Retail Media Network for a Southeast Asian Electronics Chain

Industry

Retail & E-Commerce

Geography

Malaysia, 10 states and 50+ cities

Cooperation Period

12 Months

Executive Summary

Clixlogix helped a leading Malaysian consumer electronics retailer turn fragmented shopper data and underused store screens into a measurable retail media business. The engagement started with endemic suppliers as the first buyer segment and built the data foundation, activation platform, brand buying portal, and attribution infrastructure that endemic advertisers needed before they would commit spend.

The client had run a proximity marketing initiative in the previous decade that was retired once the underlying category economics collapsed. The internal commerce team carried scar tissue from that experience into any subsequent conversation about store technology. The new brief was different in shape.

The client wanted to convert the attention their physical store estate generated into a durable revenue line. The route to that revenue was a retail media network that would let endemic brand suppliers reach verified shopper cohorts across the store, the client’s ecommerce experience, and offsite surfaces on Meta and Google.

A short strategic assessment produced three findings that shaped the recommendation.

The chain had five distinct pools of shopper data that had never been unified. Any retail media offer would fail commercial diligence from sophisticated brand advertisers without a unified shopper view sitting behind it. Point of sale transaction history, loyalty application activity, Wi-Fi guest network sessions, ecommerce browsing, and post purchase warranty registrations.
The commercial relationship with endemic brand suppliers was transactional and negotiated annually as trade marketing spend. Moving those dollars into a self serve buying motion required both a technical platform and a commercial reframing that the client sales team would need to own.
The chain’s in store digital signage estate had been installed for merchandising rotation and category promotion. Turning it into monetizable advertising inventory required addressing screen coverage gaps, standardizing creative formats, and building the ad operations discipline that brand advertisers expected.

Clixlogix delivered the retail media network across five tracks spanning the data foundation, the media platform, the brand portal, the measurement pipeline, and the audience extension motion.

The engagement produced a working platform that opened a new revenue line for the chain. Endemic brand relationships shifted from annual trade negotiation to always on self serve campaign management. The store estate moved from a cost of doing business to an asset with measurable media value.

Fig 1 – Retail media network system architecture across data foundation, media orchestration, measurement, and audience extension

About the Client

The client operates a consumer electronics retail chain with presence across 10 states and more than 50 cities in Malaysia. The store estate spans high street outlets across state capitals and mall anchored formats inside major mall groups such as Sunway, IOI, Pavilion, and Mid Valley, plus specialist electronics destinations including Low Yat Plaza in Kuala Lumpur and Digital Mall in Petaling Jaya.

Product coverage runs across appliances, personal electronics, computing hardware, and home cooling products at a broad price range. Category leaders in appliances, televisions, and mobile devices sit alongside private label offerings that give the chain margin flexibility on entry price points.

The chain occupies a mid market position with strong regional brand recognition. The supplier network reaches premium and value tier categories through relationships built over multiple expansion cycles.

The commercial history that brought the client to a retail media conversation was specific. An earlier proximity marketing initiative had been retired once the underlying category economics stopped supporting the operating cost of the program.

The internal commerce team had absorbed a strategic lesson about being early to store technology cycles. The retail media network conversation was different in kind, since the reference set was no longer speculative.

Amazon Ads, Walmart Connect, Kroger Precision, and a growing set of regional retail media networks in Malaysia had established the category. Brand advertisers had budget lines allocated to retail media as a distinct channel. The client saw a window to establish position in the regional market before the larger international players compressed the opportunity.

Any platform sitting on shopper data at this scale in Malaysia also had to align with the Personal Data Protection Act 2010 as amended by the Personal Data Protection Amendment Act 2024, which brought the Malaysian regime closer to global data protection standards through phased implementation across 2025.

The amendment introduced mandatory data protection officer appointment, data breach notification obligations, data portability rights, and enhanced cross border personal data transfer controls. Consent scope, purpose limitation, notice, retention, and data minimization all had to be enforceable through the platform itself, since brand advertisers running diligence on the network would inspect the consent posture before committing budget.

The Challenge

The engagement carried three challenges that had to be resolved in parallel for the platform to reach the commercial threshold that endemic brand advertisers required.

