The Complete Guide to Ralph

Ralph is an AI eCommerce OS that reads your entire business — 21 live data sources, every product, every order, every campaign, every customer segment — and takes real, measurable actions. Not just dashboards. Not just recommendations. Actual actions.

What Ralph actually does

The simplest mental model: Ralph is a senior eCommerce manager, an analyst, a copywriter, an SEO specialist, a campaign manager, and a data scientist — all running 24/7 in your Shopify store. Here's what that looks like in practice:

01Quick Start

From zero to your first live campaign in under 10 minutes:

What happens during the first scan

Your first intelligence scan runs automatically on connection and builds the full foundation:

• Full product catalogue pulled with all variants, pricing, images, and metafields
• Order history analysed (up to 2 years on Growth, unlimited on Scale)
• 12-dimension product scores calculated for every SKU
• Product clusters built by affinity, margin, and velocity
• Customer segments modelled: LTV tiers, cohorts, churn risk, repeat purchase probability
• 10 opportunity detectors scored across your full catalogue
• Contextual baseline built: store type detected, audience profile inferred, sensitivity map created
• First morning briefing prepared (delivered at your configured time)

02Core Concepts

Reactive tool vs. proactive OS

Most analytics tools are reactive: they tell you what happened after it happened. Ralph is designed to be proactive — it identifies what's about to happen and acts before the window closes. When a bank holiday approaches with a favourable weather forecast, Ralph doesn't wait for you to check a dashboard. It surfaces a scored opportunity with campaign ready to launch, and — with Autopilot enabled — launches it automatically.

Every piece of data Ralph collects feeds into predictions, not just reports. Order velocity feeds demand forecasting. Search Console keywords feed content strategy. Weather forecasts feed campaign timing. Customer cohort data feeds email sequencing. Nothing is siloed.

The intelligence-to-action loop

Ralph runs a continuous intelligence cycle:

The Autopilot spectrum

Ralph operates on a spectrum. You decide exactly where every feature sits:

Each feature has its own toggle. You might run Campaign Monitor on full Autopilot (let Ralph revert underperforming campaigns) while keeping Campaign Launch on Gated mode (Ralph builds the full campaign but emails you for sign-off before going live).

03A Day with Ralph

To understand what Ralph actually does, here's what happens across a typical 24-hour cycle — most of it without you touching anything:

04Integrations

Ralph connects to 21 data sources across 12 platforms. Here's every integration and exactly what Ralph reads and writes through each one:

Connecting integrations

All integrations connect via OAuth — no API keys to copy-paste, no webhook URLs to configure manually. Go to Settings → Integrations, click the platform you want to connect, and authorise via the platform's standard OAuth flow. Ralph handles token refresh automatically.

05Voice Commands

Ralph understands natural language — you don't need to learn exact syntax. The examples below are starting points. Ralph interprets intent, so variations work too. Commands process in under 2 seconds for intelligence queries and trigger background jobs for campaign builds.

06Campaign Pipeline

When you say "campaign these products," Ralph executes a precise 16-step pipeline across three AI models, hitting Shopify, Meta, Google Shopping, Klaviyo, and your blog simultaneously. The entire process takes approximately 67 seconds.

What Claude Opus does vs. Claude Sonnet

Step 4 (strategy) uses Claude Opus 4 — the most capable model in the pipeline. Opus receives your full store context: top products with 12-dimension scores, customer segments, past campaign performance, weather forecast, retail calendar signals, competitor pricing where available, and current inventory levels. It returns a structured strategy document including which products to lead with, recommended discount depth, suggested messaging angle, and expected ROAS range based on comparable past campaigns.

Steps 8, 9, 12 use Claude Sonnet 4 — faster and cost-efficient for executional tasks. Sonnet receives the Opus strategy document as context, so every piece of copy, every email subject line, and every blog post is grounded in the campaign strategy rather than written in isolation.

Context injection — what every AI call knows

Ralph doesn't send bare prompts to Claude. Every call includes a rich context packet:

• Store type, audience profile, and sensitivity map (price-sensitive vs. premium)
• Current weather forecast and seasonal context
• Upcoming bank holidays and school term proximity
• Top 10 products with full 12-dimension scores
• Last 5 campaign outcomes with ROAS and conversion data
• Customer LTV tiers and top segment characteristics
• Active GSC keywords and content gaps
• Current Meta ROAS and best-performing audiences

This context means Ralph's AI output is specific to your store, your customers, and your current moment — not generic marketing copy.

