error during reassignment.

The second approach was to use Shopify’s built-in product edit capabilities. Shopify’s admin does not have bulk attribute rename or merge tools. You can edit attributes on individual products or export a CSV and re-import with corrections, but there is no native mechanism to say “rename attribute X to attribute Y across all products where X is present, then merge with the existing Y values.” The CSV export-edit-import cycle for 4,200 products with 15,000 attribute instances, working in Excel, produced a file management problem larger than the original attribute problem.

The team also contacted two of their suppliers to request normalized re-deliveries of their product feeds. Both suppliers were cooperative but slow: the corrected feeds arrived 3 and 5 weeks later respectively. Corrected supplier data helps for future imports but does not address the existing catalog, and it does nothing for the four suppliers who were not contacted or who declined to change their feed format.

The Solution

The team migrated their catalog into MicroPIM and executed a structured 72-hour product attribute cleanup sprint. The sprint had three phases: attribute consolidation, value normalization, and blank attribute resolution.

Phase 1: Attribute Consolidation (Day 1, 14 hours)

MicroPIM’s attribute management interface shows all attributes in the catalog with their product counts. The team exported this list — 312 distinct attribute names across 4,200 products — and identified every duplicate group: the 94 cases where multiple attribute names represented the same property.

For each duplicate group, they designated a canonical attribute name following a consistent naming convention (Title Case, no abbreviations, unit in parentheses where applicable — e.g., “RAM Capacity (GB),” “Screen Resolution,” “Storage Type”). MicroPIM’s bulk attribute merge tool allowed them to select the canonical attribute and all duplicates to be merged into it, then execute the merge. The system reassigned all products from the duplicate attributes to the canonical attribute and deleted the empty duplicates.

Running the merges for all 94 duplicate groups took approximately 6 hours. The catalog went from 312 distinct attribute names to 218. The faceted navigation filter count dropped from a chaotic 312 possible filter dimensions to 218 well-defined ones, with accurate product counts per filter value.

[SCREENSHOT: MicroPIM attribute merge interface, showing the “RAM Capacity (GB)” canonical attribute selected on the left and five duplicate attributes — “RAM,” “Memory,” “RAM Memory,” “Installed RAM,” “Total Memory” — queued for merge on the right, with combined product count shown]

Phase 2: Value Normalization (Day 2, 16 hours)

With attribute names consolidated, the team addressed value format inconsistency within each attribute. For the highest-impact attributes (screen resolution, processor speed, storage capacity, RAM capacity, connectivity ports), they built normalization rules:

Screen resolution: A lookup table mapping all variant representations to a single canonical format (“1920x1080”). The table covered 34 input formats across all resolution values in the catalog.

Processor speed: A transformation rule converting MHz values to GHz (divide by 1000), stripping “GHz” from values that included the unit, and standardizing to one decimal place.

Storage capacity: Standardization to GB as the unit (converting TB values by multiplying by 1,000), with consistent formatting (e.g., “512 GB,” “1,000 GB” rather than “1TB” or “1000gb”).

USB port count: Converting all text representations (“Two,” “2x USB,” “USB x2”) to a plain integer.

MicroPIM’s bulk value replace tool applied these normalization rules across all products in each attribute simultaneously. For the screen resolution normalization alone, 34 input format variants were mapped to canonical outputs and applied across 1,847 products in 4 minutes.

After Phase 2, the “1920x1080” filter in faceted navigation returned 312 products, compared to 41 before normalization — because the same products were now consistently represented.

Phase 3: Blank Attribute Resolution (Day 3, 10 hours)

The third phase addressed the 31% of products with missing values in key attributes. MicroPIM’s bulk edit tools allowed the team to:

• Filter products by attribute completion status (products where “RAM Capacity (GB)” is empty)
• Cross-reference available data in related fields to derive the missing value (e.g., products where RAM was mentioned in the product name or description field but not extracted into the attribute)
• Apply bulk updates to groups of products where the missing value was derivable from other data in the record

For products where the missing value was not derivable — genuinely absent from all available data — the team created a priority list ranked by product traffic. The top 200 products by monthly session count with missing critical attributes were assigned for manual research. The remaining 1,300 products with missing non-critical attributes were flagged for resolution in the next supplier re-import cycle.

