<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Tesco Technology on Medium]]></title>
        <description><![CDATA[Stories by Tesco Technology on Medium]]></description>
        <link>https://medium.com/@tescotechnologycommunications?source=rss-2ce0c24b1b1b------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*SaxB6zrkF4W9km_C-ZG9yA.png</url>
            <title>Stories by Tesco Technology on Medium</title>
            <link>https://medium.com/@tescotechnologycommunications?source=rss-2ce0c24b1b1b------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 09:40:33 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@tescotechnologycommunications/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[From Monthly Releases to Continuous Deployment]]></title>
            <link>https://medium.com/@tescotechnologycommunications/from-monthly-releases-to-continuous-deployment-d67829640c71?source=rss-2ce0c24b1b1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/d67829640c71</guid>
            <dc:creator><![CDATA[Tesco Technology]]></dc:creator>
            <pubDate>Wed, 14 Jan 2026 08:21:07 GMT</pubDate>
            <atom:updated>2026-01-14T08:21:07.943Z</atom:updated>
            <content:encoded><![CDATA[<p><strong><em>By Deniz Vehbi, Head of Software Development, Tesco Technology</em></strong></p><p>The <strong>One Account</strong> team owns the web frontends for Tesco Online’s onboarding and account management journeys. Until a few years ago, the team made <strong>one production deployment per month</strong>, which took <strong>three hours to complete</strong> and had a <strong>high failure rate</strong>, often resulting in rollbacks. At that time, the applications were <strong>only available in the UK</strong>, and only a <strong>subset of engineers could make production deployments</strong>, with some only able to deploy specific applications.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/925/1*tX0WDlWGxaTq9JKPmDNeZw.png" /></figure><p>How did we achieve this transformation and subsequently go beyond by implementing continuous deployment? Along the way, several steps were taken, some of which are described below.</p><p><strong>Technical Changes</strong></p><ul><li><strong>Simplified the architecture</strong> into smaller, independently deployable micro‑apps, with reusable patterns and pipelines</li><li><strong>Testing depth and breadth</strong>: layered automated testing (unit → integration → end-to-end, plus accessibility and performance) and business-readable specs using Gherkin.</li><li><strong>DevOps Mindset:</strong> same engineers build, test, deploy, monitor, and run the services, including the Azure infrastructure.</li><li><strong>Adaptive Observability</strong>: moved away from noisy, false positive alerts to adaptive, ratio-based signals that reflect seasonality and market differences.</li><li><strong>Customer‑centric Observability</strong>: introduced journey level impact reporting for incidents, so conversations focus on journey outcomes, not just raw error counts.</li><li><strong>Everything(almost) as Code</strong>: infrastructure, dashboards and alerting all converted to code.</li></ul><p><strong>Cultural Changes</strong></p><p>· <strong>Developer Focus time: </strong>limited the number and duration of ceremonies and scheduled meetings to specific days of the week to help create focus time for development.</p><p>· <strong>Tackling skill gaps: </strong>used a skills matrix to identify areas of knowledge silos and focused on learning and knowledge sharing on areas with frequent gaps.</p><p>· <strong>Shared Domain Knowledge: </strong>introduced sub-team rotations, light-weight documentation and team knowledge sharing sessions.</p><p>· <strong>Agile hygiene: </strong>focused on having engaging ceremonies, accurate sprint boards and using agile delivery metrics to guide improvements.</p><p>· <strong>Continuous improvement mentality: </strong>supported by the team’s Tech Radar and Kaizen days with their own board and prioritisation process.</p><p>· <strong>All underpinned by greater psychological safety: </strong>where team members feel able to ask for help when needed and confidently propose new ideas.</p><p><strong>Setting Continuous Deployment as a North Star goal</strong></p><p>After establishing <strong>fast deployments</strong>, <strong>shared domain knowledge</strong>, <strong>simpler infrastructure</strong>, <strong>good agile practices</strong>, and <strong>deep and broad test coverage</strong>, we set an even more ambitious goal: doing continuous deployment, meaning each commit to main is going to production without manual testing. To do continuous deployment, you need very extensive automated testing coverage, coupled with precision monitoring to swiftly auto-rollback if any post-deployment anomaly is detected.</p><p><strong>Why did we have Continuous Deployment as a North Star?</strong></p><p>The journey towards continuous deployment readiness would add value to many teams as it leads to questioning of the <strong>breadth and depth of testing</strong>, the <strong>coverage of alerting</strong>, the <strong>quality of monitoring</strong> and the <strong>deployment, rollback durations</strong> for any given application. Even if a team ultimately ends up not enabling continuous deployment, the <strong>journey towards CD readiness</strong> helps the team mature its testing, monitoring, deployment and rollback mechanisms further and thus increase its change frequency while lowering failure rate.</p><p><strong>Did we manage to implement Continuous Deployment?</strong></p><p>In 2024, after months of test coverage improvements, alerting fine-tuning, testing of the canary deployment and rollback process, we enabled continuous deployment for one of our critical applications. Since then, over a year and a half, the application has been deployed hundreds of times without any manual verification after deployment, <strong>proving the feasibility of doing continuous deployment at scale for critical applications</strong>.</p><p><strong>In Conclusion</strong></p><p>I’m proud of what we’ve achieved over the last few years and moved from serving one country to all five Tesco markets, from 0.2 deployments per day to more than 5 deployments per day, from 3-hour deployments to 5–10 minute deployments and from a low morale, low support environment to a high morale, supportive environment where engineering is seen as a team activity where the team wins together.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d67829640c71" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Teaching Legacy Code to Speak: How GenAI Unlocks Modernisation]]></title>
            <link>https://medium.com/@tescotechnologycommunications/teaching-legacy-code-to-speak-how-genai-unlocks-modernisation-e5825dfa3b36?source=rss-2ce0c24b1b1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/e5825dfa3b36</guid>
            <dc:creator><![CDATA[Tesco Technology]]></dc:creator>
            <pubDate>Thu, 20 Nov 2025 12:50:32 GMT</pubDate>