Internal skepticism about another store technology bet

The retired proximity program had consumed operational bandwidth for a period, produced a measurement dataset that was interesting but never converted into durable commercial value, and taught the internal commerce team to apply harder diligence to any subsequent store technology proposal.

The retail media network proposition needed to survive that diligence with a defensible revenue model, a defensible technology architecture, and a defensible operational plan for the ad sales motion that would sit on top of the platform once it was live.

Shopper data trapped in silos

The five data pools sat in different systems with different identity structures, different refresh cadences, and different consent scopes.

Point of sale carried transaction level detail with customer identity attached only when the shopper presented a loyalty card
The loyalty application carried device level identity with permission based location and browsing signal
Wi-Fi guest network sessions carried device level identity captured through a splash page consent flow. That flow had been designed for guest network access, which meant the consent scope had to be re contracted before those identifiers could be used in identity resolution for advertising purposes
Ecommerce carried session identity with cookie level tracking
Warranty registrations carried verified customer identity with product level detail but no ongoing engagement

Reconciling these five identity spaces into a shopper view that endemic brand advertisers could target against required identity resolution work, consent harmonization work, and privacy governance work that all had to land before the media platform could produce commercially credible audience segments.

Endemic brand relationships anchored in annual trade marketing

Trade marketing spend from brand suppliers had historically flowed through annual negotiation, arriving as promotional co funding, category endcap fees, and campaign specific slotting fees.

Converting those dollars into self serve retail media buying required both the technical infrastructure to support audience based campaign booking and a commercial motion inside the client sales team to walk brand marketing teams through the shift.

The internal category buyers who managed the endemic brand relationships did not naturally speak the language of CPM pricing, impression estimation, or attribution windows. Clixlogix would need to support the client through the operational transition alongside delivering the technical platform.

The platform needed to:

Stand up a retail media platform that endemic brand suppliers could buy advertising through, priced and reported in the terms those brands expected from other retail media networks they already worked with.
Unify shopper data across point of sale, loyalty, Wi-Fi, ecommerce, and warranty into a shopper view that could support audience segmentation and attribution while aligning to Malaysian PDPA principles including notice and choice, purpose limitation, retention, and data minimization.
Convert the in store digital signage estate into monetizable inventory with the operational discipline to run audience based campaign booking against it.
Produce measurement transparency for brand advertisers that stood up to diligence, using DOOH appropriate impression estimation for in store surfaces and web native measurement for onsite and offsite surfaces.
Preserve strategic optionality to extend shopper data to offsite platforms for audience extension buying against Meta, Google, and regional programmatic exchanges once the onsite foundation was in place.

The Approach

Clixlogix approached the engagement in four phases.

Strategic phase. Retail media commercial model, platform architecture, and operational plan for the ad sales motion.
Foundation phase. Shopper data unification and identity resolution infrastructure.
Platform phase. Media orchestration engine, endemic brand self serve portal, and measurement pipeline.
Extension phase. Offsite audience extension surfaces integrated under a single buying interface.

Fig 2 – Four phase engagement roadmap with track sequencing across foundation and platform phases

The sequencing decision that shaped everything downstream was to start with endemic brands as the first buyer segment before opening the platform to non endemic advertisers through a future programmatic exchange integration on OpenRTB 2.6.

Endemic brands already had a commercial relationship with the client, already carried budget allocated to trade marketing spend that could migrate into retail media buying, and could absorb the platform learning curve without requiring the polish that a first party exchange presence would demand. The exchange integration was scoped as a later phase once the endemic revenue line had reached operational maturity.

The Solution

The five tracks that delivered the platform ran in parallel across the foundation, platform, and extension phases. The strategic phase produced the design decisions that governed each track.

Shopper Data Unification

The client’s category managers had been running merchandise strategy on partial data for years. Loyalty knew loyalty behavior. Point of sale knew transaction behavior. Wi-Fi knew visit behavior. Nobody had the same shopper resolved across every source. This track gave the client a single shopper view that endemic brand advertisers could pay to target against, that the internal category managers could use to plan assortment, and that the measurement pipeline downstream could join campaign exposure against actual purchase behavior.

The first track built the shopper data foundation that everything downstream depended on.

