Find the margin leak.
Fix it. Prove it.
Your systems already recorded the problem. Ward reads them, names the cause, runs the fix, and closes the case only when the KPI moves.
Nine SKU-store pairs cross the predicted-zero threshold inside 48 hours. Store 22 produce is the most urgent.
forecast.daily | Lettuce at Store 22: 84 units/day against 38 on hand, zero by 11a Thursday |
|---|---|
supplier.lead | Regional DC lead time 18 hours, no sufficient PO already in transit |
Vendor C billed above the active price list. The overage is recoverable.
invoice.line | 1,200 units at $4.28; contract clause 3.2(b) sets $4.12 |
|---|---|
dispute.amount | $0.16 per unit, $192 total, classified as off-contract pricing |
Ranked on GMROI and velocity together, not units alone, so profitable slow movers survive the cut.
gmroi.sku | 412 SKUs below $0.90 GMROI holding $1.8M in working capital |
|---|---|
assortment.gap | 31 SKUs selling in the urban cluster are unstocked in a demographically matched cluster |
Three stores in the ten-store cluster breach the 2% threshold on one SKU. The cause is upstream, not theft.
shrink.rate | Stores 4, 11, 19 at 2.4–3.1% against a 1.2% cluster baseline |
|---|---|
delivery.defect | All three take delivery from the same lane, defect rate 4.8% above goal |
We ran 400+ stores. Every tool we bought stopped at the chart.
The stockout was in the POS on Tuesday. Nobody read it until Friday, and the sale was gone. The information was already there. What was missing was the distance between a system recording something and a person doing something about it, and one to three points of revenue live in that gap.
Ward closes it. How we got here →
Ward is built on the askotter observability platform, and these nine companies run on that platform in production today. They are askotter customers, not ours. See their numbers →
Ward’s own work is in enterprise grocery at nine-figure revenue: AI strategy, change management, orchestration and the reporting layer, on live systems. Pilot status, updated as it runs →
A dashboard tells you a number moved. Ward tells you why, then moves it back.
Four layers. Adopt them separately.
Retail observability
- POS, ERP, WMS, labor, finance, supplier feeds
- Scored against each store’s own baseline
- Every number one click from its SQL
AI orchestration
- Anthropic, OpenAI, Gemini, Ollama, on your keys
- Cheapest model that clears your accuracy bar
- Forecasts on classical models, never the LLM
Workflow automation
- 43 playbooks, specced end to end
- Writes into SAP, Oracle, Blue Yonder, Relex
- Every write gated on a named approver
Security and data governance
- Federated query. No second copy of your data
- Cedar policy per agent, versioned in your Git
- Audit stream to your SIEM as JSONL
Read the architecture, or start on the onboarding path for IT →
What we are running today, and the arithmetic behind it.
What is running and what is modelled sit in separate boxes, so you can judge each on its own.
AI strategy work, in enterprise grocery.
Change management, AI orchestration and the reporting layer on top, against live POS, ERP and inventory. Stockouts, shrink, fill rate and assortment are where it starts, because that is where the dollars are.
We do not name operators or publish their numbers without permission. Reference calls get arranged while you are still evaluating. How the engagement is going →
Where 200 basis points would come from.
Four signals, worked out for a mid-market multi-store estate. Nobody has booked a dollar of it yet.
None of the four require hiring an analyst. The same arithmetic in dollars →
Here is what each option on your shortlist costs.
- Spreadsheets · $0 visible, 1 to 3% of revenue in avoidable stockouts and markdowns
- Tableau plus one analyst · $170K a year, 3 to 6 months to a first answer
- ThoughtSpot Enterprise · $400K to $1M+, built for the Fortune 1000
- Ward Connected · $60K, live in six weeks, writes back gated on a human
Connected in a week. Acting in six. Answered in ninety.
No warehouse project and no modelling engagement. Ward reads what you have, in the shape it is in.
-
48 hoursFirst insight cards
From a read-only connection. Findings on your own data, not a sandbox and not a slide.
-
Week 2Findings ranked by dollars
Stores, SKUs and vendors ranked by what they cost you, with the cause named and the SQL one click under every number.
-
Week 6First gated write
One playbook, one system of record, blocked on an approver role you named. Everything before this is read-only.
-
Day 90The number, either way
Measured KPI delta against the metric agreed on day one. If it did not move, the pilot ends.
Forty-three playbooks arrive already specced.
Each one names its trigger, its procedure, the system it writes to, and the KPI that has to move before the case can close. Your job is to pick which ones run.
Stockout escalation
SKU hits zero inside 48 hours. Raises the replenishment, tells the buyer.
- Target KPI: On-shelf availability
- Runs within 6 hrs of the trigger
Vendor invoice dispute
Invoices reconciled against receipts and terms. Drafts the dispute.
- Target KPI: COGS recovery
- Runs within 24 hrs of the trigger
Promo conflict cancellation
Two promos eating each other mid-flight. Proposes killing the overlap.
- Target KPI: Net promo lift
- Runs within 2 hrs of the trigger
What happens between the anomaly and the fix.
The AI is the part you can see.
