AI and engineering for consumer goods companies, retail chains and D2C brands. One sales picture across every channel, and the systems to act on it.
Most planning tools assume the right-hand column is already solved. Usually it is not. That is where we start.
In brief
EigenSpark builds AI systems and the software around them for consumer goods companies, retail chains and D2C brands. We forecast demand, spot stock-outs, measure promotions, track prices and plan range. We automate the supply chain that carries all of it. And we keep your labels and listings inside the rules, which change more often than anyone can track by hand.
What we cover
The rules keep moving. We check every product and every listing against the current set.
The starting position
Hundreds of distributor files, one product with five different codes, and listings that drift.
Every distributor uses its own format, item names and week endings. Every forecast and every incentive is built on that pile.
Your code, the retailer’s article code, the listing ID and the name on the pack. Until they agree, a channel comparison is a guess.
The next promotion is planned while the last one is still being reconciled. The analysis arrives after the decision.
Content, declarations and images change across dozens of listings. Nobody has the hours to check every one by hand each week.
How an engagement runs
The first step is the one everything else depends on. It is also the one nobody quotes for.
Two to three weeks turning your distributor files, modern trade reports, marketplace settlements and D2C orders into one reconciled view, with one product identity that survives across all of them. This is the piece everything else depends on.
A single decision taken end to end: a forecast a planner acts on, a promotion evaluated before the next one is committed, or a listing compliance sweep. Small enough to prove in a quarter and specific enough to measure.
Ingestion, identity resolution, the calendar and the rule engine are shared. Pricing, assortment, promotion and compliance use cases reuse them, and your team maintains the mappings without waiting for us.
Use cases
Demand and availability, pricing and promotions, listings and labels, customers and claims.
Distributor and retailer files read whatever their format, item names matched to your master, and the result reconciled against primary sales with every gap listed.
Data Engineering & Platforms →Offtake history, seasonality, promotion calendar and channel mix read into a forecast at the level a planner actually orders, with error tracked per product.
AI & Data Strategy →Store and dark store availability read from what sold and what did not, so a stock-out is visible on the day it starts.
Agentic AI →Each scheme measured against a baseline from your own channel data, with uplift, cannibalisation and forward buying separated before the next spend is committed.
AI & Data Strategy →Your prices and your competitors’ read across marketplaces on a schedule, with policy breaches and unexpected discounting flagged to the person who can act.
Agentic AI →Outlets grouped by what they actually sell, with the range each group should carry proposed against space and offtake, and the numbers behind the case shown.
AI & Data Strategy →Every live listing checked against Legal Metrology and category rules, with failures listed per product per platform and re-checked as listings change.
Continuous Compliance Monitor →Titles, images, attributes and descriptions compared against your master catalogue across every platform, so a wrong pack size is found before a customer finds it.
Document Intelligence Engine →New artwork read against the declarations it has to carry and against the last revision, with every change tagged to the rule it touches, before the plate is cut.
Document Intelligence Engine →Reviews, complaints and service transcripts read for the recurring cause behind them, grouped by product, batch and region, while the problem is still small.
Generative AI →An assistant that answers order, return, warranty and product questions from your approved material and the customer’s own history, and cites what it used.
Generative AI →Scheme claims, marketplace settlements and returns reconciled against what actually sold, with the differences returned as an exception list with money attached.
Continuous Compliance Monitor →The engineering half
Four layers of ordinary engineering. The first one decides whether anything above it works.
Nothing else on this page works until a SKU and an outlet mean the same thing everywhere. It is the least exciting layer and the one that decides the outcome.
The ERP and the distributor management system stay. Most of the effort is everything that arrives from outside them.
Built for a demand planner, a key account manager and a regulatory desk, with the assumptions shown so a forecast can be argued with.
Defined once and used everywhere, so the sales review, the supply plan and the claim settlement cannot disagree. Your team runs and owns it.
We also run the training that goes with it: SQL and data engineering, cloud, enterprise systems and cybersecurity, alongside the AI tracks. A team that cannot query its own data cannot specify a model either.
What we offer retail and CPG teams
Most companies start with the sales data or a listing check, and use three of the four.
Where a channel data programme starts.
Where a single workflow starts.
AI tracks, and the foundations under them.
The teams that own this work.
Working in a specific function? See how we help Supply Chain & Operations teams.
Security and controls
Your planner overrides the model, every number traces back to a file, and nothing publishes itself.
Forecasts, range proposals and promotion recommendations go to the person accountable for the number, with the assumptions visible and the override recorded. A model that cannot be argued with gets ignored.
A forecast, an uplift figure or a claim exception links back to the distributor file, the settlement report or the point of sale extract it came from, on the week it arrived.
SKU mappings, outlet clusters, promotion calendars and compliance rules are managed in a console by your own team. A new retailer, a new platform or an amended rule does not need a release.
Listing checks produce a worklist for your regulatory and e-commerce teams. Nothing is edited, delisted or published on your behalf.
FAQs
Because a forecast is only as good as the offtake it is trained on, and in most consumer businesses that offtake sits in several hundred distributor files, a set of modern trade reports in the retailer's own item codes, and platform reports by dark store. Harmonising that is the precondition for forecasting, promotion analysis, assortment and claims all at once. It is the least interesting slide in the deck and the one that decides the outcome.
Yes, and it is the normal starting point. Files are parsed on their own structure and content, item names are matched to your master by meaning so a renamed line still lands, and anything the system cannot place with confidence goes to a person to map once. Those decisions are remembered, so the manual work shrinks every month.
Yes. Since July 2026 imported products have to be discoverable by country of origin through searchable and sortable filters, and the mandatory declarations have to appear on the product page. We check every live listing against the rule set, report failures per SKU per platform, and re-check as listings change. The rule set is held as configuration your regulatory team edits, because these notifications keep coming.
No. Compliance checks produce a worklist for your regulatory and e-commerce teams with the specific failure and the rule it breaches. Nothing is edited, delisted or published on your behalf, and no price is changed by a system.
The compliance and catalogue work makes sense at any size, because the Legal Metrology rules carry no exemption for a small business or a D2C brand. The forecasting and promotion work needs enough history and enough channel spread to be worth it, which usually means a national distribution footprint or a serious multi-platform presence.
Much of it, from the other side. Availability, assortment by cluster, price tracking, catalogue quality, review mining and the customer assistant all carry over. Secondary sales harmonisation and trade spend are manufacturer problems and would not apply.
No, and the parts that are not are usually what makes the AI work. Ingestion, identity resolution, calendar normalisation and the rule engine are ordinary engineering, and they are most of the effort. The same is true of the training: alongside the AI tracks we run SQL, data engineering, cloud, enterprise systems and cybersecurity.
In your own cloud tenancy. Channel data is commercially sensitive and it stays inside your boundary. Where a hosted model is used for text work such as review mining, what is sent to it is agreed and recorded before build.
Related
The service this ships as, the product behind the listing checks, and the teams that own it.
The channel data layer everything else sits on.
→ ProductListing, label and claim checks against a live rule set.
→ ServicePrice tracking, availability and catalogue monitoring.
→ FunctionThe function that owns forecasting and fill rate.
→ FunctionThe function that owns trade spend and the channel.
→ PillarThe engineering side, for systems that run a channel.
→Send us one month of secondary sales in whatever formats they arrive in. We call you within 48 hours and go through how much matches up on its own, where product codes break, and what it would take to get one clean view.
Talk to us