Clixlogix delivered a customer data platform that ingested the five identity streams and produced a unified shopper profile queryable for audience segmentation and attribution. The platform sat on a cloud data warehouse with a medallion architecture separating three zones.

Raw ingest (bronze), holding source events in the shape they arrived from the source system, with typed schemas enforced at the ingest boundary through a schema registry
Cleansed and conformed data (silver), holding harmonized entity representations with consistent schemas, resolved identity keys, and applied consent scope
Served analytical tables (gold), feeding both the audience segmentation service and the attribution pipeline downstream, with query performance tuned against the minute scale refresh windows that pre booked campaign audience delivery required

The normalized event contract that flowed through the pipeline carried a consistent shape across the five source system vocabularies, so downstream consumers worked against a single set of event types (pos_transaction, ecommerce_view, wifi_session, loyalty_event, warranty_registration). Every event carried a common identity, timestamp, consent scope version, and confidence score envelope regardless of source system.

Fig 3 – Five identity streams converging into a unified shopper profile through the customer data platform

Ingestion was handled through an event streaming pipeline built on managed streaming infrastructure. Point of sale transactions, loyalty application events, Wi-Fi session events, ecommerce interactions, and warranty registrations flowed into the raw ingest zone within minutes of the source event.

Batch ingestion handled the historical backfill for the initial platform launch, replaying multiple years of transaction history into the unified profile store before advertisers could target segments with reliable behavioral depth. Streaming ingestion handled ongoing operations, with idempotency keys applied at the event level so replay and recovery scenarios did not double count shopper events downstream.

Retail media networks at this scale carry large volumes of shopper events across in store telemetry, ecommerce interactions, loyalty engagement, and payment events. The architecture separation between batch and streaming reflected an operational reality across the retail media category. Different use cases carry different latency requirements.

Audience segmentation for pre booked campaigns can tolerate refresh windows measured in minutes
Screen orchestration on in store inventory required cache lookup in the low hundreds of milliseconds for campaign changes and pauses to propagate
Attribution measurement ran overnight and could tolerate hour scale processing windows

Sizing a single ingestion tier against the strictest latency requirement would have overspent on infrastructure that most of the workload did not need. Sizing against the loosest would have starved the operational decisioning path. The split model let each workload class run against infrastructure that matched its actual latency budget, which is the sizing discipline that separates retail media networks with defensible unit economics from those where infrastructure cost erodes the media margin.

Identity resolution ran a two stage blend with guardrails.

Deterministic matching handled the identity pairs where a hard link existed. A loyalty application event carrying the customer identifier that matched a point of sale transaction with the same identifier resolved deterministically at ingest time.

Probabilistic matching handled the harder cases where identity had to be inferred. Wi-Fi session device identifiers matched against loyalty application device identifiers through timestamp proximity, catchment area overlap, and device signal correlation. Each probabilistic match produced a confidence score, and matches above a calibrated threshold entered the unified profile.

Guardrails around the probabilistic path shaped what could be done with the resulting identities.

Raw device fingerprints were never exposed to advertisers or written into audience export payloads
Confidence thresholds varied by use case. Aggregate audience targeting used a lower threshold than individual level personalization, and stale or low confidence matches were excluded from attribution join entirely
Probabilistic identities were eligible for aggregate targeting and reporting. Sensitive personalization and one to one messaging required the deterministic identity path
Consent scope on each source event determined whether the resulting identity could be used for targeting, measurement, or analytics only

Fig 4 – Identity resolution flow, deterministic and probabilistic paths with confidence scoring, time decay, and consent gate

Consulting Insight

Early builds of the identity resolution pipeline treated device signals with equal weight regardless of age. This produced a subtle failure mode. Shoppers who had visited the catchment area once but moved away kept resolving into audience segments they no longer belonged to, which degraded audience quality for advertisers who paid a premium for reach against active shoppers.

Clixlogix added a time decay function against the device signal weights in the probabilistic blend. Signals from Wi-Fi sessions or ecommerce visits older than 45 days decayed at a rate calibrated against typical shopper churn in the catchment. Signals older than 180 days dropped out of the blend entirely.

