Signals

Sherlock's daily starting point; products with meaningful sales movement ranked by dollar impact, filterable by Momentum, Tier, brand, and marketplace.

⚠️

Sherlock is in active development. If something on screen doesn't match this page, drop a note via the assistant and we'll catch up.

How the app is organized

The left sidebar has two groups:

  • InvestigateSignals. The full, filterable list of every meaningful product movement.
  • CasesZombie Ad Spend, Return Anomalies, Ad Spend Drop, Price CVR Drop, FBA Stockout, Buy Box Loss. Each Case is a curated shortcut for a specific risk pattern.

If you don't know where to start, open Signals. If you already have a hunch ("I think we have an ad-spend issue"), jump straight to the matching Case.

What's on the Signals page

When you open Signals, you'll see a list of products with these columns:

ColumnWhat it means
ProductProduct title, marketplace flag, parent and child ASIN (each with a copy icon)
TrendA small sparkline of the metric Sherlock is flagging
Growth% change; toggle the header between WoW and MoM to switch the view. Green = up, red = down
$ ImpactAbsolute revenue difference between the current period and the prior equivalent period, in the product's sales currency. Follows the WoW / MoM toggle. Click the header to sort
Rev ShareProduct's contribution to your total revenue over the last 6 months (USD)

At the top, you'll see how many products are showing (e.g., "Showing 1–50 of 184 products").

💡

The small flask icon at the far right of each row means "Recent analysis available." Sherlock has cached the investigation, so it'll open instantly. Rows without the icon will generate a fresh analysis on first click (a few seconds, sometimes longer).

Filter to find what matters

The filter bar sits just above the table. Each filter has an in-app tooltip that explains the exact rule.

Momentum

Filter by revenue trend (WoW vs MoM).

OptionExact rule
AllShow all products regardless of revenue trend
GrowingRevenue up week-over-week AND month-over-month (both > +10%)
RecoveringRevenue up this week (WoW > +10%) but still below last month (MoM < −10%)
SlowingRevenue down this week (WoW < −10%) but still above last month (MoM > +10%)
Declining (default)Revenue down week-over-week AND month-over-month (both < −10%)
Sudden spikeRevenue jumped this week well above the monthly trend; something changed recently
Sudden dropRevenue fell this week well below the monthly trend; something may have broken

Tier

Filter by L6M revenue contribution. Use this when you want to focus only on products that move the needle.

OptionExact rule
AllInclude all products regardless of revenue contribution
MeaningfulTop 50% of products by total revenue over the last 6 months
CoreTop 25% of products by total revenue over the last 6 months
HeroTop 10% of products; your highest-revenue ASINs over the last 6 months

Brand

Multi-select dropdown with a Search brands input. Tick one or several to narrow the list.

Marketplace

Multi-select dropdown listing every marketplace you have data for.

Min. 10 units

Toggle button. Hides products with fewer than 10 units sold in the last 30 days. Avoids noisy % swings on near-zero volume ASINs. Most teams leave this on for the weekly review.

Reset

Appears whenever at least one filter is active. Clears every filter in one click.

💡

Quick recipe for your weekly review: Momentum = Declining, Tier = Core or Hero, Min. 10 units = on, sort by $ Impact. That's usually the shortest list of the most expensive problems.

Top-right icons

In the top-right of the page:

  • Refresh: Re-runs the analysis with the latest data.
  • Settings: Opens the period and pre-generation panel.
  • Search products: Slides out a search field. Type a keyword or ASIN to filter the list.

Open an investigation

Click any product to open its investigation. The header tells you the date the investigation covers and, once cached, when it was last updated.

You'll find three tabs:

Overview

The main story. Sherlock always produces three outputs: Summary, Explanations, and Suggestions. From top to bottom:

  • Summary: One-paragraph plain-English explanation.
  • Explanations: A deeper breakdown of what moved.
  • Metrics table: Side-by-side comparison of the two 7-day windows.
  • Causal chain: The cause-and-effect logic Sherlock followed.
  • Suggestions: Concrete actions to take.
  • Was this helpful?: Thumbs up or down to help us improve.

Below, an Overview dashboard shows sparkline charts for the metrics most relevant to this product.

Timeline

A chronological list of events that may have affected this product:

  • Product changes (parent ASIN swap, title change, etc.)
  • Shopping events (e.g., Prime Day)
  • Seller annotations: Notes your team has logged via the Sherlock assistant

Metrics

A heatmap of every key metric over the analyzed period, grouped by:

  • Sales Performance: Orders, units, refunds, shipments
  • Pricing: Selling, listing, landed, shipping price
  • Traffic & Engagement: buy box %, sessions, conversion
  • Advertising: Ad sales, ad spend, ACOS, ad orders, ad units

Green cells sit above baseline, red cells below. Use the T0 / T1 / All toggle to focus on a single window or compare both.

Share or chat about an investigation

Top-right of any investigation:

  • Share: Copy link (live) or copy link at current date (locked)
  • Open assistant: Opens the Sherlock assistant in a side panel

Two modes for the assistant

  • Product-scoped (inside an investigation): The assistant already knows the product. Ask follow-ups, log feedback, or save a seller annotation.
  • Workspace-scoped (from any page): Deep-dive into your full DataHawk data layer for broader questions.

Tips

  • Always start with $ Impact: The biggest dollars are at the top.
  • Use Min. 10 units to cut low-volume noise.
  • Use Pre-generate top 10 in Settings before your weekly review so investigations open instantly.
  • Drop seller annotations into the assistant as you go: They give every future investigation more context.

Under the hood (for the curious)

Every Sherlock investigation runs through the same 5-step pipeline:

  1. Fetch metrics: Pull 30 days of daily arrays from Snowflake: Revenue, sessions, CVR, Buy Box, ads, inventory, price, rank.
  2. Compute evidences: Compare the last 7 days vs the previous 7, and the last 3 vs the previous 3. Build deltas, decomposition, event start dates, and scored root causes.
  3. Fetch events: External events (Prime Day, holidays), seller annotations, and product changes.
  4. Load knowledge playbook: Apply DataHawk's library of diagnostics and causality rules.
  5. Present: A single LLM call combines all of the above and produces the Summary, Explanations, and Suggestions.

On this page