Share
Explore

icon picker
Maxis Communications

Last edited 1 minute ago by Tobi Konitzer

Background documents

Ocuboost Architecture:
image (7).png

High-level Pilot Project Scope

Maxis business challenges being addressed:
Optimize acquisition costs of net new e-commerce customers
Leverage the latest in AI/ML technologies to get more value from its first-party data and enhance and extend Maxis’ existing data science capabilities.
Solutions to be deployed to deliver the pilot
Ocurate Value-based bidding on Google Ads
Access to Ocurate Portal to allow for campaign tracking
Ocuboost feature share (Ocuboost)
Real-time revenue predictions share

Maxis teams involved in the pilot
Marketing (CMO is primary exec sponsor)
Data Science (CIO, who owns Data & AI teams)
Asif Muhammad, Head of Data & AI CoE, to lead the pilot effort
Digital (reports under CMO and is responsible for Maxis and Hotlink mobile app, as well as e-commerce advertising)
Proposed pilot duration
93 days (3 months+)
Pilot Use Cases & Success Criteria
UC1 - Acquisition Optimization
Core KPI = 15% increase in Return On Ad Spend (ROAS) for net new customers on Google Ads
UC2 - Ocuboost Behavioral Features to enhance any one of the following a) Churn prediction, b) LTV/per-customer-revenues prediction and c) Prepaid to Postpaid conversion
Core KPI = % increase in performance against existing Maxis churn and LTV predictive models and baseline pre-to-postpaid subscriber conversion rate
Digital media discovery
What is the Maxis' ad spend on Meta and Google respectively, both monthly and projected for the next 12 months?
What is the average customer acquisition cost (CAC) on Meta and Google?
What is the ARPU (12-months or based on agreed-upon window for per-customer revenues) for customers acquired digitally?
Describe your current optimization efforts on Google Ads?
Do you work with an agency for digital media management?
Updated on July 3
Maxis Pilot Use Cases & Success Criteria
Name
Solution
Differentiator
Use Case
Success Criteria
Go Deeper
1
UC1 - Acquisition Optimization
Ocurate VBO
Problem statement: In a highly competitive at penetrated market like Malaysia, the vast majority of service provider web ads come from competitors (i.e. switchers) and it is extremely difficult to predict - at the moment of purchase - which ones are worth acquiring and at which price point
Outcome: acquire customers with higher LTVs/revenues at less cost automatically and consistently over time
Value: reduced customer acquisition costs, increased customer LTV, increased profitability
Core KPI: 15 % increase in Return On Ad Spend (ROAS)
Validation of ML accuracy predicting per-customer revenues for newly acquired customers (initial KPI)
+15% ROAS increase vis a vis control group (core KPI)
Open
There are no rows in this table

Maxis PoC Phases and Capabilities
1
Phase
Sub Phase
Name
Outcomes
Requirements
Activities
Deliverables
Delivery Start
Duration
Phase 1 - Occurate
11
Phase 0
Ocurate Initial Onboarding
Integration into key data sources
Share access to Ads API (Marketing)
Install Ocuboost (IT/website owner) ~ deployment doc here
Ocuboost will reconcile languages on the backend
Share CRM data (ecommerce director) ~ overview doc here