The decay function preserved probabilistic reach for active shoppers while catching stale identity linkages that would have degraded audience segments for advertisers. Probabilistic identity is a function of both signal strength and signal freshness, and freshness deserves first class treatment in the resolution algorithm.

Consent harmonization was handled through a consent registry aligned to Malaysian PDPA principles of notice and choice, purpose limitation, and data minimization. Each identity source carried its own consent scope, captured during the source system’s own opt in flow and versioned in the registry. Scopes were versioned so that a change in the loyalty application privacy policy triggered a new scope version, and shoppers who had consented under the previous version retained that scope until they opted into the new one.

The registry applied scopes at query time when audience segments were being built for advertiser use. A brand advertiser building a segment for a laptop campaign could only reach shoppers whose consent scope permitted product level marketing on the surfaces the campaign would run against. Consent scope violations were caught at the segment build step upstream of any campaign delivery, which produced a defensible audit trail for the Personal Data Protection Commissioner and for brand advertiser diligence reviews.

The business outcome of this track for the client was concrete. The category planning team gained a unified shopper view for the first time, which unlocked cross channel merchandise decisions the client had not been able to make on partial data. Endemic brand advertisers gained a targetable audience that came with a defensible consent posture, which cleared the diligence bar that had been blocking the retail media revenue conversation.

In Store Screen Orchestration

The client operated more than a thousand digital signage screens across the store network, running the same category rotations and promotional loops that had been in place for years. Those screens generated no revenue and produced no data. This track turned the screens into advertising inventory that endemic brand advertisers could buy against, priced against verified shopper audiences and reported through DOOH appropriate impression estimation that endemic advertisers already used for their out of home buying.

The second track turned the in store digital signage estate into monetizable inventory.

Clixlogix assessed the existing screen deployment across the store network and identified coverage gaps at high dwell locations. Screen inventory was standardized to two aspect ratios and three duration windows that endemic brand advertisers could plan against.

Fig 5 – In store screen inventory zones with daypart overlay

The orchestration engine ran precomputed playlist decisioning against the store inventory. Audience segments were precomputed at the store and daypart level on a rolling five minute refresh cycle. Screen playback drew from those precomputed segments. Per screen real time bidding logic borrowed from web ad serving was intentionally out of scope for the in store inventory, since the DOOH context called for operational decisioning shaped for scheduled playlist inventory.

The engine ran against an in memory cache that carried the precomputed segments, the campaign inventory booked for the daypart, and the active brand safety and competitive separation rules. Cache lookup latency was budgeted in the low hundreds of milliseconds so campaign pauses, creative swaps, and rule changes propagated through the network within the next screen refresh cycle. That mattered when a brand advertiser needed to pull a creative or an ops team needed to adjust separation rules mid campaign.

Cache lifetimes were tuned to the freshness characteristics of each data type. Audience segments held for five minutes to absorb consent scope changes and segment recomputation. Inventory availability held for 60 seconds so booking confirmations and campaign pauses reached the decisioning path within the next screen refresh. Creative asset metadata and brand safety rule sets held for longer windows since they changed on operational cadences. Frequency cap counters ran on an append only counter model updated as impression events landed.

Brand safety and competitive separation rules ran as data driven policies configurable through an operations console.

The client ad operations team could adjust separation windows, frequency caps, and category exclusions without requiring engineering deploys. Rule changes propagated to the decisioning cache on the next refresh cycle.

Impression measurement anchored to opportunity to see modeling drawn from screen playback verification and audience presence signal. Playback verification confirmed that a booked creative had actually played on the scheduled screen at the scheduled time. Audience presence signal came from Wi-Fi analytics and store footfall counters and produced an estimated audience count in the vicinity of each screen during each playback window.

Together these produced a DOOH standard impression estimate that endemic advertisers reconciled against the impression metrics they already used for other out of home buying. Web native viewability metrics from Meta and Google were preserved for the onsite and offsite surfaces that supported them, keeping each measurement standard within the domain it was designed for.

Consulting Insight

The strategic reframe that unlocked this engagement was helping the client see their store estate as an audience the chain had already earned. The second half of that reframe was seeing the in store digital signage as inventory they could sell to brand advertisers who wanted access to that audience.

