Flockr Deployment & Integration

Going live

Deployment & integration

Two technical inputs from your side — a browser tag and a catalogue feed. The rest is what Flockr runs. Going live is a phased sequence with a deliberate calibration period before any messages render to shoppers.

Two inputs. Two modes. A calibration period that earns its keep.

Block 2 — what you do, what Flockr runs

Division of responsibility

Two technical inputs from your side. Everything downstream — signal computation, scoring, message selection, attribution, threshold tuning — is what the engine and the Flockr team run.

Your side

Browser tag installed — one async JS snippet on every shoppable page.
Catalogue feed connected — via standard ecommerce connector or custom feed.
Active surfaces approved — PLP, PDP, cart, search, recommendations — toggleable per surface.
Mode toggle — flip from data mode to live mode in Settings when calibration is complete.
Copy review — sign off on message templates before live mode.

Flockr's side

Signal computation — per product, per request, across nine time windows.
Scoring & selection — ranks competing candidates per slot, picks the strongest.
Threshold tuning — signal eligibility floors calibrated to your traffic.
Adaptive scoring — weights tuned continuously from observed conversion data.
Attribution model — the two-group session-level computation, three lift measurements.
Knowledge base maintenance — the source of truth Signal answers from, kept current with engine behaviour.

Block 3 — the two technical inputs

A browser tag and a catalogue feed — nothing else

No SDK in your application code. No data warehouse integration. No scheduled job to maintain. The tag captures shopper behaviour; the feed keeps product truth current.

Input 1

The browser tag

One async JS snippet, loaded once on every shoppable page. Non-blocking; never delays the page load. Captures shopper behaviour and renders messages when the engine returns them.

Captures — page views, product impressions with viewport visibility, ATBs, purchases, click events
Fires — Flockr requests on qualifying pages, returns messages in milliseconds
Reversible — removing the tag stops all activity; no residual state in shopper sessions

Input 2

The catalogue feed

Product data ingested via standard ecommerce connectors or a custom feed. Refreshed continuously. Becomes the engine's product truth — what's in stock, what's available, what counts as a new arrival.

Connectors — Shopify and major ecommerce platforms; custom feed for bespoke stacks
Refresh cadence — periodic full sync plus low-stock and inventory webhooks for fast updates
Includes — SKU, title, price, inventory state, status, category, lifecycle attributes
Used for — product truth, scarcity calculation, lifecycle classification, eligibility checks

Block 4 — the integration sequence

Five phases from setup to steady state

Going live happens in stages. Data mode runs first — the engine is computing signals, not showing messages. The mode flip happens once thresholds are calibrated to your actual traffic.

Setup

Tag installed. Catalogue feed connected. Portal access provisioned for your team.

Data mode

Calibration

Engine collects signal data. Thresholds tune to your traffic patterns. No messages render to shoppers.

Data mode

First surface live

Mode flips. A single browse surface (typically PLP) starts rendering messages. Attribution events begin firing.

Data → Live

Surface rollout

Additional surfaces enabled progressively — PDP, search, recommendations, cart. Each surface independently toggleable.

Live mode

Steady state

All approved surfaces live. Adaptive scoring engaged. Attribution data accumulating across the full message footprint.

Live mode

Why a calibration period. Signal eligibility floors depend on traffic. Hardcoded defaults would either be too restrictive (no messages on smaller stores) or too permissive (irrelevant messages on larger ones). The data-mode phase tunes thresholds to what's actually happening, so the first message a shopper sees is one with a genuinely strong signal behind it.

Block 5 — ongoing operations

Continuous

Adaptive scoring

The scoring engine tunes signal weights per surface and per family from your store's actual conversion data. Operators see the result reflected in Settings → Scoring (read-only). Signal can explain why weights shifted when asked.

Weekly

Review & opportunity surfacing

The Flockr team reviews activation, attribution, and signal mix each week. Scarcity that's slipping. Overstock with promotional potential. Surfaces that are underperforming. Surfaced to a designated point of contact rather than buried in a dashboard.

On demand

Self-serve & direct support

Signal handles most operational questions in-portal — configuration explanations, performance figures, briefs to share with the team. For genuine engineering questions, a direct channel to the Flockr team. The portal's own activity is itself the audit trail.

A note on the calibration period. Data mode is not a wait state. Signal data is being collected, product truth is being built, lifecycle classifications are settling, threshold floors are being tuned to your traffic. By the time messages start rendering, the engine has weeks of behavioural evidence behind every selection it makes — not generic defaults applied blindly. The mode flip is the moment shoppers begin to see the work; it isn't the moment the work begins.