            <atom:updated>2025-11-20T12:50:32.983Z</atom:updated>
            <content:encoded><![CDATA[<p><em>Written by Arshad Hameed, Software Dev Engineer-III, Tesco Technology</em></p><p><strong>We taught a 40-year-old COBOL system to speak.</strong></p><p>Modernisation didn’t stall on compute — it stalled on comprehension. So, we made the code the source of truth and used GenAI to turn tens of thousands of lines into plain-English narratives, flowcharts, and even a chatbot. Engineers get faster onboarding, PMs get clarity on business logic, and leaders get decisions grounded in reality — not guesswork.</p><p>The payoff? Weeks saved per program and a foundation to accelerate modernisation beyond understanding.</p><p><strong>The First Mile of Modernisation: Making Sense of the Past</strong></p><p>“Modernise the warehouse management system.” On paper, it sounded like a platform shift. In reality, the hard part wasn’t Kubernetes or cloud networking — it was comprehension. The system we inherited had been alive for four decades, expanded by many hands, and documented by few. What documentation existed lagged production, and the few SMEs who remembered the deeper seams were already busy keeping the lights on. Understanding — not compute — was the blocker. So, we placed a bet: treat the code as the only non fiction in the room and use GenAI to make that code speak. Not as snippets, not as screenshots, but as plain English explanations, execution flowcharts, and a searchable knowledge base that anyone — from an SDE ramping up to leadership making a sequencing call — could trust. Phase 1 of the journey was simple to say and hard to do: turn legacy COBOL into living, validated knowledge.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*eX_rzADnhwz_Nq00RduMvg.png" /></figure><p><strong>Our thesis: the source of truth is the source code</strong></p><p>COBOL programs in this WMS routinely reached 10–30K lines, with copybooks silently ballooning effective size; months could evaporate just building a reliable mental model of a single large program. We stopped chasing perfect docs and omniscient SMEs. Instead, we began with what actually runs in production: code. The job for GenAI was to translate that code into narratives, execution flows, and query-able facts — then invite SMEs to validate the few things only humans can: domain entities, abbreviations, and semantics. This shifted SME time from endless walkthroughs to one-time validations that compound.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*2WeCERgMQNXZJ2x7o8HRsw.png" /></figure><p><strong>Making the code speak: the pipeline we built</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*2WJbN-dHQM51Jl1IidBP3g.png" /></figure><p>We started at the syntax level and climbed toward meaning. COBOL was parsed into an Abstract Syntax Tree (AST) using Proleap and projected into a Knowledge Graph in Neo4j. We modelled Paragraphs — think of them as functions in modern languages — as primary nodes, with Sections and Divisions as containers. Edges captured relationships like contains, uses, and calls (including PERFORM, GOTO, and CALL). In simpler terms, we broke each program into connected chunks: small blocks grouped into larger ones, linked by control flow and dependencies. This let us explain leaf nodes first, then roll up to program-level behaviour without losing the plot. Next, we tackled domain meaning. COBOL’s naming constraints encouraged cryptic abbreviations, so we extracted entities, deduplicated them, and had SMEs validate once for reuse.</p><p>Validated entities and embeddings went into Qdrant, forming a semantic backbone for explanations and queries. Execution flow was another challenge. Chunked summaries drifted without an anchor, so we generated Mermaid-based flowcharts at the Section level and stitched them into program-level flows. These were fed back into prompts, keeping narratives aligned with real control paths — including PERFORMs, GOTOs, and CICS events like EXEC CICS START (COBOL’s way of jumping to another part of the program or another program).</p><p>Finally, we made it discoverable. Explanations and entities were indexed in Qdrant, layered with <a href="https://github.com/HKUDS/LightRAG">LightRAG</a> for precise retrieval. The result: a chatbot where you can start with “What does this program do?” and drill into specifics like “How is inventory reallocation triggered under exception X?” — with links back to code and flows.</p><p><em>New to COBOL? Check out this quick </em><a href="https://www.geeksforgeeks.org/cobol/cobol-basic-syntax/"><em>guide</em></a><em> for program structure and key control flow statements like PERFORM and GO TO.</em></p><p><strong>Architecture at a Glance</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/940/1*F9QgDxRIp28tVvU71HR9NQ.png" /></figure><p><strong>What changed: early signals you can feel (and measure)</strong></p><p>The most important effect was social: shared understanding arrived sooner. New contributors could orient on a program with a narrative and flowchart instead of spelunking through copybooks. SMEs were no longer bottlenecks for basic comprehension; their time moved to high leverage validation of domain semantics. And for leaders, decisions started resting on verifiable behaviour, not guesswork. We also measured it. Across programs with 10K+ LOC, we reached 80–85% knowledge coverage, cut at least 15 days per program from requirements discovery, and got a full explanation generated in under a day including manual review. Those numbers aren’t a finish line; they’re proof that translating code into knowledge can move the critical path.</p><p><strong>Challenges That Shaped the Solution</strong></p><p>We hit a scale wall on very large programs — ~20K+ LOC with copybooks — where even generous output limits couldn’t coax complete, single shot narratives. We tried context batching and the Review Agent; they helped a bit, but not enough. Our next step is modular documentation: a main page with linked subpages to capture complex behaviour without losing readers in a single monolith. Control flow complexity was another culprit. Chunked explanations can drift from reality unless you explicitly encode the flow. Generating execution flow separately and injecting it into the narrative context kept explanations grounded in what actually runs. Entity extraction initially buried SMEs under candidate terms. Unigram splitting and cross program deduplication cut the review list down to the set that compounds, and the burden continues to shrink as coverage grows. We also learned to ignore comments — validating them can be a project of its own — and instead trusted code plus SME validated entities.</p><p>Search introduced its own edge cases. Broad words can pull too many documents. Relevance thresholds, document limits, and <a href="https://github.com/HKUDS/LightRAG">LightRAG</a> turned the firehose into a focused stream. There’s more tuning to do, but the retrieval now feels like a conversation, not a treasure hunt.