The Art of Restocking

What is a replenishment?

At its core, a replenishment is a bulk shipment of products from a brand to Ankorstore's warehouse.

Before joining the fulfillment service, brands would handle each retail order individually - picking products, packing boxes, and shipping them one by one. This meant hours spent on logistics rather than growing their business.

With Ankorstore's fulfillment service, brands send us their inventory in bulk, typically once a month. When retailers place orders, our warehouse team handles everything: picking the products, packing the boxes, and shipping them to retailers.

The journey of replenishment declaration

When I started at Ankorstore, I joined what many called 'the startup within the startup': a lean team of engineers who had accomplished something remarkable. In just one month, led by one of the company's founders, they had built the entire fulfilment service software infrastructure from scratch as a proof of concept.

By the time I joined, this proof of concept (PoC) was already serving several dozens of brands. However, the self-service experience was, to put it mildly, very basic. Users had to navigate complex inventory decisions with nothing but a bare-bones form to help them.

From this starting point, we began a journey of continuous improvement. Each step was shaped by our users' needs, feedback, and pain points, as we transformed that basic form into a powerful stock management tool.

This case study will take you through this journey:

1. From PoC to MVP - How we built the foundations
2. The corners project - How we refined it
3. What's next - Our exploration into predictive inventory management

1. From PoC to MVP

Our Proof of Concept revealed an immediate pain point: submit a form, and your data vanished. Brands managing hundreds of products (SKUs) were forced to double-track everything - once in their spreadsheets, once in our system. A classic case of software creating more work, not less.

Usability testing

I built a high-fidelity prototype of an enhanced replenishment experience and tested it with five fulfillment brands. While the improved interface was well-received, our user interviews uncovered a deeper issue: a fundamental trust gap in our system.

The problem went beyond usability. Our inventory status calculations were too simplistic compared to brands' sophisticated forecasting needs. As a result, they built their own prediction models in spreadsheets and relied heavily on Ankorstore's Inventory Managers for guidance. This meant that our employees would have to support every single user through lengthy consultations on stock forecasting. This is simply not scalable.

The message was clear: no amount of UI polish would matter until we addressed the underlying data intelligence gap.

Three Key Features

1. The Missing Puzzle Piece: Replenishment Details

Imagine sending your merchandise to a new warehouse without being able to compare what you planned to send against what actually arrived. That was our user's reality. We fixed this blindspot with a dedicated replenishment details page, showing both declared quantities and warehouse receipts.

The impact was immediate: it became our third most-visited fulfillment page, revealing just how crucial this basic functionality had been missing.

2. At the Right Time and Place: Contextual Stock Information

We enhanced the declaration form with crucial data points: current storage levels and a traffic-light system of tags that visually flags items at risk. Although the users didn’t trust our traffic-light system to be accurate to their business requirements , they seemed to welcome it nevertheless.

3. Two Steps Instead of One

A simple review step transformed our replenishment accuracy. Instead of asking brands to submit quantities blind, we now show them a focused view of just their declared items, so they can double-check that everything is as expected.

The result? Brands have caught and corrected mistakes in over 10% of all replenishments before they hit our warehouse floor. What previously would have been shipping delays and inventory mismatches now gets fixed with a quick click.

Results of the MVP

The impact of these improvements rippled through our fulfillment ecosystem. The biggest transformation came from giving brands visibility and control over their data: the detailed replenishment tracking became their go-to reference, reducing the need for constant email exchanges with our Inventory Managers, and the two-step flow caught over 10% of the mismatches.

While we still had work to do on our prediction algorithms and other advanced features, these foundational improvements marked a crucial first step in rebuilding trust. Brands could finally see, track, and manage their inventory flow end-to-end within our platform - translating into faster fulfillment times and less churn from the fulfilment service.

2. The corners project

Did it ever happen to you that you’ve built an MVP, knowing you'd have to cut corners, only to watch those compromises become permanent features? It’s way more common than we’d like to admit in the agile world: we ship the MVP, then immediately pivot to the next priority, leaving those "temporary" solutions firmly in place.

But this story is different. This is about going back and adding those corners we had to cut.

After three quarters of major improvements across our fulfillment ecosystem - new warehouse, expiry management tools, order automation - our replenishment MVP came back in focus as a painful experience. The catch? Each individual issue seemed too small to justify the "epic" title. Yet collectively, these "small" problems were costing us big.

Building the case

To convince our team's main stakeholders that it was worth carving out some time to improve the user experience, I started by gathering data. First, I launched a targeted Hotjar survey that caught brands right after completing their replenishments, capturing:

• satisfaction scores (from 1 to 7)
• improvement suggestions (a free form)

The product metrics revealed deeper problems: a 30% form abandonment rate wasn't just users giving up - Hotjar recordings showed a pattern of brands starting the form, realizing they needed to analyze their stock data first, abandoning it, then returning later. Combined with a 12-minute average completion time, we were watching our users create makeshift workflows to compensate for our tool's limitations.

