Demand Forecast
Tiki is an e-marketplace with its Retail business operated by Tiki itself. Demand forecasting is a crucial part of any retail supply chain. We built a decision-support system from scratch to automate parts of this process and help our users—Tiki’s internal employees—make more informed decisions.
Summary
Challenge
Manual demand forecasting no longer worked well, falling short in accuracy and coverage. A more reliable and effective solution was needed to improve decision-making.
What I did
Conducted a user observation to analyze the user's current method.
Formed the product concept with PM, DA & Dev.
Designed the UI.
result
The research insights informed the developing & design process.
After launching the first version, overestimation rate dropped significantly to only around 10%.
what i've learnt
Designing decision-support systems effectively requires a thorough understanding of user's decision making process (typically experts in the task), plus domain knowledge, plus technical expertise.
Context
Inventory is a crucial investment in retail. Overstocking ties up cash, while understocking leads to missed revenue. The right inventory level meets customer demand, a challenge assigned to the Procurement team. Each month, they must forecast demand and plan stock procurement and replenishment.
The manual approach worked in the early days when there were fewer products and cost optimization wasn’t a major concern. However, as the business scaled to 100,000+ SKUs, the Procurement team faced stricter coverage and accuracy requirements. The manual process became inefficient, making
automation solutionessential.As our tech team was still forming while business needs were very urgent, our approach was to build a first simple version with whatever initial impact we could make rather than waiting for a complete, more advanced AI/ML solution that may be more impactful but takes months or even years to finish.
Research
Given the needs to automate and optimize this decision-making process and the above context, we started with
studying and analyzing the current workflowto identify opportunities for us to jump in.I conducted a
Contextual Inquirywith 12 users, each one in charge of different product type (Mom and baby products, Health & beauty products, Electronic devices, …). This method involved observing their end-to-end workflow and interviewing (semi-structured) them during the process.Questions to be answered
How is the demand forecasting task currently performed? What the process is like? What calculation they make?
Why they do what they do?
Identify the problems or challenges which could lead to low coverage and accuracy.
Findings
After collecting the data, my
thematic analysisrevealed their methods & process, as well as the potential risks it could cause to the task’s accuracy & coverage. The findings, as well as my recommendations for next steps, are put together in the below report.The research was complexbecause not only I have to record their processes, I also have to analyze their calculations, data references and forecast strategies, which requires a lot of knowledge regarding business strategies, data, and terms within the domain.My report was a base for our team (PM, Data Analyst, Developers) in further discussions to identify opportunities and define the product concept.
Design the product concept
High-level design
Product goals were defined first as a
design compass. Success metrics were also defined along the way as the product idea is formed.We approached this stage by producing a
rough idea of how the product worksin the form of user flow, questions + rough answers and/or quick sketches.The goal is to visualize the overall logic of the product and identify the key steps users go through to complete their tasks.
We then gradually add more detail to these rough ideas until the directions are clear enough for each role to focus on their part — the PM on more detailed product logic, and me on more detailed UI design.
Outputs from this stage usually include:A half-finished product concept document.
A rough design for main screens with important elements addressed (quickly put together using design systems + notes).
Product features
Given the above findings, we decided to build a decision support system with core functions to:
Generates prediction suggestions using simple forecasting methods*.
Allow users to review and refine suggestions with reference data displayed on the UI. Some reference data are directly pulled from sources users usually refer to, while others are pre-calculated to save effort and reduce human error.
We also identified supporting features to improve usability and forecast coverage, such as Advanced filtering, Data sorting, Marking products as reviewed or not, Reason for too small/too large numbers, etc.
*Our first version generated predictions using simple time series method along with a set of rule defined based on the research findings. More advanced Machine Learning methods were introduced in later stages.
Read more at: engineering.tiki.vn
Our design process
The decision-making process involved
brainstorming between the PM and me, followed byalignment discussions with key stakeholders— including the BI team, development team, and business representatives — to ensure that the data and features in the tool were both meaningful and feasible.Since users still make decisions manually, we
worked closely with the BI teamto determine which data to display and how to present it so that it would best support their forecasting judgments.
From concept to final UI
In this project, I spent a lot of time on
prioritizing which data mattered mostfor decision-making and designing a clearvisual hierarchyto support it. This required a deep understanding of each data point — what it represented, how it was calculated, and how users relied on it to make forecasting decisions.The interaction design also demanded special attention, especially for core functions such as viewing, editing, and submitting forecasts, since various
edge casescould occur and needed to be thoughtfully covered.Given how data-heavy and complex the product was, my goal as a designer was to create an interface that
visually felt comprehensive yet simple, empowering users without overwhelming them.At this stage, I worked mostly independently but stayed in close contact with the PM. As design and logic were closely connected,
many details changed along the way. Some ideas that seemed clear in early discussions turned out differently once visualized, so we often had to revisit and refine them together.
Launch
As we launch our first version,
different metrics were trackedto measure the product success, such as user's review & edit behaviors, product coverage, system's forecast and user's forecast accuracy (compared retrospectively to actual demand).Forecast accuracy improved gradually over time. After 2 months,
overestimation reducedfrom a large margin to just around 10%.In terms of
product adoption, the transition from Excel to our tool was pretty smooth. We reached 80% adoption across 13 category teams after 1 month and by the second month, the demand forecast process was completely transferred to our tool. Despite the unfamiliar interface and task complexity, the learning curve was low—only a few of the 30 users needed assistance.