</p><p><strong>How different roles use it</strong></p><p>For an engineer, the experience starts with the flowchart: where does control enter, what conditions branch, which paragraphs do real work? The accompanying narrative points to paragraphs and sections by name, so it’s easy to jump to the COBOL and verify. For a PM, the story surfaces business meaning: why a reallocation happens, where exceptions are handled, which data elements carry which decisions. For a leader, the value is in the map: you can see integration points, system boundaries, and the parts of the codebase that carry the most business risk — enough to sequence modernisation with eyes open. The crucial part is that everyone is reading the same source anchored truth.</p><p><strong>The door we just unlocked</strong></p><p>Phase 1 was about understanding. But once the code speaks, new possibilities appear. A knowledge base enriched with validated entities and real execution paths can map business capabilities, seed acceptance tests, highlight integration seams, and even guide AI assisted code generation with traceability back to the original COBOL. That’s the leapfrog: not just preserving legacy behaviour but using its truth to accelerate what comes next.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/505/1*17klkiKqUnk9OPjz7Q2cJQ.png" /></figure><p><strong>Epilogue: a system that can finally explain itself</strong></p><p>The quiet victory here isn’t a tool or a model. It’s the change in posture — from manual discovery and scattered lore to assisted comprehension and shared reality. In a 40 year old WMS, that shift is everything. The code didn’t get simpler. It just started telling the truth in a way humans can use. And that’s how modernisation stops being a guessing game and starts being a plan.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e5825dfa3b36" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How GS1 standards unlock innovation at Tesco scale]]></title>
            <link>https://medium.com/@tescotechnologycommunications/how-gs1-standards-unlock-innovation-at-tesco-scale-a69911dcc115?source=rss-2ce0c24b1b1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/a69911dcc115</guid>
            <dc:creator><![CDATA[Tesco Technology]]></dc:creator>
            <pubDate>Thu, 16 Oct 2025 12:31:35 GMT</pubDate>
            <atom:updated>2025-10-16T12:31:35.466Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*HMlgFiu9tBhVwIOWdPrXoA.png" /></figure><p>By Mark Barnes, Principal Engineer and Richard McKeating, Enterprise Architect, Tesco Technology</p><p><strong>What do we mean by Tesco scale: the complexity we manage</strong></p><p>Tesco is big. With millions of customers, hundreds of thousands of colleagues, and thousands of stores and distribution centres, even the smallest change can have a big impact. Rolling out innovation at this scale isn’t just about having a good idea; it’s about making it work across a vast, complex ecosystem. And that’s before you factor in rising regulatory pressures and ever-increasing expectations from customers and stakeholders alike. Let’s explore what it really means to deliver change in one of the world’s largest retailers, and why doing so successfully requires not just technical excellence, but a deep understanding of scale, impact, and the delicate balance between speed and stability.</p><p>Managing compliance and traceability at Tesco’s scale is not a simple task. Regulations such as those governing sustainability and product provenance require us to maintain accurate, auditable data across a vast network of suppliers, systems and processes. This isn’t just about capturing information; it’s about ensuring that data flows seamlessly through our ecosystem, supporting risk management and reporting without disrupting day to day operations. At this scale, even a single product can involve multiple data points and partners, which is why standards and interoperability are critical to making compliance achievable and efficient.</p><p>Beyond compliance, the sheer volume of data we manage every day adds another layer of complexity. We have the transactional data needed to allow you to get the latest prices at the tills and pay for your shopping, together with the 10s of thousands of other shoppers also checking out at the tills at the same time. We have the operational data behind the scenes to make sure that toilet roll you just purchased will be replenished in the background for the next shopper and will be ordered from the suppliers. We also have the analytical data to support the business in forecasting demand, running experiments and measuring the key performance indicators to ensure the business is operating well.</p><p>To support the business aspirations for innovation, we are continually trying to modernise our technology platforms, simplify our operations and ensure we deliver the outstanding customer experience that has kept<strong> </strong>Tesco the number one supermarket in the UK for 30 years. We must adopt a mindset of incremental improvements while ensuring all our systems, old and new, developed or bought, continue to talk to each other. And our data volumes above indicate why this is a challenge, a single distribution centre will generate millions of events per day as products move through the operation. If we start adding in the rest of the supply chain internationally and the final delivery to stores and customers, data standards are not a luxury, they are a critical success factor and the only way to ensure an operation the size of Tesco can keep functioning efficiently.</p><blockquote><strong>Tesco’s proud history of innovation around standards</strong></blockquote><p>At Tesco’s scale, it might seem tempting to think we can create our own rules; in reality, success depends on working within global standards. Our suppliers serve multiple retailers worldwide, and the branded chocolate bar on a Tesco shelf is the same one you’ll find anywhere else. Where Tesco makes a difference is in using its scale to accelerate the adoption of those standards and innovate on top of them. A great example is our initiative to introduce Outer Case Codes (OCC) on product cases. By championing this GS1 standard across our supply chain, we improved traceability and efficiency for millions of cases moving through our distribution network, while also setting a precedent that benefited the wider industry. This approach reflects our philosophy: standards are not constraints; they are enablers of innovation at scale. The result? Fewer errors, fresher products, faster replenishment, and ultimately better value for customers through lower costs and reduced waste.