Overview
01 · Activation
Social Proof
02 · Substrate
Demand IntelligenceCohorts
03 · Engine
EngineSignal AI
04 · Measurement
AttributionPortal & Analytics
06 · Integrations
Integrations overviewCommerce platformsCustomer EngagementCustomer data platformsAd platformsData warehouseCustom & developer
Pricing
Contact
Book a demo
Platform overview  →
Three layers of demand intelligence — API, portal, assistant
01 · Activation
Social Proof  →
Messaging across 8 surfaces
02 · Substrate
Demand Intelligence  →
Live per-product demand model
Cohorts  →
Segments from live behaviour
03 · Engine
Engine  →
Real-time selection pipeline
Signal AI  →
Ask your demand data anything
04 · Measurement
Attribution  →
Incremental orders & revenue
Portal & Analytics  →
Live visibility & reporting
06 · Integrations·Connect Flockr to the rest of your stack
Overview  →
Commerce platforms  →
Customer Engagement  →
Customer data platforms  →
Ad platforms  →
Data warehouse  →
Custom & developer  →
The conversational layer

Not a chatbot. An analyst.

Signal AI is embedded in the Flockr Portal. Full knowledge of how Flockr works, direct access to your live data — ask plain-language questions, get answers grounded in source, applied to your store, current to within the hour.

14 live-data toolsRAG over docsStreaming
Signal· dripmoda · live data×
What’s the top scarcity risk right now?GJ
S

The most urgent scarcity risk is the Volta Black Tracksuit.

5 active carts, 95 views in the last 24 hours, accelerating demand, only 2 units left.

Recommended actionRestock immediately.
2 sources
Ask Signal anything about Flockr…
By design

Four things, by design.

Each one a deliberate engineering choice — what Signal is, and what it isn’t, by construction.

01 · Grounded

Source-grounded

Every answer cites the chapters it drew from. Visible source pills below each response — expand to show exactly what informed the answer.

3 sources5 sources
02 · Configured

Knows your configuration

Surface enablement, breakpoints, pipeline overrides, mode — Signal answers for your account specifically.

“Looking at your configuration, the active browse surfaces are PLP, predictive search, and recommendation…”
03 · Live

Live data, not generic

14 tools fetch real figures from your data when needed — conversion lift, scarcity risk, signal performance, momentum leaderboards. Cached for repeat speed; never stale.

14
Demand · AnalyticsAttribution · Realtime
04 · Read-only

Read-only by design

Signal can explain, show trade-offs, draft an internal comms message — but it doesn’t modify anything. Changes happen through the relevant Portal pages, deliberately.

ReadWrite
Architecture

How Signal answers.

Each question runs through three stages. Answers stream as they generate — word by word, no spinner.

1Rewrite

Resolve pronouns & follow-ups

Pronouns and short follow-ups are resolved against the last two turns into a self-contained question.

user:"why?"rewritten:"why is scarcity winningso many slots on PDP?"
Resolves against last 2 turns
2Retrieve

Vector search + live-data tools

Vector search across the knowledge base finds the top 7 chunks. Tools called inline if live data is needed.

retrieve:Demand intelligenceSelection engine+ 5 more chunkstool_calls:getSignalPerformance(7d)
RAG + 14 live-data tools
3Stream

Word-by-word with sources

Answer streams word by word. Markdown rendered. Sources cited in a pill below the response.

The adaptive model has observed that scarcity signals on PDP are converting at a higher rate than purchase signals for your catalogue…
4 sourcesStreaming
Streaming + source citations

Three stages. One answer.

two stores

Two stores. Two questions.

Knowledge base for how does X work. Fourteen live-data tools for what’s happening with X right now.

knowledge_base· cosine similarityk = 7
Knowledge base

Pre-embedded chapters

Every chapter of Flockr’s documentation embedded as vectors. Cosine similarity finds the seven most relevant chunks for the question.

Queryhow is attribution calculated
01Home
0.89
02Demand layer
0.86
03Selection engine
0.82
04Attribution
0.79
05Demand intelligence
0.77
Top7 chunkspassed to contextAuto-updated
live_data_tools· 14 endpointscache: 1h
Live-data tools

14 tools, grouped by source

Real figures from your data, fetched when the question demands it. Each tool cached for one hour per query.

Demand Intelligence5 tools
trendinghealthfadingscarcity / overstockrank leaderboards
Analytics6 tools
overviewsignalsmessagesproductsconversion liftrevenue / AOV
Attribution2 tools
CTR liftpurchase by product
Realtime1 tool
30-min snapshot
Cached1hr per query· shared across team

Updated automatically when adaptive scoring drifts beyond threshold — Signal stays current with what the engine actually does.

