Service Channel
Feb 2022

Work Order Anomaly Detection

An AI model that runs in the background and generates notifications when potential anomalies are detected on a WO.

Industry

SaaS

Platforms

Web-App,
Responsive Web

My Role

Lead Innovation Growth Product Designer

Project Context

Crash Course 101 about Service Channel

Service Channel is a SaaS cloud-based service automation platform for multi-site facility management to help manage the entire process of facilities management, including finding contractors and suppliers.

Service Channel is a marketplace with two primary users: The Subscriber and The Provider. The Subscribers are multi-locations that you commonly know, like, Chipotle, Louis Vuitton, etc. They purchase SC as their facility management software. The providers offer services like plumbing, repair, maintenance, etc.

*SC=Service Channel

Project Summary

This is an Innovation project that doesn't adhere to a standard design process; instead, it focuses on determining the business value of new ideas. It involves participation and leadership in the Voice of Customer (VOC) initiative to create a product that boosts recurring revenue for the Service Channel. The 'Work Order Anomaly' was an experimental idea developed during the innovation process. This process uses artificial intelligence and involves close collaboration with a dedicated cross-functional team.

Overview

About the Project

As competition in the facility management sector grew, Service Channel began to lag, unable to meet market demands.

To optimize our investments and time, we adopted a user research approach known as "Lean Portfolio Management" (LPM). Throughout this LPM research series, we gathered over 1400 data points and spent 60 hours engaging with our customers. Amidst an abundance of ideas for Service Channel's next product, a data-centric charter emerged, designed to provide our users with data-driven insights for more informed facility management decisions.

My Team

Product Manager
Data Scientist
Executive Board (CEO , CFO, & CPO...etc.)
Director Of Innovation
Duration

13 month

Growth Accelerator Charter
This charter was written by the executive board as a hypothesis on where there is an opportunity for value based on the data gathered in the LPM stage.

Team Formation

Fun Fact, my team was formed via a series of DisC Assessment to determine our compatibility as a pod to deliver results.

The Problem

ServiceChannel Lacked a clear view of personas, workflows, and Jobs to be done in its core markets resulting in a product portfolio that isn’t optimized for revenue growth

The Challenge

HMW identify untapped opportunities in the market that not only provide customer value but also offer monetization potential, leveraging over a decade's worth of Service Channel's historical facility management data?

The Innovation Process

The innovation process starts with analyzing the market, identifying potential customers and finding gaps. This is confirmed through customer conversations, or VOC. Understanding the issues leads us to devise potential solutions and test them with customers, checking their fit and the customers' willingness to pay.

The Innovation Process
The innovation process that my team & I followed, it’s a hybrid of lean UX and the double diamond with a mix of product strategy

How do we know we are on the right path of creating the next  ✨ IT ✨ product?

These four gates symbolize our updates to our growth board, detailing our findings. The growth board is comprised of C-suite executives, innovation experts, and strategists to know if we were on the right track of creating the next IT product.

Innovation Stages

Similar to start-up funding, each stage of innovation represents the allocation of investment funds to either advance or discontinue our project.

1. Market Discovery

Understanding the process, Team Formation, Market Analysis, SME VOC.

2. Idea Gate + Problem Worth Solving Gate

Identify a customer problem worth solving in a market that is strategically interesting.

3. Customer Validation Gate

For the concept developed in the previous stage, we aim to present it to real customers who will find value in it, ensure it has a viable business model, and ensure there are no feasibility obstacles.

4. Business Model Validation Gate

We have an appropriate level of risk associated with building and selling our solution at scale as a sustainable business.

Market Discovery

I collaborated with our data scientist on the team to create a MEKKO chart, segmenting customers by their annual repair and maintenance expenditure and their respective sectors. The MEKKO showed our future facility management data solution would primarily target Hyper-enterprise and Enterprise segments.

This belief is based on an assumption/evidence that:
‍
1. Hyper-enterprise & enterprise spend more on R&M, so we assume they have more spending power overall.
2. Due to the scale of their operations, there are also more potential savings they could get from data and insights.
3. There's a greater need in larger organizations to advocate upwards, and provide solid evidence when doing so.

Voice Of Customer

User Research

I crafted a research plan, and through 15 VOC sessions, we pinpointed key issues, focusing on benchmarking and data-driven decisions. At the end of this process, two distinct personas emerged: a transactional Facility Manager (FM) overseeing daily operations, and a strategic FM in charge of decision-making, aligning with our charter's focus on data-driven insights for strategic facility decisions.

“I want Service Channel to flag for me if a WO is too
expensive based on industry average”
- Elizabeth, All Birds

“How can we make benchmarking comparison apple to apple, Panda Express inside a mall is different from Panda Express in a parking lot”
- Roger, Panda Express

Under the hood

The Challenge

The two personas, transactional and strategic, each presented unique challenges. While their pain points shared similarities, the differences were significant enough to require separate solutions.

‍Instead of choosing to focus on one persona, we decided to explore the problems of both personas until we could gather enough information to determine which persona's issue presents a monetizable and impactful problem worth solving.

HMW
How Might We

Some of the HMW we were trying to solve as a team were:

1. HMW alert facility managers that a proposal isn’t within industry standards?

2. HMW help facility manager optimize their provider network?

3. HMW help facility managers know if they are spending within the industry average?

4. HMW help facility managers detects issues within a WO before they occur?

Ideation

I led a team workshop called 'Assumption Mapping'.

