Introduction
DataHawk's AI-powered diagnostic engine for Amazon sellers; watch your business daily, spot what changed, explain why, and recommend what to do next.
Think of Sherlock as your always-on analyst. It reads your data, spots what changed, explains why, and recommends a next step in plain English.
What Sherlock detects
Sherlock has two surfaces in the left sidebar: Investigate and Cases.
Investigate
- Signals: Your daily starting point. A single, filterable list of every product whose sales moved in a meaningful way week-over-week. Pick a Momentum (Growing, Recovering, Slowing, Declining, Sudden spike, Sudden drop), filter by Tier (Hero / Core / Meaningful), brand, or marketplace, and Sherlock will rank the list by dollar impact so the biggest problems sit at the top.
Cases
Cases are curated shortcuts, each built around a single, well-defined risk pattern. They run on the same engine as Signals but pre-filter the list to one story at a time.
-
Zombie Ad Spend: Ad groups spending for 10+ days without a single conversion. Pause them or fix them before they drain the budget.
-
Return Anomalies: Products whose return rate has jumped at least 50% above the 90-day baseline. Often the first sign of a listing mismatch, sizing issue, or quality problem.
-
Ad Spend Drop: Products where ad spend fell ≥10% week-over-week, pulling traffic and revenue down with it.
-
Price CVR Drop: Products where ASP rose ≥5%, conversion fell, and revenue followed. The classic "we priced ourselves out" pattern.
-
FBA Stockout: Products that went out of stock on FBA this week, lost the Buy Box, and saw revenue collapse.
-
Buy Box Loss: Listings where ASP rose ≥5% and Buy Box dropped ≥10pp.
How Sherlock works
Every investigation runs through the same 5-step pipeline:
Fetch metrics
Pull 30 days of daily data: revenue, sessions, CVR, Buy Box, ads, inventory, price, rank.
Compute evidence
Compare the last 7 days vs the previous 7, and the last 3 vs the previous 3. Build deltas, decomposition, and scored root causes.
Fetch events
Pull external events (Prime Day, holidays), seller annotations you've logged, and product changes.
Load the knowledge playbook
Apply DataHawk's library of diagnostics and causality rules.
Present
A single LLM call combines all of the above and produces three outputs: Summary, Explanations, Suggestions.
Chat with Sherlock
Every investigation has an assistant built in. Open it to:
- Ask follow-up questions in plain English.
- Log a seller annotation (e.g. "we launched a media campaign 7 days ago") that will appear in the Timeline and feed every future investigation.
- Give thumbs-up or thumbs-down feedback so Sherlock improves over time.
You can also open the assistant from any page in aim menu of Signals to deep-dive into your full DataHawk data layer.
What you need to get started
Sherlock works out of the box once your DataHawk workspace is connected to Amazon Seller Central and/or Amazon Advertising. No extra setup required.
Sherlock is Amazon-only for now. Walmart support isn't available yet.
Where to go next
Signals
Your daily filterable list of every meaningful product movement.
Zombie Ad Spend
Find wasted ad dollars.
Return Anomalies
Catch return spikes early.
Ad Spend Drop
Spot quiet ad pullbacks.
Price CVR Drop
Catch the 'priced ourselves out' pattern.
FBA Stockout
React fast when FBA goes empty.
Buy Box Loss
Find listings priced off the Buy Box.
Frequently Asked Questions (FAQ)
Common questions about DataHawk MCP; getting started, security, multi-workspace, troubleshooting, and roadmap.
Signals
Sherlock's daily starting point; products with meaningful sales movement ranked by dollar impact, filterable by Momentum, Tier, brand, and marketplace.