The retired proximity program had approached store attention as an outbound push channel for retailer promotions, which put the technology investment on the retailer's cost line and made the return dependent on incremental basket lift the retailer captured directly.

The retail media model reversed the economics. Brand advertisers pay for access to the audience, the technology investment sits against a revenue line, and the incremental basket lift becomes a proof point Clixlogix and the client can sell against. A retailer with earned foot traffic is a media company that has not yet been priced.

The business outcome of this track for the client was concrete. The store screen estate moved from a cost center running merchandising rotations to a revenue line with a defensible pricing benchmark. Ad operations became a repeatable workflow the client team could scale into additional store zones without new engineering cycles.

Endemic Brand Self Serve Portal

Category marketing managers at appliance brands, television brands, and mobile device brands wanted to buy retail media the same way they bought Meta and Amazon Ads. Planning windows measured in hours. Campaigns launched, paused, and adjusted through a portal. Reporting available on demand. The traditional trade marketing motion at the client had operated on multi week negotiation cycles routed through email and phone with the sales team. This track gave endemic brand advertisers a buying interface that met the digital media expectation and gave the client sales team the operational capacity to serve dozens of advertisers concurrently through the same infrastructure.

The third track delivered the buying interface that endemic brand advertisers used to plan, book, execute, and report on their campaigns.

Fig 6 – Endemic brand portal user journey from audience selection through to reporting

The portal handled the full campaign lifecycle.

Audience segment selection from the shopper data foundation
Inventory selection across the store network
Creative upload and approval workflow
Campaign scheduling and pacing controls
Budget management
Post campaign reporting

The user experience was designed for the endemic buyer profile the client was starting with. Category marketing managers at appliance brands, television brands, mobile device brands, and computing hardware brands.

These buyers had digital media buying experience elsewhere but had not previously bought retail media at this level of platform sophistication. The portal prioritized three things. Clarity of the audience and inventory options available to the buyer. Transparent pricing that let the buyer model campaign investment against expected reach and frequency before commitment. Reporting that answered the questions those buyers would need to answer to their own internal stakeholders.

Inventory availability queries ran against a read through cache backed by the media platform’s booking service.

Direct queries against the booking database would have blown latency budgets during peak booking windows, since a single availability query across the store network and a full daypart calendar could produce large intermediate result sets.

The cache held pre computed availability slices at the store cluster, daypart, and format level, with expiry keyed to the booking transaction volume. Cache misses triggered a background refresh that updated the affected slice without blocking the user request.

Support for agency users was included from the first release. Endemic brands with agency partners for their broader media buying needed to grant portal access to those agencies under a role permission model backed by OpenID Connect for advertiser user authentication and OAuth 2.0 for API authorization. That let the brand retain approval authority while delegating operational execution.

Consulting Insight

The sequencing decision to start with endemic brands as the first buyer segment before opening the platform to non endemic advertisers through a future exchange integration was one of the most consequential calls in the engagement. The exchange integration would have doubled the technical scope of the foundation phase. It would have introduced brand safety and inventory quality questions that the client ad operations team was not yet staffed to answer, and delayed the revenue proof point by at least two quarters.

A new media platform earns credibility with known buyers before it earns credibility with anonymous ones.

The business outcome of this track for the client was concrete. Endemic brand advertisers received a buying surface they could operate against without depending on a client sales team phone call for every campaign, which materially lifted the number of campaigns the client could serve concurrently. The client sales team shifted from transactional booking work to advisory work on campaign strategy, which is a higher margin activity.

Attribution and Measurement Pipeline

Every brand advertiser buying retail media asked the same question. Did the campaign shift purchase behavior in a measurable way, and at what cost per incremental purchase. The regional retail media category at the time was answering that question inconsistently, with several competing networks running on measurement methodologies that ranged from opaque to actively misleading. This track gave the client a defensible answer at a level of transparency that endemic brand advertisers had been asking for and had not been getting from other retail media buys they were running.

The fourth track delivered the measurement infrastructure that let brand advertisers reconcile their retail media spend against downstream commercial outcomes.

