Skip to main content

Growth Intelligence overview

In plain English

Daily Ops tells you what to do. This section tells you what's going on and why.

It's split into five parts, and you can think of them as five questions:

SectionThe question it answers
TodayWhat should I do right now?
RevenueWhy did sales go up or down?
CustomersAre the customers I'm getting any good?
MarketingWhich of my marketing is actually working?
Goals & AlertsAm I going to hit my targets?

You don't need to visit all five. Most days, "Today" is enough. Come to the others when you have a specific question — usually "why did that number move?"

One thing to know about the numbers here: they're deliberately cautious. When the app estimates you could recover £3,000, it means £3,000, not a hopeful £10,000. Each estimate also comes with a confidence label — high, medium, or low. Trust the high ones. Treat the low ones as ideas worth testing, not plans.

How to use it

  1. Click "Intelligence" in the left-hand menu. It opens on "Today."
  2. Work the five actions listed there. That's the main job — most days you can stop here.
  3. Choose your time period at the top — 7 days, 30 days (normal), or 90 days.
  4. If a number surprises you, click "Revenue" and read the breakdown. It tells you which channel or which change caused it.
  5. Once a quarter, click "Customers" to check whether newer customers are as good as older ones.
  6. Set a target under "Goals & Alerts" so the app can warn you early if you're going to miss it.

⏱ ~15 min/week · 💳 Pro+ · 🎯 Understand the business, then act on it

Why this matters for your business

Most analytics tools answer questions you didn't ask. They open on a wall of charts, each technically informative, none of them telling you what to do. You leave knowing revenue is down 4% and still not knowing whether that's acquisition, retention, a channel problem, or a Tuesday.

Growth Intelligence is built the other way round. It opens on the five highest-leverage things you could do today, ranked by estimated impact divided by effort. If you only ever look at that screen, the tool has done its job. Everything behind it — attribution, cohorts, elasticity, forecasts — exists to make that first screen correct, and to be there when you need to understand why.

The second design principle is honest uncertainty. Every estimate here carries a confidence label, every projection states its assumptions, and the numbers that can't be computed properly say so rather than guessing. Customer-acquisition cost, for instance, is labelled "ads only" because the platform can see ad spend and influencer fees but not your agency retainer — so it under-counts, and it tells you it under-counts. A tool that quietly reports a flattering CAC is worse than one that admits the gap.

What this typically unlocks

OutcomeTypical result
Time to "what should I do today?"Instant — it's the landing screen
"Why is revenue down?" investigation~2 minutes instead of a half-day
Quantified opportunities you're not acting onContinuously surfaced, with $ and confidence
Retention visibilityFull cohort matrix, not a single repeat-rate number
Goal trackingProbability of hitting target, not just progress-to-date
Weekly reporting effort−80% — the digest composes itself

What you actually get

Sixteen surfaces, grouped into five sections. The section is a navigation convenience; the URL always carries the specific surface, so any view is deep-linkable.

SectionSurfacesWhat it answersDeep dive
TodayAction Center, SnapshotWhat should I do right now?Action Center
RevenueRevenue, Attribution, Forecast & AIWhere did revenue come from, and where is it going?Revenue command center
CustomersCustomer Segments, Customer IntelligenceWho buys, who stays, who's worth acquiring?Customer intelligence
MarketingBoost Products, Keyword Routing, Growth Score, Blog Distribution, Influencer SEO, A/B Winners, Campaign Calendar, Competitor AlertsWhich marketing is working?Marketing signals
Goals & AlertsGoals & AlertsWill I hit my targets?Goals & alerts

The recoverable band

Pinned above the navigation, on every surface, is the recoverable money total — the same honest figure described in Revenue Leaks. It's there deliberately: it's very easy to spend an hour in an analytics tool and forget there was money sitting on the table the whole time.

Time windows

WindowBest for
7 daysReacting to something that just changed
30 days (default)Normal operating rhythm — enough signal, still current
90 daysTrend and seasonality work, cohort analysis

Every window comes with a matched prior period for comparison. A 30-day view compares against the previous 30 days, not against "last month" — so a 31-day month never appears to beat a 28-day one on arithmetic alone.