The Action Plan

I've defined 18 critical improvements needed - from basic sorting capabilities to CSV exports. While each fix seemed minor in isolation, together they would transform the replenishment experience from a friction point to a powerful self-service tool.

Any work on 'existing features' would compete with new feature development, so I needed the buy-in from my team's stakeholders. My pitch to logistics managers and product leadership centered on two key points:

1. Direct user quotes from our survey highlighting daily frustrations
2. The potential to reduce our operational team's workload by enabling true self-service

The argument worked. What started as 'UX polish' became our squad's primary focus.

Implemented Solutions

Over the next two months, the squad maintained an ambitious pace: one new feature every three working days. Here are our three highest-impact release out of the 18:

1. Search and Filtering
   Our search functionality transformed how brands handled large catalogs, cutting average session time by 40%. It became an essential for a quarter of our users, who now rely on it for every replenishment.
2. One-click Delivery Notes
   Every replenishment needed a printed delivery note for the warehouse to process their merchandise effectively. But brands were stuck manually recreating the replenishment declaration in their own spreadsheets. By automatically generating PDF delivery notes, we eliminated this double-work for 45% of our users. What used to cause warehouse rejections and processing delays now takes just one click.
3. Save replenishmet as draft
   Building a replenishment declaration is a chore: it takes centralising data from multiple sources. I've often seen hotjar recordings of users starting to complete it, and then going away to find the last piece of the puzzle. Many times their session ended before they could submit the replenishment. Saving drafts reduced form abandonment from 30% to 11.5% by letting brands step away to gather information without starting over.

Results

Altho we have moved the needle on a lot of direct product metrics:

• Form abandonment plunged from 30% to 11.5%
• Average completion time dropped from 11 to 7 minutes
• User satisfaction rose from 5.5 to 6 (out of a total of 7)

Yet our initial goal - reducing total replenishment completion time - remained unchanged. Conversations with logistics experts revealed why: transit time dominated the 7-day process, making our interface improvements virtually invisible in the grand scheme.

But we'd accidentally sparked a more valuable behavior change: brands began replenishing more frequently, shifting from every 30 days to every 20 days on average. This meant more consistent stock levels and, ultimately, higher fulfillment service adoption - our true north star metric. Sometimes the most significant improvements don't come from solving the problem you initially set out to fix.

What's next? Exploring Predictive Inventory Management

Once we’ve covered all basics and even brought some quality of life improvements, it was time to get back to solving the most fundamental problem with replenishments, the one that we discovered in over very first usability test: the lack of support in stock prediction.

Our users were asking for CSV exports of stock data, but we had to dig deeper to understand the real need behind that. Every month, the brands' operations experts are building complex excel sheets of stock modelling in order to determine the quantities of merchandise to send to our warehouse. This takes hours of work, and more often than not, the prediction turns out to be wrong.

We aren't just providing warehouse space and shipping services - we are sitting on a goldmine of data that could help brands make better inventory decisions. Every order, every seasonal trend, every stock movement through our warehouse is a data point that could inform future predictions. And we're a tech company that can make this data actionable.

Wizard-of-oz testing approach

Building trust in automated predictions is a delicate process. Rather than investing massive resources into a complex algorithm upfront, we opted for a more measured approach: using a specialized software in algorithmic prediction, Pigment, to start testing stock suggestions.

Our approach to testing this hypothesis is deliberately incremental. Instead of immediately building prediction capabilities into our platform, we're starting with a 'wizard-of-oz' experiment: sharing Pigment's predictions with our bigger brands through spreadsheets. While not scalable, this approach lets us learn crucial information:

• How much do brands trust our predictions?
• What additional context do they need to make decisions?
• How do our predictions compare to their own forecasting?

Early results are promising. Brands who've received our predictions report spending less time on forecasting, though they still cross-reference the numbers with their own calculations. This hybrid approach - where we provide predictions but brands maintain control - might actually be the sweet spot we're looking for.

Based on these learnings, I'm prototyping how these predictions could be integrated directly into the replenishment flow. My main goal is to enhance the current users's own expertise with data-driven insights, while allowing them to keep control.

The journey continues, and we're excited to learn from each step along the way.

Reflections

We started with a basic form that simply got the job done. We refined it into a tool that made work easier. Now, we're exploring how to make it genuinely intelligent. The evolution of our replenishment system reflects a deeper truth about building products: sometimes the biggest impact comes not from introducing new capabilities, but from truly understanding and supporting existing workflows.

As we continue to develop our predictive capabilities, this understanding will remain our north star: success isn't measured by how sophisticated our predictions become, but by how seamlessly they fit into our brands' daily operations.

The next chapter of this story will be written together with our users, one replenishment at a time