Auto-revert & campaign safety

Every campaign that changes prices automatically schedules an auto-revert. You configure two revert triggers in Settings:

• Performance revert: If ROAS drops below your floor (e.g., 2.0x) for more than your configured grace period (e.g., 48 hours), Ralph reverts all price changes and notifies you.
• Scheduled revert: Regardless of performance, all price changes revert at campaign end date (default +14 days). Campaigns don't accidentally run forever.

07Morning Briefing

Every morning at your configured time, Ralph generates a personalised 6-segment briefing covering everything that happened overnight and everything you should focus on today. On Growth and Scale plans, it's delivered as audio via ElevenLabs TTS — synthesised in your choice of voice.

Customising your briefing

In Settings → Briefing you can configure delivery time (04:00–09:00), enable or disable segments, choose voice (6 ElevenLabs voices on Growth+), and set verbosity (concise 2–3 min or detailed 5–7 min). The briefing tab UI in the app is seekable — click any segment tab to jump to it, and replay individual segments.

08Autopilot

Autopilot lets Ralph take autonomous actions within rules you configure. Every feature is off by default and individually toggled. You can run full autopilot on some features while keeping others in supervised or gated mode.

09Intelligence Calendar

The Intelligence Calendar is Ralph's unified view of your past, present, and future — overlaying campaign performance, content publishing, retail event triggers, and AI-scored opportunity windows in a single month/week view.

What it shows

• Campaign timeline bars: Every live or past campaign shown with start/end, colour-coded by performance (green = ROAS above target, amber = borderline, red = auto-reverted)
• Content calendar: Blog posts published or scheduled, with traffic attribution shown after 7 days
• Retail trigger markers: Bank holidays, school term breaks, key retail dates, weather windows — each with an AI opportunity score
• Opportunity scores: Future dates with high opportunity scores are highlighted — click any day to see Ralph's full reasoning and launch a campaign directly from the calendar

Day detail panel

Clicking any day opens a panel with: exact opportunity score (0.0–1.0) with breakdown by signal (weather, calendar, product readiness, past performance), active campaigns on that day with hourly performance, blog posts published and their traffic, and a direct "Launch Campaign" button if an opportunity is ready.

10Product Clusters

Ralph groups your products into clusters automatically, updated every night. Clusters are the primary unit for campaign launches — you can campaign an entire cluster with a single command. Each cluster has a lead product, a collective opportunity score, and a recommended action.

Expanding a cluster

Click any cluster card to expand a full view: individual product intelligence scores, velocity sparklines, margin breakdown, campaign history, and opportunity score with reasoning. From the expanded view you can launch a campaign, add products, or manually re-assign a product to a different cluster type.

1112-Dimension Product Scoring

Every product is scored across 12 intelligence dimensions, updated nightly. These scores power clustering, campaign recommendations, morning briefing insights, opportunity detection, and every AI-generated piece of copy. The score for each dimension is 0.0–1.0.

12Opportunity Scanner

The Opportunity Scanner runs 10 parallel detectors every night across your full catalogue and data sources. Each opportunity is scored 0.0–1.0 and surfaced in your morning briefing, notification feed, and the Intelligence Calendar. Opportunities above your Autopilot threshold can be launched automatically.

How opportunity scores are calculated

Each detector scores 0.0–1.0 on a composite of: timing quality (is this the right moment?), product readiness (right products in stock?), historical evidence (has something like this worked before?), and risk factors (return risk, stock depth, cannibalisation). Scores are multiplied by a confidence weight based on data quality.

An opportunity score of 0.9+ means Ralph is very confident this will perform well and all conditions are aligned. Below 0.6 means the opportunity exists but conditions aren't optimal — Ralph surfaces it but with lower urgency.

13Blog Intelligence Engine

Ralph treats your blog as a commercial asset, not a content calendar obligation. The Blog Intelligence engine correlates every article with actual revenue, tracks which posts drive product purchases, identifies the content topics that convert — and generates new content targeting your highest-value search gaps.