</p><p>That philosophy has shaped Tesco’s long history of innovation around standards. From the launch of Clubcard in the 1990s, which accelerated the industry’s shift to data‑driven loyalty, to the early scaling of online grocery and Click+Collect, our progress has always relied on consistent product and location data. In stores, Scan as You Shop put barcode scanning directly in customers’ hands, transforming the checkout experience. To support stock control for our F&amp;F clothing range, we introduced RFID scanning technology into our stores and distribution centres<strong>. </strong>More recently, we’ve piloted GS1‑powered 2D barcodes on selected lines in the UK to improve date‑code accuracy and provide richer product information. Each of these milestones demonstrates how Tesco uses its scale to make standards practical, drive adoption across the supply chain, and unlock new capabilities for colleagues and customers alike.</p><p>Those programmes succeeded because the information behind the experiences could be exchanged reliably. Barcodes and common identifiers were the first step. Here we leveraged GS1 identifier standards to ensure consistency across the supply chain. We embraced electronic interchange between third parties through the adoption of the EDI standard to get shipping advice, delivery advice and provide receipts. The next step in this journey is the adoption of EPCIS and CBV, which provide a shared language to track the movement of products through the entire supply chain. This promises a future of a machine-readable, unambiguous language that will scale across our partners.</p><blockquote><strong>Why GS1, Why EPCIS, Why Now?</strong></blockquote><p>For decades, Tesco has relied on mainframes and batch processes to manage the enormous data volumes behind our operations. That approach worked when overnight updates were good enough, but today’s retail environment demands something different. Customers expect real-time availability, regulators require instant traceability, and our supply chain needs to respond dynamically to disruptions. This shift from batch-based to event-driven operations is more than a technical upgrade; it is a fundamental change in how data flows across thousands of systems and partners. EPCIS provides the standard to make that possible, defining events in a consistent, machine-readable way and pairing them with a consistent business vocabulary through CBV. Together, they create the foundation for interoperability, compliance and innovation at scale.</p><p>EPCIS offers more than just a technical upgrade:</p><p>Facilitates the transition from batch to real-time.</p><p>EPCIS provides a structured way to capture and share events as they happen, whether that is a pallet leaving a supplier, a case arriving at a distribution centre, or a product being sold at checkout.</p><p>Core Business Vocabulary.</p><p>CBV standardises terms such as shipping, receiving, commission and aggregate, reducing ambiguity and integration errors across thousands of systems and partners.</p><p>Modern, developer-friendly formats.</p><p>EPCIS 2.0 supports JSON, making it easier for engineers to adopt and integrate using modern APIs and event-driven architectures.</p><p>Regulatory compliance made simpler.</p><p>From food safety and pharmacy regulations to sustainability, EPCIS provides the traceability backbone needed to prove provenance, manage recalls and maintain audit readiness.</p><p>A stabilising integration layer.</p><p>EPCIS acts as a decoupling layer for legacy systems, allowing them to publish and consume standardised events without requiring every system to migrate in lockstep. Anti-corruption layers and lightweight adapters make integration cost-effective.</p><p>Faster onboarding of partners and solutions.</p><p>When third-party solutions are GS1/EPCIS-compliant, integration costs drop dramatically. Instead of bespoke mappings, we can agree on a Tesco capture profile and start exchanging data quickly.</p><p>Why now? Because the future of retail is real time, and EPCIS is one of the standards that makes it possible.</p><p><strong>A high‑level example: data events captured by different parties</strong></p><p>GS1 have created a beautiful interactive animation of the different GS1 standards and where they apply in the supply chain for different verticals at <a href="https://discover.gs1.org/">https://discover.gs1.org/</a> which is well worth visiting to understand how different GS1 identifiers are used across the industry.</p><p>Let’s follow the journey of a batch of Tesco Beef Burgers, starting at the supplier’s production line and ending in a customer’s shopping basket. Along the way, a series of digital events quietly unfold, each one capturing a moment in the product’s lifecycle. It all begins on the farm where the cattle are reared and the desire for traceability to support our quality and welfare standards.</p><p><strong>At the Supplier</strong></p><p>The digital journey begins at a meat processing plant, where the burgers are freshly produced. As each batch is made, the supplier assigns a unique digital identity; not just a barcode, but a rich identifier that includes the product type, batch number and use-by date. This moment is captured as a commissioning event, marking the product’s entry into the digital supply chain. Associated with this event will be links to certifications for the animals.</p><p>Once the burgers are packed into cases, they’re stacked onto pallets ready for Tesco. This packing process isn’t just physical; it’s digitally recorded too. The cases are linked to their pallet using an aggregation event, creating a traceable relationship between individual packs and the larger shipment.</p><p>When the pallet is ready to leave, a shipping event is triggered. This records the fact that the burgers are on the move, linking the shipment to the original purchase order and the despatch advice Tesco receives. The burgers are now officially in transit, and the digital thread continues.</p><p><strong>On the Road</strong></p><p>Whether handled by the supplier’s own fleet or a third-party logistics provider, the journey to Tesco’s distribution centre creates additional events. These could include trailer loading, hub arrivals or temperature readings, all captured as transport-related events, enriching the product’s story with real-world context.</p><p><strong>At the Tesco Distribution Centre</strong></p><p>As the pallet rolls into a Tesco DC, a receiving event is generated. This confirms the arrival and updates the product’s status from “in transit” to “active”. It’s matched against the dispatch advice and receiving documents, ensuring everything lines up.</p><p>Later, when store orders come in, the burgers are picked from pallets and loaded into roll cages and then onto trailers. These movements are captured through aggregation and shipping events, linking the burgers to their next destination, the store. Temperature sensors in the DC record conditions during handling, adding another layer of assurance.</p><p><strong>On the Way to Stores</strong></p><p>Tesco’s transport fleet plays a key role here. As the burgers travel to stores, trailer telemetry captures temperature data and location updates. These are recorded as shipping events, maintaining visibility and ensuring cold chain compliance. The store order remains linked throughout, tying the journey back to the original demand.</p><p><strong>At the Store</strong></p><p>When the roll cage arrives at the back of the store, a receiving event marks the burgers’ arrival. They’re now ready for stocking and sale.