[SCREENSHOT: MicroPIM blank attribute report showing 4 key attributes with product counts missing values, sorted by estimated traffic impact, with bulk edit action buttons for each attribute group]

The Results

The 72-hour cleanup sprint covered the core catalog with the following verified outcomes at end of Day 3:

• Attribute names reduced: From 312 to 218. The 94 duplicate attribute groups were fully merged.
• Faceted navigation accuracy: A post-cleanup audit of 20 randomly sampled filter values showed average accuracy of 96.3% — products appearing in filter results that should appear. Pre-cleanup, the same sample averaged 41.2% accuracy.
• Value format consistency across key attributes: 99.1% of products with a value in each key attribute now had the value in the canonical format.
• Blank attribute rate for key attributes: Reduced from 31% to 14%. The remaining 14% were genuinely data-absent and flagged for supplier re-request.

The storefront impact was measurable within two weeks:

• Faceted navigation engagement: Sessions that used at least one filter increased from 18% to 34% of total storefront sessions.
• Filter-assisted conversion rate: Products purchased by sessions that used faceted navigation converted at 3.8%, compared to 2.1% for non-filter sessions. (The pre-cleanup equivalent metric was not tracked, but filter usage was so low that the comparison would not have been meaningful.)
• Organic structured data: Google Search Console showed improvement in product rich result eligibility within 30 days of the cleanup. Previously, 340 products had structured data errors in Google’s index. Post-cleanup: 28.

Key Takeaways

• Attribute cleanup at scale requires bulk merge and normalization tools. Manual product-by-product correction at 15,000 attribute instances is not feasible within a reasonable timeframe.
• Attribute name duplication is the first problem to solve. Until duplicates are merged, value normalization work is ineffective because the same product can have the right value in the wrong attribute.
• Value format inconsistency directly breaks faceted navigation. A filter that returns 27% of the products that should match is not a filter — it is a misleading UI element that erodes customer trust.
• Blank attributes should be triaged by traffic impact. Not all missing values are equal; focus first on the attributes that affect filter navigation and structured data for your highest-traffic products.
• The cleanup is a one-time project, but preventing recurrence requires import-time attribute normalization rules. The cleanup creates the standard; normalization rules on new imports maintain it.

Broken faceted navigation is a conversion problem with a data root cause. If your catalog has grown through multiple supplier integrations without a product attribute cleanup step, the damage is accumulating with every new import. Start the cleanup process at app.micropim.net/register — the attribute audit takes less than an hour to run, and it will show you exactly where the consolidation work needs to begin.

Related Reading

• Product Attributes and Custom Fields
• Attributes Builder
• Case Study: Baby Products Standardized 25,000 Attributes

Frequently Asked Questions

When MicroPIM merges duplicate attributes, does it overwrite existing values on products that have values in both the duplicate and the canonical attribute?

No. The merge process follows a defined precedence rule you set: either the canonical attribute value takes precedence, or the duplicate attribute value takes precedence, or you are prompted to review conflicts. For the electronics reseller’s cleanup, they set canonical-attribute precedence because the canonical attribute had been populated first and was more likely to contain correct data. Of the 94 merge operations, 312 products had values in both attributes — those conflicts were resolved using the precedence rule without manual review.

How does MicroPIM prevent the same attribute inconsistencies from reappearing when new supplier feeds are imported?

Configure attribute normalization rules at the import level. When a new feed arrives, incoming attributes are mapped to your canonical structure and values are transformed to your standard format before reaching the catalog. Products with unmappable values go to a review queue rather than entering the catalog as raw data.

Our Shopify store has 94 different filter options in our sidebar right now. Is that all coming from attribute duplication?

Not necessarily — it depends on whether all 94 represent duplicates or legitimate distinct attributes. Run a product attribute report that shows all attributes with their product counts. Attributes with low product counts (5, 10, 20 products when you would expect hundreds) are almost always duplicates or legacy attributes from a previous import. In a typical multi-supplier electronics catalog, 30 to 50% of apparent attributes are duplicates or orphans that can be cleaned up.

Can the attribute cleanup be done without migrating to MicroPIM as the primary catalog system?

The cleanup tools exist within MicroPIM, so the working process is: import your existing catalog into MicroPIM, perform the cleanup there, then export the cleaned catalog back to your storefront. Some businesses use MicroPIM as a catalog staging and cleaning layer without fully replacing their storefront’s native product management — often the quickest path to a clean catalog without a full platform migration.