How it works

• Landing site attribution: Ralph tracks the landing_site field on every Shopify order. When a customer arrives via a blog post URL and converts, that revenue is attributed to the post.
• GSC correlation: Ralph maps GSC keyword data to your blog URLs — showing you which posts rank, for which queries, and what traffic they're driving. Then cross-references that traffic with order data to calculate Revenue Per Visitor by content topic.
• Topic scoring: Blog topics are scored by: current search volume, your current ranking (or gap), estimated traffic potential, and revenue-per-visitor from similar content. The top-scoring topics are surfaced as content opportunities.
• Content gap detection: Queries with 500+ monthly searches where you have no content but stock matching products are flagged as high-priority content gaps. Ralph writes these posts automatically during campaign pipelines.

Article scoring

Every blog post in your store is scored on: total attributed revenue, revenue trend (growing or declining), search visibility trend, content freshness, and internal link depth. Articles with declining traffic but historically high revenue are flagged for a refresh.

14Contextual Sales Intelligence

Weather, school terms, bank holidays, your audience type — these things move more revenue than most store owners realise. A Sunday in late May with 20°C and a bank holiday Monday is not the same as a Sunday in February. Ralph knows the difference, and acts on it.

Every signal below gets packed into every Claude call Ralph makes — strategy, copy, campaigns, recommendations. Nothing generic. Everything tied to what's actually happening with your store, right now.

Context signals Ralph tracks

How contexts combine — a real example

Consider this scenario Ralph might detect on a Wednesday in late April for an outdoor/active store:

Ralph surfaces this on Wednesday morning with full reasoning and a "Launch Campaign" button. With Autopilot at Scale tier, it launches automatically and notifies you. The campaign runs Thursday–Monday, auto-reverts Tuesday.

Sensitivity map explained

The sensitivity map tells Ralph how your customers respond to price signals. It's calculated from: average discount depth on your top 20 past campaigns, price elasticity distribution across your catalogue, AOV vs. category median, and customer segment composition. A score of 80/100 (price-sensitive) means Ralph leads offers with the discount percentage. A score of 20/100 (premium) means Ralph leads with quality language and uses discounts only for velocity rescue, not brand campaigns.

Bank holiday × audience behaviour

Ralph has modelled bank holiday behaviour patterns by audience type:

• Family audience stores: Bank holiday weekend = +38% browsing, purchases skew towards experiences and outdoor/active. Peak purchasing window is Saturday morning.
• Fashion/gifting stores: Bank holiday = increased gifting intent, AOV up, longer decision cycles. Email open rates 22% higher Sunday evening.
• Outdoor/active stores: Directly correlated with weather forecast quality. Sunny bank holiday = highest opportunity score of the calendar year.
• Children's stores: School holiday proximity drives intent. Purchases accelerate 3–5 days before holiday start, not during — Ralph times campaigns to the buying window, not the event itself.

15Unified Cross-Channel Intelligence

Shopify shows orders. GA4 shows sessions. Meta shows ROAS. GSC shows keywords. None of them talk to each other — so you're left manually stitching tabs together trying to work out what's actually going on. Ralph kills that entirely. One model. Every channel. Everything cross-referenced.

The four intelligence tables

Product graph — what it enables

The product graph is the most powerful part of unified intelligence. Before Ralph, you'd need to manually cross-reference Shopify's bestsellers list with GA4's product views with Shopping's impression data with Meta's ROAS reports. Ralph does this automatically for every SKU and answers questions like:

• "Which products have high Shopping impressions but low conversion — landing page problem or pricing problem?"
• "Which products rank well in GSC but don't convert from organic — content problem or product problem?"
• "Which Meta product ads are driving the highest ROAS compared to their Shopify organic performance?"
• "Which products sell well on Meta but poorly on Shopping — audience mismatch or feed quality issue?"

Channel attribution matrix

Ralph builds a full attribution matrix for your store — showing revenue, ROAS, CAC, and LTV by acquisition channel. This informs where to invest next campaign spend and which channels are over or under-indexed for your store type.

Example data — your store's numbers will vary and are calculated from your actual channel data.