</p><p>Finally, when a customer picks up a pack and scans it at the checkout, a selling event is triggered. Since the pack carries a 2D barcode, the system captures not just the product type but also the batch and use-by date, closing the loop on a journey that possibly began less than 24 hours earlier at the meat processing plant.</p><p>This event chain forms a digital thread for each product. It supports day‑to‑day operations (availability, waste, substitutions) and strategic outcomes (sustainability evidence, digital product passports, customer transparency). For our customers, it guarantees you are getting a great quality product that has been properly processed and handled.</p><h3>Looking beyond: what wider adoption unlocks</h3><p>Adopting GS1 and EPCIS standards is not just about solving today’s challenges; it is about creating a foundation for the future. As more partners align on common data standards, the benefits compound, unlocking new capabilities that go far beyond compliance and operational efficiency.</p><p><strong>End-to-end inventory visibility</strong></p><p>When suppliers, carriers, distribution centres and stores publish consistent ship, receive, stock and sell events, we can build near real-time views of inventory across the network. This improves exception handling, reduces out-of-stocks and ensures better on-shelf availability. The benefits multiply as more partners participate.</p><p><strong>Waste and date-code intelligence</strong></p><p>Next-generation GS1-powered 2D codes can carry dynamic data such as batch and expiry date. Combined with EPCIS event history, this enables precise markdowns, do-not-sell controls and targeted recalls. For perishable categories, even a small improvement in date accuracy can deliver significant waste reductions and cost savings.</p><p><strong>Supplier onboarding at speed</strong></p><p>Instead of bespoke EDI dialects and unique field mappings for each partner, we can agree a Tesco capture profile for EPCIS events, GS1 label requirements and a minimal API test suite. This shortens the time from first meeting to first delivery and makes continuous improvement measurable.</p><p><strong>Customer transparency</strong></p><p>Customers increasingly expect to know where products come from, how they were made and how to dispose of the packaging. Standards make those answers portable: scan a code and see provenance, allergens and recycling guidance presented consistently across channels.</p><p><strong>AI and IoT enablement</strong></p><p>AI thrives on clean, granular, time-stamped events. EPCIS provides forecasting and optimisation models with the signal they need without heavy data wrangling. IoT sensors can attach temperature or shock readings to logistics events, giving cold-chain assurance and improving quality control.</p><p><strong>Sustainability and fleet transformation</strong></p><p>As we continue to electrify home-delivery vans and pilot heavy-duty electric vehicles, event data that includes route, asset and energy context becomes the evidence base for decarbonisation KPIs. Standards reduce friction between fleet systems, telematics and supply-chain platforms, accelerating progress towards net zero.</p><p>In short, wider adoption of GS1 and EPCIS standards is not just an IT upgrade; it is a strategic enabler that drives efficiency, transparency and sustainability, delivering better outcomes for customers, communities and the planet.</p><h3>Closing thoughts</h3><p>Modernising a large, complex technology estate is never a simple task. This is not the typical challenge of upgrading a single legacy product to a newer version; the real challenge at Tesco scale lies less in modernising individual systems and more in modernising the operations, the interactions and the data that flow across them. It is about creating a connected ecosystem where old and new can coexist without disruption.</p><p>At Tesco, we use technology as an enabler to serve our customers, communities and planet a little better every day.</p><p>If solving challenges like these excites you, why not join us and help us shape the Tesco for tomorrow.</p><h3>References</h3><p>- Tesco history: <a href="https://www.tescoplc.com/about/our-history/">https://www.tescoplc.com/about/our-history/</a></p><p>- Scan as you Shop: <a href="https://www.tescoplc.com/tesco-rolls-out-scan-as-you-shop-service-in-more-stores/">https://www.tescoplc.com/tesco-rolls-out-scan-as-you-shop-service-in-more-stores/</a></p><p>- Tesco introduces Outer Case Codes: <a href="https://www.gs1my.org/?q=tesco-outer-carton-case-occ-project">https://www.gs1my.org/?q=tesco-outer-carton-case-occ-project</a></p><p>- Tesco introduces RFID for F&amp;F: <a href="https://www.gs1uk.org/insights/news/ff-clothing-at-tesco-rolls-out-rfid-technology">https://www.gs1uk.org/insights/news/ff-clothing-at-tesco-rolls-out-rfid-technology</a></p><p>- Discover GS1 — An Interactive GS1 Standards App — <a href="https://discover.gs1.org/">https://discover.gs1.org/</a></p><p>- GS1 Identification Keys (SSCC, GTIN, etc): <a href="https://www.gs1.org/standards/id-keys">https://www.gs1.org/standards/id-keys</a></p><p>- GS1 Application Identifiers: <a href="https://ref.gs1.org/ai/">https://ref.gs1.org/ai/</a></p><p>- GS1 EPCIS &amp; CBV Standards: <a href="https://www.gs1.org/standards/epcis">https://www.gs1.org/standards/epcis</a></p><p>- GS1 EPCIS Sandbox: <a href="https://epcis-sandbox.gs1.org/">https://epcis-sandbox.gs1.org/</a></p><p>- GS1‑powered QR trials (industry news): <a href="https://www.gs1uk.org/insights/news/Tesco-in-early-trials-of-next-generation-barcodes">https://www.gs1uk.org/insights/news/Tesco-in-early-trials-of-next-generation-barcodes</a></p><p>- Packaging Europe coverage — QR trials: <a href="https://packagingeurope.com/news/tesco-trials-gs1-qr-codes-in-select-uk-stores/12806.article">https://packagingeurope.com/news/tesco-trials-gs1-qr-codes-in-select-uk-stores/12806.article</a></p><p>- EU General Food Law — traceability (EC 178/2002): <a href="https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32002R0178">https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32002R0178</a></p><p>- Digital Product Passports under ESPR: <a href="https://environment.ec.europa.eu/topics/circular-economy/product-passports_en">https://environment.ec.europa.eu/topics/circular-economy/product-passports_en</a></p><p>- Tesco Careers: <a href="https://careers.tesco.com/">https://careers.tesco.com/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a69911dcc115" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Transforming Shopping at Tesco: The Tech Revolution]]></title>
            <link>https://medium.com/@tescotechnologycommunications/transforming-shopping-at-tesco-the-tech-revolution-13055b407fff?source=rss-2ce0c24b1b1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/13055b407fff</guid>
            <category><![CDATA[supply-chain]]></category>
            <dc:creator><![CDATA[Tesco Technology]]></dc:creator>