In it, we spent 10 minutes thinking about potential solutions for the problems we identified. The riskiest guesses naturally ended up in the high importance and uncertainty section, also known as 'Leap of Faith Assumptions'.

The workshop's goal was to narrow down our guesses and decide where to focus our efforts to test out ideas.

Outcomes, Leap Of Faith Assumptions:

1. Transactional persona would benefit from a Proposal Recommender to approve/decline a proposal.
‍
2. Strategic persona wanted to see if they were within industry average, and they would benefit from a spend benchmark solution
‍
3. Transactional and strategic persona’s would benefit from a detection of WO Anomalies.
‍
4. Both Strategic and Transactional persona would benefit from a Provider Network Evaluation

Assumption Mapping

Experimentation

We refined our assumptions and evaluated their impact using 'Fit' and 'Wow' scores, measured on a 1-10 scale. 'Fit' assesses how well a solution aligns with the user's organization, while 'Wow' gauges solution excitement. Together with qualitative data, these scores helped us decide if an idea was worth pursuing.

FIT Score
WOW Score
1.
Proposal
Recommender
HMW alert facility managers that a proposal isn’t within industry standards?

A proposal is a quote from a provider( the person who does the fixing) however how do Facility managers know that this proposal is within the industry average?I created a lo-fi concept to serve as a tool to keep transactional facility managers informed whether a proposal is within the industry average or not by leveraging SC historical data.

Proposal Recommender
FIT
WOW
WHY

The issue: the solution lacked market appeal and its potential savings were marginal, making it more suitable as a minor enhancement by our product team. Compared to other problems, its impact was minor. Hence, it was passed to the product team as a feature enhancement, and killed from our project.

2.
Provider Network Evaluation
HMW help facility manager optimize their provider network?

A subscriber like Walmart, can have over 700 providers (those who do the fixing) on-call in their network so when something goes wrong, they are able to quickly attend to it. However how do Walmart know that their network of providers is optimized to good performing providers who aren’t scamming them and are doing satisfactory work?

This was considered actually a shark bite rather than a mosquito bite between all verticals and size segments, we had high fit and wow scores for this solution around 10 for both fit and wow.

Provider Network Evaluation
FIT
WOW
WHY

This problem was substantial, with many edge cases and data facets. As a result, a charter was developed for this opportunity, using my low-fidelity designs to propel further experimentation and testing, so it was killed for my team.

3.
Spend
Benchmarking
HMW help facility managers know if they are spending within the industry average?

A proposal is a quote from a provider( the person who does the fixing) however how do Facility managers know that this proposal is within the industry average?I created a lo-fi concept to serve as a tool to keep transactional facility managers informed whether a proposal is within the industry average or not by leveraging SC historical data.

FIT
WOW
WHY

While the experiment was successful in concept and we wished to develop it further, the feasibility was challenging.

1. Service Channel had issues with 'dirty' data ingestion, and we couldn't use our customers' data for benchmarking.

2. The data couldn’t be standardized. For instance, when comparing janitorial spending between CVS and Walgreens, each company might have a different definition of the janitorial trade. This meant that the benchmarking solution couldn't be self-service and required manual adjustments to account for these variables.

4.
WO Anomaly
Detection/
WO Intelligence
HMW help facility managers detects issues within a WO before they occur?

A work order is when a location like Gap for example reports they need a provider to come on-site to do fixtures. People who work at these locations and report these problems don’t have experience in facility management and often enter inaccurate data that later the facility manager needs to be on top of.

The WO intelligence concept provided a holistic overview and analytics based on comparing similar WO’s and providing suggestions.

WO Intelligence
FIT
WOW
WHY

This experiment yielded promising results and valuable feedback from our customers, even in its low-fidelity form. For instance, the 'Work Order Intelligence' merely provided static insights like a dashboard, rather than being proactive.

This led to the need for an iteration that utilized historical data to predict potential issues and necessary actions.

Interactive Prototype
From WO Intelligence to WO Anomaly detection

Outcome &
Takeaways

Neither Pursue or Kill but rather PIVOT

We failed to narrow our scope early for two user groups, and prematurely favored 'benchmarking' as a solution, leading to skewed research data. Essentially, we were fitting a problem to a preconceived solution instead of the other way around

1. AI model effectiveness hinges on data upkeep.
Service Channel competitive advantage was it’s wealth of 10+ years of data, but because there was no data upkeep, we faced garbage in garbage out scenario.

2. Development is expensive
‍We did save $480,729 dollars by pivoting earlier prior to development.

3. Additional Discussions
‍Points which are not covered but could be worth discussing

  • IFD (Intensity, Frequency, Density) Calculation.
  • Presenting to Executive leadership
  • Agile Fast Sprint
Thanks for making it this far. Here’s a GIF for you.

More Projects

Workflow Dashboard

Re-designing the dashboard to allow for workflow management and efficiency

Customer 360

Migrate from legacy systems into Salesforce platform to improve the customer and employee experience.

Yello Mango’s MVP

Deliver low-cost, affordable meals to working professionals that fits their preference.

Equity Multiple-Resources

Provide discoverability of real-estate investment resources to help new investors be well-versed with this space.