16Customer Intelligence & Segmentation

Most stores have customer data. Few actually use it. Ralph builds a live model of every customer — where they came from, what they bought, how likely they are to buy again, how close they are to churning, and how much they're worth over the next 12 months. That model runs in the background every night and feeds everything: your briefing, your email sequences, your campaigns.

LTV model

Ralph calculates predicted 12-month LTV for every customer using a probabilistic repeat purchase model built from your store's cohort data. Inputs include: order frequency, AOV trend, product category affinity, recency (days since last order), and acquisition channel. The model is recalibrated monthly as new order data flows in.

Customer segments

Churn prediction

Ralph calculates a churn risk score (0–100) for every customer daily. Score above 70 = high risk — customer appears in your morning briefing's stock/customers segment and is automatically added to your At Risk Klaviyo segment for win-back sequencing (with Email Autopilot enabled).

CAC by channel

Ralph calculates true CAC by acquisition channel, accounting for ad spend, email send costs, and Shopify transaction fees. This CAC is crossed against 12-month predicted LTV to give you a payback period and LTV:CAC ratio per channel — the most important metric for deciding where to spend your next campaign budget.

17Dynamic Pricing Engine

Discounting blind is one of the most expensive things a store can do. Too much, and you leave margin on the table. Too little, and the campaign doesn't move. Ralph calculates the right number for every product, every time — based on how that specific product actually responds to price changes, how much stock you need to shift, and what the seasonal context is pushing demand to do.

The pricing formula

For each product in a campaign, the pricing engine calculates:

// Simplified representation of Ralph's pricing model

optimal_discount = f(
  price_elasticity,      // 0.0–2.0 — how demand responds to price
  inventory_urgency,     // days_of_stock / target_days — how fast to move
  margin_floor,          // minimum acceptable margin % (your setting)
  competitor_delta,      // your price vs competitors where data available
  campaign_context       // seasonal uplift factor from contextual intelligence
)

// Example: Trail Runner Pro
// Elasticity: 0.82 | Stock urgency: 0.6 | Margin floor: 35%
// Seasonal uplift (bank holiday + sunny): +1.24x
→ Recommended discount: 12% → Expected volume lift: +34%
→ Net revenue impact: +£2,840 over 14-day campaign window

Inventory urgency pricing

When stock urgency is high (velocity rescue scenario), the pricing engine shifts from margin optimisation to volume optimisation. The floor is your configured minimum margin — Ralph will never discount below it. Above that floor, the optimal markdown is calculated to clear stock within the target window without leaving money on the table.

Mid-campaign dynamic pricing

With Autopilot enabled, Ralph monitors live demand signals during a campaign and adjusts prices dynamically within your configured bounds (e.g., max ±5% from campaign price, never below margin floor). If a product is selling faster than forecast, Ralph may reduce the discount to protect margin. If conversion drops, it may increase the discount within limits.

Price floors and ceilings

In Settings → Pricing, configure per-category price floors (minimum price Ralph will ever set), price ceilings (maximum campaign price — useful for premium brands that never want prices lowered below a brand threshold), and maximum discount depth per category. Ralph respects these in every automated pricing decision.

18Order Tracking & Intelligence

Ralph provides deep order intelligence — not just "how many orders today" but pattern analysis, fulfilment monitoring, abandoned checkout recovery, and anomaly detection that surfaces issues before they affect your customers.

Real-time order monitoring

Ralph maintains a live order feed via Shopify webhooks (created on setup). This means order data updates in seconds — not the 15-minute polling cycle of most integrations. The live feed powers:

• Velocity calculations: Hourly velocity updates during campaigns feed the dynamic pricing engine and campaign monitor
• Stock level sync: As orders come in, inventory levels update instantly — Ralph warns of stockout risk with real-time accuracy
• Revenue attribution: Orders are attributed to campaigns, channels, and customer segments in real time
• Anomaly detection: Unusual order patterns (sudden spike, sudden drop, unusual geographies, high-risk order signals) are flagged immediately

Abandoned cart recovery

Ralph listens to Shopify's checkouts/create and checkouts/update webhooks to detect abandoned carts. Three recovery triggers are configurable:

All recovery emails are written by Claude with full product context — not generic "you left something in your cart" templates. The product name, category-appropriate language, and your brand tone are all incorporated.