            <pubDate>Wed, 04 Jun 2025 10:16:16 GMT</pubDate>
            <atom:updated>2025-06-04T10:18:26.804Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*URjc3-b0n5bkiEbIYDhMng.png" /></figure><p>By Priyadarsanie Ramasubramanian, Director — Engineering, Technology — Supply Chain and Abhishek Tripathi, Head of Software Development, Tesco Technology</p><p>Once upon a time, shopping was a mundane chore. Then, technology arrived, flipping it on its head and transforming the ordinary into the extraordinary, both online and offline. Imagine a world where e-commerce claims more than 25% of global retail sales, with customers spending a staggering 90 billion hours on shopping apps. In this new reality, Tesco stands at the forefront, crafting a shopping journey so personalised, convenient, and enjoyable, it feels like having a personal shopper in your pocket. Our motto? Shop for ‘Anything, anywhere, anytime.’</p><p>Let me take you on a journey through the key areas where technology has revolutionized the way we shop.</p><p><strong>The ‘Phygital’ Approach</strong></p><p>Picture an ‘Omnichannel’ blend of offline and online shopping, where boundaries blur and convenience reigns. Imagine using augmented reality (AR) and virtual reality (VR) on your phone to see how a couch would look in your living room. Meanwhile, our physical stores have transformed into digital playgrounds. In just five years, self-checkout kiosks have surged by 300%. You routinely tap your phone or scan a QR code to pay, earn loyalty points, or receive personalized offers. Virtual assistants guide you to the right aisle, making shopping feel like a high-tech treasure hunt.</p><p><strong>Personalisation</strong></p><p>In this new era, personalisation is key. At Tesco, we understand that each customer is unique, and so is their shopping journey. Say goodbye to one-size-fits-all recommendations! Our sophisticated AI delves into your purchase history and browsing habits, suggesting items tailored to your preferences, not just popular choices.</p><p><strong>Immersive Experiences with AR and VR</strong></p><p>AR and VR are the dressing rooms of the future. Try on clothes, accessories, and makeup virtually. Explore virtual stores from your couch. Retailers create virtual showrooms that mimic the in-store experience, complete with interactive displays and virtual assistants. It’s shopping, but with a sci-fi twist.</p><p><strong>Voice Assistants and AI Chatbots</strong></p><p>Voice assistants like Alexa, Siri, and Google Assistant are turning your kitchen into a command center. Need groceries while baking a cake? Just ask. AI chatbots are stepping up their game too, moving beyond scripted responses to engaging in meaningful chats. They handle the repetitive tasks, letting human agents tackle the trickier issues. Deploying them also reduces the need for a large customer service team, leading to significant cost savings.</p><p><strong>Frictionless Checkout</strong></p><p>Imagine walking into a Tesco store, grabbing what you need, and walking out without waiting in line. With ‘Get-Go’, our frictionless checkout system, it’s like magic. Real-time image processing and sensors track your items, creating a virtual cart and invoicing you automatically as you leave. No lines, no waiting, just pure convenience.</p><p><strong>Social Commerce</strong></p><p>Social commerce is exploding, merging social media with e-commerce. Platforms like Instagram and Facebook let you shop from friends and influencers. Live shopping events enable brands to host product reviews, answer questions, and sell directly from the stream. Shopping apps allow users to review products in creative ways, making reviews more authentic and relatable.</p><p><strong>Supply Chain Revolution</strong></p><p>The tech transformation isn’t just customer-facing. IoT-enabled smart shelves track inventory in real-time and alert managers when stock is low. These systems optimize stock levels, reducing shortages and overstock. Robots in warehouses handle picking, packing, and shipping, making operations faster and more cost-effective.</p><p><strong>Sustainable Shopping</strong></p><p>Tesco is committed to being carbon neutral in our own operations by 2035 and net zero across our whole footprint by 2050. The retail sector accounts for about 5% of global greenhouse gas emissions, but technology is helping us shop greener. Our transport platform optimizes delivery routes, cutting fuel use and emissions. Algorithms select right-sized boxes to minimize waste. AI predicts demand, preventing overproduction and food waste.</p><p><strong>Conclusion</strong></p><p>Emerging technologies are crafting a shopping experience that’s more interactive, efficient, and personalized. As these technologies evolve, they’ll continue to reshape retail, making shopping more fun and convenient for consumers while giving retailers powerful tools to streamline operations and engage customers innovatively. Who knows? Maybe one day, shopping carts will read your thoughts, or malls will transform based on your mood. Feeling blue? How about a beach day at the mall!</p><p>And so, our shopping adventure continues, driven by technology, towards a future filled with endless possibilities.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=13055b407fff" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Putting the AI in retail… introducing Tesco’s award-winning Data Science team]]></title>
            <link>https://medium.com/@tescotechnologycommunications/putting-the-ai-in-retail-introducing-tescos-award-winning-data-science-team-1afcf5737d52?source=rss-2ce0c24b1b1b------2</link>
            <guid isPermaLink="false">https://medium.com/p/1afcf5737d52</guid>
            <dc:creator><![CDATA[Tesco Technology]]></dc:creator>
            <pubDate>Thu, 30 Nov 2023 11:24:07 GMT</pubDate>
            <atom:updated>2023-11-30T17:14:26.515Z</atom:updated>
            <content:encoded><![CDATA[<p>Hear from our Data Science Director, <a href="https://www.linkedin.com/in/robarm/">Rob Armstrong</a> about the work our Data Science team get up to…</p><p><em>Personalising billions of search and recommendation events<br>Optimising millions of home delivery schedules round the clock<br>Consistently improving millions of price reductions<br>Enhancing millions of online order picks across thousands of stores<br>Continually optimising thousands of truck routes between depots and stores</em></p><p>… just a fraction of what Tesco’s <a href="https://www.theorsociety.com/membership/awards-medals-and-scholarships/presidents-medal/previous-awards/#2022">award winning</a> Data Science team empowers every week. And when combined, it all helps reduce thousands of tonnes of CO2 emissions per year, while increasing business efficiency and profitability.</p><p>But we didn’t achieve this scale and impact overnight. So, as I celebrate my seventh anniversary of leading the team, I thought it would be a good opportunity to reflect on our journey as well as the key ingredients that have helped us get here.