Fig 7 – Attribution pipeline, exposure events joined against transaction streams to produce lift through geo holdout comparison

The pipeline tracked exposure events from the in store screen orchestration engine and the onsite ecommerce placement engine. It resolved those exposure events against the unified shopper identifier from the data foundation. It joined them against subsequent purchase events in the point of sale and ecommerce transaction streams to produce campaign level lift measurements.

Attribution methodology supported two reporting modes. Last touch reporting for operational campaign optimization. Incremental lift measurement for annual commercial planning.

Last touch reporting was available to endemic brand advertisers through the self serve portal within one business day of campaign completion. The pipeline ran a scheduled batch job overnight that aggregated the previous day’s exposure and transaction events, joined them at the shopper identifier level, and populated the reporting surfaces for morning access.

Incremental lift measurement required a holdout methodology that Clixlogix implemented through geo based holdouts at the store cluster level.

Individual shopper level randomization would have introduced consent complications, since exposing a shopper to a campaign creates a data processing event that requires a lawful basis under Malaysian PDPA, and randomly selecting shoppers for exposure without a defensible processing basis was not viable.

Geo holdouts sidestepped this by treating clusters of stores as the unit of randomization. Some store clusters ran the campaign, others did not, and the difference in shopper behavior between exposed and control clusters produced the lift measurement. Cluster level randomization better fit the operational and consent constraints of running lift studies at the client scale. Matched baseline selection ensured exposed and control clusters carried comparable pre campaign trajectories, which improved confidence in the exposed vs control comparison at the sample sizes available to a regional network.

The lift calculation ran against the attribution warehouse and produced both the point estimate and the confidence interval for advertiser reporting.

The pipeline surfaced results through advertiser facing reporting in the self serve portal, through account facing reporting to the client ad sales team, and through executive facing reporting to the client leadership. Each surface presented the same underlying measurement data at the level of aggregation and interpretation appropriate to the audience, which preserved measurement consistency across the organization.

Consulting Insight

The regional retail media landscape at the time the client entered the market was fragmented, with several competing networks running on measurement methodologies that ranged from opaque to actively misleading.

Clixlogix recommended the client take the opposite position and publish attribution methodology openly. Documentation of the geo holdout design, the confidence interval reporting, and the limits of the measurement was made available to endemic brand advertisers as part of the platform onboarding materials.

Endemic brand advertisers who had been burned by misleading measurement elsewhere responded to the transparency by shifting spend into the client network at rates that exceeded the initial commercial projections. In a category where competitors compete on measurement opacity, the first player to compete on measurement clarity captures disproportionate share of advertiser trust.

The business outcome of this track for the client was concrete. Endemic advertisers who had remained skeptical of retail media measurement generally moved committed spend into the client network specifically because the measurement posture cleared their internal diligence bar. Measurement transparency became a commercial differentiator with a defensible revenue impact.

Onsite and Offsite Audience Extension

Brand advertisers wanted one campaign that reached the same consented shopper cohort across the client’s physical store, the client’s ecommerce experience, and the shopper’s other digital surfaces on Meta and Google. Building three separate campaigns to reach the same cohort was operational overhead the brand marketing teams did not want to carry.

This track extended the unified shopper foundation to activate the same eligible audience cohort across those surfaces, with cohort level lift attribution reconciling outcomes back to the source campaign across the exposure surfaces the campaign ran on.

The fifth track extended the shopper data foundation to onsite ecommerce advertising placements and to offsite programmatic buying against Meta Marketing API, Google Ads API, and regional demand side platforms.

Fig 8 – Onsite and offsite audience extension, hashed identifier onboarding to Meta and Google plus regional DSP integrations

Onsite extension delivered sponsored product placements on the client ecommerce experience and category page banner inventory. Endemic brand advertisers could book these placements through the same self serve portal that governed the in store screen inventory, which produced a single buying interface for both physical and digital retail media.

Offsite extension activated shopper segments from the unified data foundation against advertiser targeting integrations at Meta and Google, and against regional demand side platforms for programmatic display and video buying.

The identity resolution work in the data foundation produced pseudonymized identifier lists that could be onboarded to the target platforms under privacy compliant matching. The onboarding process used SHA-256 hashing of email addresses and mobile advertising identifiers. Hashing is pseudonymization, and pseudonymized identifiers remain personal data under most modern privacy regimes including Malaysian PDPA. The hashing removed raw identifier exposure between the client and the platform while preserving the ability to match identities against target platform user graphs.