Order intelligence queries

Ask Ralph in plain English about your orders:

Fulfilment monitoring

Ralph tracks fulfilment status across all orders and surfaces delays in the morning briefing. If fulfilment time is increasing (e.g., warehouse processing time creeping from 1 day to 3 days), Ralph detects the pattern and alerts you — before customer service complaints arrive.

19Smart Product Filter

The Smart Product Filter is Ralph's AI-powered product search and filtering interface on the Products page. Unlike standard eCommerce product management (filter by name, SKU, or collection), Ralph's filter understands intent and intelligence-based criteria.

Filter by intelligence dimensions

Instead of "show me products in the Footwear collection priced above £50," ask Ralph to filter by business logic:

Filter combinations

The UI filter panel lets you combine multiple intelligence dimensions at once. Example filter set:

• Velocity score: above 0.6 (growing faster than seasonal norm)
• Stock risk: low (14+ days of stock at current velocity)
• Margin: above 35%
• Lifecycle: Growth or Maturity
• Campaign history: not in active campaign

This returns exactly the products Ralph would recommend for a growth campaign. You can then campaign the filtered set directly from the Products page.

Bulk actions from the filter

Once you've filtered to a product set, bulk actions are available: Campaign these products, Update SEO metadata, Add to collection, Export to CSV, Tag for review. All bulk actions respect your Autopilot settings — campaign launches can require your confirmation or run automatically.

20Rich Snippets & Schema Markup

Rich snippets are Google's enhanced search results — star ratings, price, availability, review count, breadcrumbs. They dramatically improve click-through rate from organic search. Ralph generates and maintains structured data (JSON-LD schema markup) for every product, collection, and blog post, automatically injected via Shopify metafields.

Before and after Ralph

Schema types Ralph generates

How Ralph injects schema

Ralph generates schema markup as JSON-LD and injects it via Shopify metafields — no theme editing required. The schema is maintained automatically: when a product's price changes (e.g., campaign discount goes live), Ralph updates the Offer schema to reflect the sale price and expiry date. When a campaign ends and prices revert, the schema updates back.

21Weather Intelligence

Weather drives more purchase intent than most store owners credit it for. It's not just outdoor gear — knitwear, homewares, gifts, activewear all have measurable weather sensitivity. Ralph works out your store's exact sensitivity coefficients from historical data, then watches the forecast and acts on it before the window opens.

Building your weather sensitivity map

On first setup, Ralph analyses your last 24 months of order data against historical weather data for your store location. For every product category, it calculates a weather sensitivity coefficient — how much does demand change per degree of temperature change? This produces a sensitivity map like:

Live forecast integration

Ralph queries the Weather API hourly for the next 7 days. When a forecast is detected that crosses a sensitivity threshold for one of your product categories, and stock depth is sufficient to support a campaign, the Weather Windows opportunity detector fires — scoring the opportunity and surfacing it in your briefing.

Multi-region support

If your customers are distributed across the UK (or internationally), you can configure up to 5 regional forecast locations. Ralph weights the forecast signal by your customer geographic distribution — a sunny forecast in London means less for a store with 70% Scottish customers than one serving primarily southern England.

22Reporting & Analytics

Forget switching between Shopify analytics, GA4, Meta Business Suite, and Klaviyo reports. Ralph pulls everything into one place, correlates it, and tells you what it means. Every report comes with Ralph's interpretation above the chart — because numbers without context are just noise.

Report types

Report customisation

Every report supports custom date ranges, comparison periods (previous period, same period last year, or a specific custom range), and filtering by product, collection, channel, or customer segment. Reports can be scheduled for weekly or monthly email delivery in PDF format.

AI-interpreted insights

Above every significant chart, Ralph adds a natural-language interpretation: "Revenue is up 27% vs. last month, driven primarily by the Spring Outdoor campaign (£8,400 attributed). However, organic conversion rate has dropped from 2.4% to 1.8% — which suggests the campaign traffic may be lower-intent than your baseline organic visitors. Worth reviewing landing page alignment." This turns data into direction.