</p><p>But before we begin, I want to credit our many successes to the wonderfully talented data scientists, engineers, product managers and all the other Tesco colleagues who have been at the centre of this journey. Without their excellence, collaboration, and determination, none of this would have been possible.</p><p><strong><em>How our data science story began…</em></strong></p><p>The year was 2016. The UK was still in the EU, Obama was President, and I joined Tesco to lead its fledgling Data Science team. The team was part of the customer organisation and comprised 10 talented data scientists, hired mostly straight out of academia, who were still finding their feet in the corporate world and understanding how to generate value for the business using data science.</p><p>Plotting where the team sat in 2016 based on the type of data science work they were conducting, you’d find them in the ‘researching interesting problems’ zone. And they faced challenges when it came to productionising and scaling their solutions to release real world value. These challenges included the availability of centralised productionised data sets, the availability of downstream systems and APIs with which to integrate and deploy solutions and the absence of a product-led mindset when developing scalable software solutions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/903/1*wh0Pf02pE6KCuGxVZsmXpQ.png" /><figcaption><strong>Figure 1: A spectrum of Data Science team activities</strong></figcaption></figure><p><strong><em>So how did we get here…</em></strong></p><p>When I look back over our seven-year journey, there are several key ingredients which have helped shape how we think about data science in Tesco and how it can add value to the business. They’ve also helped the team grow organically year on year by securing increasing investment built on the ROI achieved through our diverse portfolio of AI applications. And this has ultimately helped shift the perception of data science from being a cost centre to profit centre.</p><p>I’ve visualised these key ingredients in Figure 2, and I’ll dig into them in a little more detail below.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/518/1*0B6l7f-7up4-3W4-SoJVDg.png" /><figcaption><strong>Figure 2: The key ingredients to our data science success</strong></figcaption></figure><p><strong><em>Treating data science as a core technology capability</em></strong></p><p>Prior to 2017, the Data Science team was part of the customer organisation and while it didn’t only work on problems in the customer space, it also didn’t have full remit to solve problems across the whole retail business. Therefore, the strategic decision to move data science into the technology organisation and treat it as a core capability was foundational to the team’s growth and ultimate value generation.</p><p>This shift enabled the team to have a more horizontal remit and work on problems across the entire retail business. As a result, the team were able to identify common technical threads in how to solve similar problems and efficiently scale solutions.</p><p>Identifying these threads led to the establishment of our data science domains (see Figure 3) as well as the development of several core data science platform capabilities that act as the foundations for multiple technology products delivering ongoing efficiencies and performance improvements.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/903/1*H9cC9DeGHACBX7N899tAXQ.png" /><figcaption><strong>Figure 3: Our data science domains</strong></figcaption></figure><p><strong><em>Adopting an agile product methodology for developing algorithmic software products</em></strong></p><p>In parallel and alongside the wider technology organisation, the Data Science team adopted an agile product methodology for developing technology products. This involves working as part of broader agile developments teams made up of all the expertise required to discover and develop solutions including product managers, software engineers, analysts, data engineers and UX designers (to name a few).</p><p>Adopting this approach shaped how we operate and has enabled us to provided clarity on the role that data science plays at Tesco.</p><p>We summarise it as follows:</p><p><strong><em>“The Data Science team is responsible for developing the algorithmic (AI) features of Tesco’s technology products which support or automate decision making”</em></strong></p><p>This clarity of purpose has helped us to communicate our role in the product discovery and development process and determine what should, and often more importantly should not, form a part of our data science portfolio.</p><p><strong><em>Experiment, experiment, experiment</em></strong></p><p>The importance of experimentation and how it has helped influence the culture and performance of our team can’t be overstated. I would equate doing data science in the absence of a robust experimentation approach to driving a car without a speedometer. You might have a general sense of how fast you are going, but you won’t know for sure if you were going too fast until you receive a fine for going 50mph in a 30mph zone. The same can be said for building data science solutions; it’s too late if you find out for the first time that a model isn’t performing once you’ve rolled out at scale.</p><p>For a long time, due to the relative maturity of the online space and the presence of industry standard approaches and platform for A/B testing, online was our most mature area for experimentation. As a result, it was relatively straightforward to measure the performance of different models by providing portions of online traffic with different treatments.</p><p>In the physical world, it has been more complex to introduce automated, and always on, mechanisms to measure model performance across representative test and control groups.</p><p>In the physical space, these mechanisms weren’t always available straight out of the box. It’s been a journey working with our Analytics and Experimentation teams to develop and implement robust, reusable, and automated approaches to ensure we build experimentation into all of our development areas.</p><p>Establishing these approaches has ultimately enabled us to understand the performance, and therefore the value, of our developments in a lot more detail and it allows us to make important prioritisation decisions on where to focus our constrained resources.</p><p>Looking forward, we’re laying the foundations for more advanced simulation and digital twin capabilities which will help us to further prioritise which of the many resource and time intensive in-store trials are the best candidate to invest in.</p><p><strong><em>Implementing a robust Data Science Career Framework</em></strong></p><p>Even now, more than a decade after Harvard Business Review’s article claiming the role of <a href="https://hbr.org/2012/10/data-scientist-the-sexiest-job-of-the-21st-century">Data Scientist as the sexiest job of the 21st century</a>, there is still no single globally recognised definition of what constitutes a Data Scientist and every company I’ve worked at has had a slightly different definition to fit their internal context. It was therefore important to craft our own definition and build a robust career framework to ensure data scientists had clarity on their roles, career path opportunities and how they add value to the business.