Matching happened through customer match style workflows on Meta and Google, and through data clean room integrations where the client and the target platform both used a supported clean room environment. Consent scope from the client foundation governed which shoppers could be included in each onboarding manifest, keeping the offsite extension aligned to Malaysian PDPA purpose limitation.

This let endemic brand advertisers extend a retail media campaign into offsite surfaces without breaking the audience continuity the platform was selling. A campaign built against a specific consented cohort could activate across in store screens, onsite ecommerce placements, and offsite Meta and Google buying, with cohort level lift measurement from the attribution pipeline reconciling outcomes across the exposure surfaces.

Aggregate and cohort level attribution was the correct grain for the offsite surfaces. Meta and Google exposure and conversion data returns at aggregate and cohort level through their measurement APIs, and the client platform did not attempt user level joins across surfaces where those joins were not supported by the source data.

The business outcome of this track for the client was concrete. Endemic brand advertisers could run a single campaign that reached the same eligible audience cohort across the surfaces the cohort actually used. That is the retail media proposition that separated the client from every competitor in the regional market who could only sell in store or only sell onsite.

Results

Within the first operational phase, the retail media network reached endemic brand commercial maturity and produced measurable outcomes across the commercial, operational, and strategic dimensions that mattered to the client leadership.

48% Media Revenue Growth in the First Quarter of Endemic Operation

A previously unmonetized physical store estate began generating measurable media revenue, with endemic brand spend running 48% ahead of the initial commercial projection through the first quarter of live operation.

14 Endemic Brands Active Across the Category Range

Endemic brand supplier participation reached 14 active advertisers within the first operational window, spanning appliances, televisions, mobile devices, and computing hardware.

Campaign Setup Time Reduced From Weeks to Hours

Endemic brand marketing managers moved from multi week trade marketing negotiation cycles to campaign setup measured in hours through the self serve portal, replacing email and phone booking with always on campaign management.

100+ Campaigns Served in the First Operational Month

The self serve portal handled more than 100 endemic brand campaigns in the first operational month, with the client sales team shifting from transactional booking work into advisory work on campaign strategy.

Attribution Transparency as a Commercial Differentiator

Published attribution methodology, including geo holdout design and confidence interval reporting, produced share of wallet gains against competing regional networks running on opaque measurement.

Onsite, Offsite, and In Store Extension Under One Interface

Onsite ecommerce placements, offsite Meta and Google audience buying, and in store screen inventory converged under a single buying interface with continuous audience attribution across surfaces.

Data Foundation Established for Downstream Use Cases

The unified shopper view opened downstream use cases beyond retail media, spanning merchandise planning, category expansion analytics, and personalization use cases the client leadership had not previously been able to pursue.

The engagement produced a durable strategic asset for the client beyond the direct commercial results. The data foundation, the ad operations discipline, and the measurement infrastructure all persist as capabilities that support use cases the client leadership continues to expand into. The retail media network itself is one commercial application of a broader data and platform investment that positions the client for the next generation of retail commerce competition.

The endemic brand phase produced sufficient operational maturity and commercial evidence for the client to authorize the next phase of the platform roadmap. The next phase covers non endemic advertiser onboarding through a programmatic exchange integration on OpenRTB 2.6, expansion of the in store screen inventory into additional store zones and additional format types, and deepening of the offsite audience extension motion into additional regional platforms. The engagement transitioned into a sustained partnership model that governs ongoing platform evolution alongside the client internal commerce and media teams.