ID integration

We will use UUID for ID integration, which is observable directly through Ocuboost, and can be shared as part of the CRM data, allowing Ocurate to merge the back-office CRM data with the event-type data Ocuboost collects updated on July 3
Data residency
Which region is preferred for AWS/GCL deployment? (assumption: Malaysia)
30/07/2024
1 day
Phase 1
ML development
Ocurate trains custom Deep Learning algorithm for per-customer revenues
Data assumptions meeting; agreement on the time window for LTV/per-customer revenues used to optimize ad spend updated on July 3
30/07/2024
45 days
Phase 2
Preliminary KPI: ML accuracy
Validation of ML accuracy predicting per-customer revenues for newly acquired customers based on scaled Mean Average Percentage Error (MAPE) (i.e. MAE/Actuals)
Validation meeting
Validation collateral
14/09/2024
1 day
Phase 3
Integration into Maxis data lake
Integrate real-time revenue predictions into Maxis data lake for free use
Access to real-time data stream of revenue predictions
14/09/2024
1 day
Phase 4
Integration into Google Ads APIs
Integrate real-time revenue predictions into Google ads API to train Google’s bidding algorithm
14/09/2024
1 day
Phase 5
Integration into Ocurate Portal
Integrate data into Ocurate portal to allow tracking of campaign performance including value-basedbidding
Access to Ocurate Portal
14/09/2024
1 day
Phase 6
Train Google impression bidding algorithm based on real-time revenue predictions
Google Ads API ready to place bids based on Ocurate ML
15/09/2024
30 days
Phase 7
A/B test: ROAS increase
Matched market test to compare ad effectiveness in terms of ROAS, i.e. predicted LTV/per-customer-revenues with a time-window agreed upon in the data assumptions meeting: CAC, of value-based bidding vs. conventional/current bidding updated Jul 3
Carve out a single acquisition campaign or a blend of acquisition campaigns anywhere within the alphabet ad real estate (Search, Video/Youtube, Shopping Google PMAX, Youtube, Performance Max, Smart Ads) that generates at least 500 conversions every two weeks. updated Jul 3
Ocurate dashboard allowing tracking of ROAS increase
15/10/2024
15 days
Phase 8
Core KPI: ROAS increased by 15%
Register a 15% increase in ROAS vis-a-vis control group
Currently, the KPI for Maxis’ acquisition campaigns is CPA (Cost per acquisition/customer acquisition cost), for this pilot, KPI will be moved to ROAS (predicted LTV/per-customer-revenues with a time-window agreed upon in the data assumptions meeting: CAC) for both the treatment and control arm updated Jul 3
01/11/2024
1 day
Control
Project Management
Regular cadence created to capitalise on early momentum
Successful delivery of outcomes and deliverables
documented risks and issues and mitigation actions monitored and tracked
weekly status meeting allowing early intervention in case of issues or delays
risks and issues log
project management plan
project progress reports
meeting minutes
30/07/2024
93 days
Phase 2 - Feature Share
4
Phase 0
Data Discovery
Assessing and evaluating fitness for purpose of source data (Ocuboost and derived shared features from CRM) to enable and drive the commercial outcomes for Maxis
Discovery: Understand Maxis data platform?
Prioritization of features identified
Data Discovery
Data Review
Definition / write-up of initial key decisions / KDR around clients data
Initial Data Dictionary /ICD

02/08/2024
30 days
Phase 1
Data ingest refinement
Evaluate source data initially shared
Analyze predictive power of features collected in phase 0 for various Maxis production models
Refine feature share
02/09/2024
15 days
Phase 2
Enrichment Service
Production Deployment Enrichments
Share data back to Maxis data lake

Data share pipeline from Ocurate VPC to Maxis production servers
02/10/2024
15 days
Control
Project Management
Regular cadence created to capitalise on early momentum
Successful delivery of outcomes and deliverables
documented risks and issues and mitigation actions monitored and tracked
weekly status meeting allowing early intervention in case of issues or delays
risks and issues log
project management plan
project progress reports
meeting minutes
02/08/2024
88 days

Timeline
1

Resourcing

Ocurate will allocate
Dedicated data scientist
Dedicated marketing specialist for campaign integration
Dedicated dev ops support/data engineer

Compute and privacy

Ocurate embraces practices that promote security, observability, and accountability. We choose technologies and services which are standard within the industry, make heavy use of managed cloud services, and ascribe to the principle of least privilege.
Ocurate uses managed, serverless cloud infrastructure to build its products and data processing systems. All of Ocurate’s cloud resources are hosted in AWS and Google Cloud in the appropriate geographic regions necessary to comply with local regulations (e.g. GDPR data in Europe, PDPA data in Singapore, etc.). Client and customer data is encrypted both in transit and at rest. Access to such data is strictly limited only to the Ocurate personnel who need it to perform the specific data processing contracted by our clients, and all accesses are auditable and can be traced to the individual who initiated the access.



Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.