</p><p>We launched v1 of our Data Science Career Framework in 2016 which we lived and breathed for six years before launching v2 in early 2022.</p><p>Our main learning from the v1 career framework was the importance in finding a balance between detail and usefulness. In v2 we wanted to capture the essence of what it means to be a Data Scientist but not try to be exhaustive and capture every single skill as these will naturally change over time as the field of data science continues to evolve.</p><p>We framed our v2 career framework around four core data science skills under which we capture more detailed expertise and experiences.</p><p>The four core data science skills we focused on were:</p><p>1. <strong>Defining data science problems:</strong> The ability to translate business problems into data science problems and generate hypotheses by effectively and comprehensively gathering requirements.</p><p>2. <strong>Solving data science problems: </strong>The ability to identify and apply the right tools and modelling approaches for building high quality solutions to data science problems by effectively leveraging Tesco’s technology stack.</p><p>3. <strong>Validating data science solutions:</strong> The ability to robustly test and validate the impact (with respect to the business metrics of interest) of data science solutions/products selecting, defining, and implementing appropriate validation approaches.</p><p>4. <strong>Deploying data science solutions:</strong> The ability to architect, deploy and integrate reliable data science solutions, at the appropriate scale, that smoothly integrates with Tesco’s infrastructure.</p><p>We used these four skill areas (plus a wider set a core technology skills) to outline the expectations of our data science roles ranging from our entry level associate data scientists to our most senior principal and heads of data science.</p><p>Having a robust career framework has allowed us to standardise on our role expectations and how we can best support the learning and development of our data scientists and set a consistent bar for our global recruitment processes.</p><p><strong><em>Partner with business experts and don’t operate as an intellectual island</em></strong></p><p>Whenever you bring together a group of technical experts who are at the top of their respective fields, there is a risk they’ll believe they can solve any business problem in isolation (as an island). We’ve learnt the hard way that this is rarely the case, and it is often a sure-fire path to failure.</p><p>We have learnt that the most successful data science solutions are built on a deep partnership between technical experts and business experts who often have a lifetime of real-world operational experience which should be leveraged at every opportunity.</p><p>This type of partnership can come in many forms, but we’ve found that there is no substitution to sitting with business experts to observe and learn about their processes and policies and in return share our thoughts on engineering a data science solution. This also shouldn’t be viewed as a one-off activity but rather an ongoing sharing and re-validating exercise as a solution is developed.</p><p><strong><em>Establishing a Data Science Value Pipeline</em></strong></p><p>The final key ingredient is around how we have viewed our Data Science team through the lens of a ‘Value Pipeline’. It’s about connecting all the activities our team of talented data scientists and engineers are working on, and understanding how each activity ultimately supports the next stage in the pipeline to the end goal of delivering real world value via productionised algorithmic product features.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/903/1*nVxCtGVT0e1Tl6zNXlJKGA.png" /><figcaption><strong>Figure 4: Our Data Science Value Pipeline</strong></figcaption></figure><p>Here’s a brief description of the stages in our Value Pipeline:</p><p>1. <strong>Awareness activities:</strong> building internal awareness of the problems that data science can help to solve. These activities come in many different forms but typically focus on bringing to life our portfolio of developments via meet ups, roadshows, training events, blog posts, newsletters, conferences and academic talks.</p><p>2. <strong>Research activities:</strong> crafting research problems to help drive future technical advances across our development portfolio. Over the years we have invested in building strong ties with academia, ranging from offering PhD internships on highly focused technical topics through to welcoming visiting scientists and sponsoring multi-year PhD projects in broader technical fields that align to our areas of interest.</p><p>3. <strong>Innovation projects:</strong> the innovation stage of our pipeline provides a framework for identifying, prioritising, and working on rapid hypothesis tests with an aim of fostering new problem spaces outside of those already covered by our existing product teams. We measure success in relation to how many high value innovation ideas we’ve been able to test and not on whether the innovation project test was successful. Where a test has been successful, we then look to form a product team to take the development forward.</p><p>4. <strong>Development:</strong> at this stage we are developing the algorithmic features of our technology products, working as part of product teams and following agile product methodologies and software development principles. Our aim is to test and learn often to minimise investments and maximise returns.</p><p>5. <strong>Live and iterating:</strong> This is ultimately where we hope all our value pipeline activities lead. At this stage our developments are live and we are continuing to iterate as part of the product teams to ensure they remain performant and we are releasing additional value as we make new research and innovation discoveries.</p><p><strong><em>In conclusion</em></strong></p><p>The one constant during our first seven years has been the need to adapt and evolve as a team. And when looking ahead to the next seven, I don’t see this changing any time soon…</p><p><em>If you want to hear more about the pricing optimisation solution for which the team won the Operational Research Society President’s Medal you can check out the following article in The OR Society magazine ‘Impact’: </em><a href="https://www.tandfonline.com/doi/full/10.1080/2058802X.2023.2202575"><em>https://www.tandfonline.com/doi/full/10.1080/2058802X.2023.2202575</em></a></p><p>Article written by Rob Armstrong, Data Science Director at Tesco.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1afcf5737d52" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>