Technologies and Tools

Category	Tools
Shopper Data Foundation	Segment (customer data platform for identity ingestion and event streaming), Snowflake (cloud data warehouse on medallion architecture with bronze, silver, gold zones), Apache Kafka via Confluent Cloud (event streaming ingestion pipeline with schema registry and idempotent producers), Hightouch (reverse ETL for audience activation from warehouse to advertiser platforms)
Identity and Consent	LiveRamp (deterministic identity graph and hashed identifier resolution), OneTrust (consent management platform with versioned scope tracking aligned to Malaysian PDPA principles), custom consent registry service (query time scope enforcement upstream of audience segment build)
Media Orchestration	Broadsign (digital out of home ad server for in store screen inventory scheduling and playback), Kevel (ad decisioning engine with in memory cache for operational decisioning on precomputed playlists), custom creative management and approval workflow with brand safety and competitive separation rule engine
Brand Self Serve Portal	React with Next.js (web application front end), NestJS on Node.js (backend API serving audience selection, inventory booking, creative upload, campaign management, reporting), Redis (read through cache for inventory availability at store cluster and daypart level)
Attribution and Measurement	Snowflake (attribution warehouse holding exposure and transaction join tables), custom incremental lift service (geo holdout methodology at store cluster level with confidence interval reporting), Looker (advertiser and executive reporting surfaces), published methodology documentation for endemic buyer onboarding
Audience Extension	Meta Marketing API and Google Ads API (customer match style audience onboarding through hashed identifiers), LiveRamp Safe Haven (data clean room environment for privacy compliant matching against target platform user graphs), The Trade Desk and regional DSPs (programmatic display and video buying for offsite extension), custom onsite ad placement engine embedded in client ecommerce experience
Infrastructure and Delivery	AWS as underlying cloud, Amazon EKS (containerized microservices for platform components), GitHub Actions (CI CD pipelines with automated test gating), Datadog (observability with distributed tracing across CDP, ad decisioning, and attribution services), PagerDuty (incident response with runbook automation)
Authentication and Authorization	`OpenID Connect` for advertiser user authentication, `OAuth 2.0` for API authorization including B2B API access, TLS 1.3 for data in transit, Stripe (advertiser invoicing and revenue recognition for the media billing surface)
Standards and Compliance	`OpenRTB 2.6` for programmatic bidding when the exchange integration goes live, Malaysian Personal Data Protection Act 2010 as amended by the 2024 Amendment Act (notice and choice, purpose limitation, retention, data minimization), `SHA-256` for identifier pseudonymization at offsite onboarding, ISO 27001 aligned controls for data governance and data breach notification

Services Delivered

Digital Engineering, Consulting Service, Retail Media Strategy, Data Engineering, Identity Resolution, Consent and Privacy Governance, Ad Serving Engineering, Measurement and Attribution Engineering, DevOps, Backend Engineering, QA

Team Composition

Retail Media Strategy Lead, Data Architect, Media Platform Engineers, Brand Portal Engineers, Attribution Engineer, In Store Deployment Coordinators, Ad Operations Specialist

Have a question for our team or need help with your project?

Our team can share client references, scope your project, and answer any question about your delivery.

Related Case Studies

More engagements where our delivery teams shipped similar outcomes for clients across industries. Read on for context on the patterns we reused, the trade offs we navigated, and the metrics that landed in production.

Statewide EV charging platform engineering for a Florida network operator

EV Charging Platform Engineering with Route Optimization for a Florida Operator

Mobile App Development Digital Engineering Devops IT strategy consulting IoT Automotive & Mobility IT Consulting

Voice AI, Prompts, and Memory for a Real Estate Platform

Generative AI & ML AI Agent Real Estate AI

View All Case Studies

Retail Media Network for a Southeast Asian Electronics Chain

Executive Summary

About the Client

The Challenge

Internal skepticism about another store technology bet

Shopper data trapped in silos

Endemic brand relationships anchored in annual trade marketing

The Approach

The Solution

Shopper Data Unification

In Store Screen Orchestration

Endemic Brand Self Serve Portal

Attribution and Measurement Pipeline

Onsite and Offsite Audience Extension

Results

48% Media Revenue Growth in the First Quarter of Endemic Operation

14 Endemic Brands Active Across the Category Range

Campaign Setup Time Reduced From Weeks to Hours

100+ Campaigns Served in the First Operational Month

Attribution Transparency as a Commercial Differentiator

Onsite, Offsite, and In Store Extension Under One Interface

Data Foundation Established for Downstream Use Cases

Technologies and Tools

Have a question for our team or need help with your project?

Related Case Studies

EV Charging Platform Engineering with Route Optimization for a Florida Operator

Voice AI, Prompts, and Memory for a Real Estate Platform

About

Services

Solutions

Resources