Freshness and failure

The hub shows when its data was last updated and gives you a manual refresh. When a panel's underlying engine fails, you get an explicit error rather than an empty state that reads as "you have no data" — an important distinction when the difference is "nothing to report" versus "we couldn't check."

How it works (without the technical bits)

Rules, not black boxes

The action queue is nine hand-built rules, not a machine-learning model. That's a deliberate trade-off, and worth understanding because it shapes how much you should trust the screen.

A model might rank marginally better. But this is the "do this now" surface, and a recommendation you can't interrogate is a recommendation you won't follow. Every item here states the condition that triggered it and the arithmetic behind its estimate. When the tool says "Reorder 4 SKUs at stockout risk — est. $3,200", you can check the four SKUs and the maths. When it's wrong, you can see why it's wrong, which is what lets you trust it the next time.

AI is used where it genuinely helps — a three-line plain-English narrative summarising the dashboard — and not where it would obscure a calculation you need to audit.

Conservative by construction

Every estimate in this hub is built to under-promise:

EstimateRate appliedWhy
Cart recovery10% of open cartsWhat recovery sequences actually return
Dormant winback8% reactivationReactivation is hard and most tools overstate it
Safe price increase5% on inelastic SKUsAssumes minimal volume loss
Cross-sell uplift15% AOV, capped at 30% of ordersPrevents claiming cross-sell touches every order

The cap on cross-sell is the clearest example of the philosophy. Without it, a shop with 40 high-confidence product pairs would show an opportunity implying cross-sell influences the entire order base. The cap makes the number smaller and true.

How this differs from Daily Ops

They're complementary, and the boundary is clean:

Daily OpsGrowth Intelligence
PurposeExecuteUnderstand, then execute
ContentFindings from every engine, one ranked queueAnalysis, plus its own action queue
Ranking unitMoney at stake ÷ effortEstimated impact ÷ effort
DepthOne line per findingFull driver decomposition
Use it whenWorking the daily listSomething needs explaining

If you only have ten minutes a day, use Daily Ops. Come here when a number needs a reason, when you're planning a quarter, or when you want to know what you're leaving on the table.

Real merchant scenarios

Scenario A — "Revenue is down 11%" answered in 90 seconds

Setup. Home-fragrance brand, revenue down 11% month over month. The founder's assumption was ad-market cost inflation.

Revenue diagnosis showed:

DriverContribution to the drop
Email channel revenue−73% of total drop
Volume (orders) vs AOVOrders down 14%, AOV flat
Repeat rateDown 6 points
Anomalies in period1 critical

Root cause. The critical anomaly was a deliverability drop — their sending domain had been flagged three weeks earlier. Email revenue had collapsed, and email was their retention channel, so the repeat rate fell with it.

Not an ad problem at all. The paid channel was flat. Had they acted on the original assumption, they'd have cut ad spend and made the quarter worse.

Scenario B — The opportunity nobody had quantified

Setup. Coffee subscription business, 2,800 active subscribers.

Opportunity radar:

OpportunityEstimateConfidence
Subscription dunning recovery$8,400High
Cart recovery$3,100High
Dormant winback$2,700Medium
Cross-sell pairs$1,900Low

The surprise. Dunning was the largest and they hadn't known it existed as a category — failed renewals had simply looked like churn. High confidence, because the estimate is mechanical: failed contracts × their actual monthly price. No modelling assumption at all.

Result. Dunning sequence built that week. Recovered $7,600 of the estimated $8,400 over the following month.

Note the ordering of confidence. Cross-sell showed the smallest number and the lowest confidence, correctly — it rests on an affinity model and an uplift assumption, where dunning rests on arithmetic.

Scenario C — Cohorts contradicting the headline

Setup. Apparel brand, repeat rate flat at 31% for six months. Read as "retention is stable."

Cohort matrix said otherwise:

Acquisition monthMonth 3 retention
November38%
December34%
January29%
February24%
March21%

What was happening. Retention was deteriorating steadily. The flat aggregate was arithmetic coincidence — a growing number of new customers, each retaining worse, averaging out to the same number.