Repeat questions get answers nearly instantly — your team’s questions reuse each other’s heavy queries.

Worked examples

From the live portal.

Two example exchanges. The first explains a metric from the knowledge base. The second uses live data and a follow-up to generate a brief for the ops team.

Example 01 · Knowledge question
Signal· dripmoda · live dataLive mode
how is attribution calculatedGJ
S

How Flockr Attribution is Calculated

Flockr uses a two-group session-level attribution model.

The Two Groups

GroupDefinition
PrimarySessions where a shopper clicked a Flockr-messaged product on a browse surface
BaselineEvery other session in the period — bot sessions excluded from both

For your store, the active browse surfaces are PLP, predictive search, and recommendation.

The Order Constraint

Orders only count toward primary when the purchased product matches the one the shopper clicked. No halo inflation:

  • If a shopper clicked product A on the PLP (with a Flockr message visible) and bought product B, that purchase is not a primary-group order.
3 sources
Client-specific

“For your store” — answer applied to the actual configuration, not generic.

Sources visible

“+ 3 sources” pill expands to show which chapters informed the answer.

Structured markdown

Tables, headings, lists, bold all rendered — not flat text.

Example 02 · Multi-turn + asset creation
Signal· dripmoda · live dataLive mode
which products are most at riskGJ
S

There are currently 390 products flagged as scarcity risks and 122 overstock opportunities.

Highest-priority scarcity risks

The most urgent is the Volta Black Tracksuit — 5 active carts, 95 views in the last 24 hours, accelerating demand, 2 units left. The classic high-priority combination: strong demand against critically low stock. Restock immediately.

Close behind: Terrain Trek Grey Set — 4 active carts, 2-day runway, 1 unit remaining.

4 sources
write a briefGJ
S

Here is a restock brief you can pass directly to the ops team, based on live data as of the last 30 minutes.

Urgent Restock Brief

Generated: 4 May 2026 · Source:Flockr Demand Intelligence
Priority 1 — Order immediately (runway 2 days)

These products will sell out within ~48 hours at current sell-through rates. All have active demand.

  • Terrain Trek Grey Set (Quarter Zip / Short) — 1 unit remaining, accelerating demand, 4 shoppers holding it in their cart, 65 views in the last 24 hours.
  • Junior Utility Orange Set (T-Shirt / Short) — 1 unit left, 2 active carts. Demand is fading, but still worth a restock check given cart activity.
+ Priority 2 (runway 4–5 days) and Priority 3 (runway 7 days) sections continue below
4 sources
Follow-up resolution

“Write a brief” doesn’t say what about. Signal resolves it against the prior turn — the products at risk.

Live-data tools

Both turns call the scarcity-risk tool. Numbers reflect current state, not yesterday’s snapshot.

Generates assets

Signal produces work products, not just answers — briefs structured for the team’s workflow.

A note on the examples above. The conversation reproductions are illustrative excerpts of how Signal answers in the live portal — structured markdown, source pills, client-specific phrasing. Actual responses run on your knowledge base, your configuration, and your live data.

Operational layer

Built to hold up.

Three operational properties that keep Signal usable, accurate, and trusted — day to day, week to week.

01 · Performance

Streaming & caching

Responses stream word by word — no spinner, no waiting screen. Prompt cache and analytics cache (1h each) mean the second question is faster than the first, and your team’s queries are shared.

Stream · word-by-wordCache 1h
02 · Mode-aware

Knows live vs data mode

When a question depends on rendered-message data and you’re in data mode, Signal explains exactly why the figure is empty by design — rather than failing silently. It tells you the right next step instead.

Data→LiveSignal knows
03 · Current

Per-client knowledge

When the adaptive scoring model drifts beyond threshold, the knowledge base auto-updates. Signal stays accurate to what the engine actually does — governance built in, not bolted on.

In sync
~drift

Performance. Mode-awareness. Currency.

Ready to ask

Ask Signal your questions.

Signal is included in Demand Plus and Enterprise. Book a discovery call to see it answering questions about your catalogue, in your configuration, with your live data — in a few minutes, not a setup project.

Book a discovery call→Or read the pricing
Try asking SignalLive
KnowledgeHow is attribution calculated?
Live dataWhich products are losing momentum?
Follow-upWhy is scarcity winning slots on PDP?
AssetDraft a restock brief for the ops team.
And anything else aboutyour demand data