Every vendor demos the tip: a model that answers, a dashboard that watches. The nine tenths underneath decide whether it clears your security review.
What a demo shows you
What has to be true before it runs on your estate
Nine tenths of this product never appears in a demo.
Access control
SAML SSO against your IdP, SCIM provisioning, RBAC per role.
Data governance
Retention per data class. Residency in US, EU, or JP. Named sub-processors.
Lineage
Every answer carries the SQL it ran and the tables it touched.
Approval gates
Nothing writes to SAP, Relex, or Oracle without a named approver.
Audit trail
Detection, cause, playbook, write-back, approver. Streams to your SIEM as JSONL.
Cost control
Your LLM keys, your spend caps, failover across providers.
Reconciliation
Figures checked against the system of record before they publish.
Integration plumbing
POS, ERP, WMS, and the write-back path into each. Takes other vendors nine months.
Your team owns this after week six.
Picked a model vendor, a warehouse, and a policy framework already? Ward inherits all three.
It adopts your decisions
- Your model contracts, your keys
- Your warehouse stays the system of truth
- Your SSO, SCIM, and RBAC roles
Your team configures it
- A scope change is a pull request
- Write access opens one playbook at a time
- Versioned policy, so it rolls back
It survives review
- Full architecture packet before the pilot
- Read-only until you turn a write on
- SOC 2 Type II underway, questionnaire pre-answered
Your tier is set by how complex your stack is.
A $20M group on three POS systems is harder to run than an $80M single-brand chain. Revenue is a lazy proxy for the work.
Tier follows stack complexity: how many systems have to talk to each other, how many brands you run, and how custom your schema is. Not headcount, and not revenue.
- Month-to-month, 30 days notice
- No rollover clause
- Every tier is the full platform
Every vertical gets its own tuning.
The maths changes with velocity, regulation, and SKU density. Pick your vertical or your job.
Ask an AI about Ward instead.
Our architecture, integrations, and security posture are public and machine-readable. Ask a model with no stake in the answer.
What is Ward (getward.ai) and how does its closed-loop retail analytics compare to Power BI, Looker, and ThoughtSpot for a mid-market multi-store retailer?
Pre-filled. Or read llms.txt, the source we publish for models.
The things buyers actually ask.
An AI analytics platform for multi-store retail. It reads the systems you already run, names where margin is leaking, and runs the playbook that fixes it. A case closes only when the KPI moves.
Everything a dashboard does happens before the interesting part. Ward names the cause, attaches the playbook, writes back once a named human approves, then measures the KPI. The work after the chart is the product.
A general model can write a query. It does not know your planogram, vendor terms, or markdown ladder, and it cannot write back to SAP.
No. Store, ops, and finance users ask directly. If you have a data team, they keep their warehouse and Ward reads from it.
Not by default. Ward connects read-only. Write-back is enabled per playbook, gated on a named approver, and logged before and after.
It stays where it is. Ward queries your warehouse in place: SaaS, your VPC, or an on-prem connector. Full deletion on request, covered in the DPA.
Model-agnostic, running on your keys. Each query routes to the cheapest model that clears the accuracy bar, with failover across providers.
TLS 1.3, AES-256 at rest, SSO/SAML, SCIM, RBAC, customer-managed keys. SOC 2 Type II underway. Security questionnaire pre-answered.
Underneath it, as the retail-specific layer. Ward brings observability over your retail data, orchestration across the model providers you have already contracted, workflow automation into your systems of record, and the governance around all three. If you have picked a model vendor, a warehouse, and a policy framework, Ward inherits all three.
Cedar policies and agent charters live in your repo, so scoping an agent is a reviewed pull request rather than a support ticket. Each charter declares its scope, sources, allowed actions, owner, and version. Playbooks are specced objects you tune per vertical and per store cluster, and write scope opens one playbook at a time against an approver role you define.
A read-only service account per source and an identity integration. No ETL to schedule, no warehouse to stand up, no model to host. Ward queries Snowflake, BigQuery, Redshift, Postgres, and SAP HANA in place. After week six the configuration is yours and most changes happen in Git.
The application talks to an abstraction layer instead of one provider's API, so Anthropic, OpenAI, Gemini, and Ollama are all config. Each task routes to the cheapest model that clears your accuracy bar, with failover across providers. An eval harness scores any candidate model against your real cases first, so promoting a new one is a measured decision.
Read-only connection in week one. First insight cards in week two. First playbook running against a system of record by week six.
$24K to $240K+ a year plus one-time setup. The tier follows how complex your stack is. Most mid-market retailers land on Connected at $60K.
Single-store operators, and anyone shopping for self-serve BI. Ward is for multi-store retailers where one unnoticed leak costs more per month than the software does per year.
Close criteria are agreed in writing up front. If the KPIs do not move, the pilot ends. Better a no in 90 days than a churn in 18 months.
See what your own data has been hiding.
Read-only connection, first findings in two weeks. Or score yourself in three minutes, no call required.
Read-only to start · Your LLM keys · SOC 2 Type II underway
Find out what your data has been hiding.
Tell us about your operation. We’ll show you the problems Ward catches, and the ones your current tools miss.