Cause. A discount-led acquisition push starting in December had brought in progressively more deal-driven customers. LTV by channel confirmed it: discount-acquired customers had roughly 40% of the lifetime value of organic ones.

Action. Shifted acquisition mix back toward organic and referral. Month-3 retention on subsequent cohorts recovered to 33% over the following quarter.

Why the aggregate lied. A single repeat-rate number cannot show a trend across cohorts. This is the canonical case for cohort analysis, and it's invisible on any dashboard that reports one number.

Scenario D — A goal that was never going to happen

Setup. Merchant set a quarterly revenue goal of $400,000. Six weeks in, they were at $164,000 — 41%, which felt roughly on pace.

Feasibility forecaster:

Goal $400,000 revenue · quarter
Current $164,000 (41%)
Projection $355,000
Probability 18% → unlikely
Confidence high (42 daily samples)

Why "unlikely" when it looks close. Linear pacing puts them at $355,000 — 89% of target, in the at-risk band. The probability model accounts for daily variance and finds only an 18% chance of closing an $45,000 gap in six weeks at their observed run rate and volatility.

Action taken at week 6, not week 12. They pulled a planned Q4 promotion forward. Final: $391,000 — still short, but $36,000 better than the projection, and the shortfall was known and communicated in advance rather than discovered at quarter end.

Best practices

Land on Today and stay there most days. The deeper sections are for questions, not routine.

Use 30 days as your default and 7 only for incidents. Seven-day windows are noisy enough to invent trends that aren't there.

Read confidence labels before acting on a number. A low-confidence estimate is a hypothesis to test, not a plan.

Set at least one goal. The tracker and feasibility forecaster do nothing without a target, and they're the two surfaces most likely to change a quarter's outcome.

Diagnose before reacting to any revenue move. The decomposition takes two minutes and regularly contradicts the obvious explanation.

Turn on the weekly digest. It's the same analysis, delivered — which means it gets read.

Don't treat "ads only" CAC as your true CAC. It excludes agency fees, content, and tooling. Use it for trend, not for absolute unit economics.

Don't sum the opportunity radar and call it a forecast. They overlap, and each is an upper bound on its own assumption.

Don't read a flat aggregate as a stable trend. Check the cohort matrix — Scenario C is common.

Don't set a goal you can't measure daily. Repeat rate and AOV goals produce lower-confidence forecasts because they don't bucket cleanly by day.

Plan tiers

CapabilityFreeStarterProAgencyEnterprise
Action Center + opportunity radar
KPI strip with prior-period deltas
Revenue diagnosis
Attribution breakdown
Cohort retention + LTV by channel
Marketing signal surfaces
Goal tracker + feasibility forecast
AI executive narrative
Weekly executive digest
Multi-store roll-up

Frequently asked

How is this different from Daily Ops? Daily Ops is the execution queue across every engine. This is the analysis — why numbers moved, who your customers are, what you're leaving on the table. Daily Ops tells you what to do; this tells you what's going on.

Why does the hub open instantly but some tabs take a moment? The landing screen needs no heavy aggregation. Surfaces that analyse attribution and product intelligence run real scans and load on demand, so opening the hub is never slow.

Why is my CAC lower than I calculate manually? It counts ad spend and influencer fees — the costs the platform can observe. Agency retainers, content production, and tooling aren't included, which is why it's labelled "ads only."

Can I change the estimate rates? No. Fixed conservative rates keep estimates comparable over time and across stores. A tunable rate is one that eventually gets tuned until the dashboard looks good.

Does the AI narrative make decisions? No. It summarises what the deterministic engines already computed, in three lines. Every number it mentions is independently visible and auditable on the dashboard.

How far back does cohort analysis go? Twelve months of acquisition cohorts by default.

Why do two surfaces show slightly different revenue? Check the window and the basis. Attributed revenue (credited to a channel) and gross revenue differ legitimately. Each surface labels which it's showing.

Can I export this? Yes — the weekly digest delivers by email or Slack, and the underlying metrics are available programmatically via the Admin API.

See also