The intelligence layer
for modern marketing

ONYX is the intelligence layer between fragmented marketing systems and coordinated execution. Synthesizing marketing and commercial data to help teams prioritize what matters most.

Interactive preview. Click through the modules in the sidebar.
app.onyxintel.co/overview

Overview

All channels · Last 90 days
● Live Sync
Customer base value · 90d
$3,071
Avg customer lifetime value
Retention alignment
71%
Outreach → highest-value segments
High-value repeat mix
27.3%
▲ 9.9 pts over 2 months
Campaigns active
8
4 above baseline
4 below baseline
Customer value trend 90d
● Customer value ($)◦ Repeat mix (%)
DIRECTIVE ▸ Value and repeat mix climbed in step over 90 days: retention, not acquisition volume, is driving the gain. Keep budget weighted toward the repeat engine.
Engagement-to-repurchase gap By stage
01 · Landing
$544K
02 · Browse
$324K
03 · Detail
$226K
04 · Checkout
$69K
05 · Confirmed
$0
PEAK DROP-OFF ▸ Landing loses $544K. A directive is attached at every stage.
ONYX InsightImproving

High-value repeat mix up 9.9 pts over 2 months. The Repurchase Sequence is the primary driver at $7,518 avg customer value; the Seasonal Push is the drag at $1,125. Scale the first, reduce the second, and the portfolio baseline moves with it. See full analysis ▸

Weekly Brief

Auto-generated every Monday
7findings7actions3watches
ONYX InsightCritical

The Repurchase Sequence produces customers at $7,518 avg LTV 90, 2.45x the portfolio average, on 16% of the monthly budget. A third of its own cohort has gone dormant.

The findingsThe actionsThe watchlist
1
Budget share vs. LTV index vs. dormancy rate within cohort
16% budget share at 2.45x LTV index · 33.9% of cohort dormant · 80.1% above portfolio conversion avg
HIGH
The signal

Holds 16% of the $53,000 monthly budget while producing $7,518 avg LTV customers with a 100% repeat purchase rate and zero one-and-done outcomes.

Next step

Reallocate toward this campaign and deploy a retention sequence for the 22 dormant repeat buyers before they cross the dormancy threshold.

2
High-Value Upsell Trigger: a confirmed retention engine at the lowest cost per order
$4,382 avg LTV 90 at $100.17 cost per order · LTV nearly doubles from 90d to 180d
HIGH
3
Budget misallocation: 32.5% of spend producing below-average LTV customers
$17,225/month allocated to campaigns averaging $1,332 avg LTV 90 · 49.9% below portfolio avg
ACT
See all 7 findings ↓

Segments

Segment intelligence · Momentum
● Live Sync
● Accelerating · 14 segments● Declining · 6 segments
San Francisco
REPEAT PURCHASERS
90D96
30D23
7D4
AVG ORDER $1,328CANCEL 5.9%
New York
FIRST-TIME BUYERS
90D12
30D4
7D0
AVG ORDER $1,530CANCEL 8.1%
Market LTV ranking 90d avg customer value
San Francisco
$6,920 · +125.3%
New York
$5,361 · +74.6%
Chicago
$5,278 · +71.9%
Austin
$3,260 · +6.2%
Untapped market signal

San Francisco repeat purchasers hold an estimated $151K in 90-day recoverable revenue at current momentum, yet allocation sits 6.4 pts under their value weight. New York first-time buyers are the first watch: zero orders in 7 days against 12 in the prior 90. See all 8 markets ▸

Ask ONYX

Plain-English answers from your connected data
Evidence shown on every answer
VP
Why did blended CAC rise last week even though Meta CAC improved?
O
Blended CAC rose 6.3% to $43.90 because your channel mix shifted, not because any channel got worse. Meta CAC improved 4%, but its share of new customers fell from 58% to 46% as spend rotated into Google PMax, where CAC runs $12 higher. Net effect: mix drove +$2.61 of blended CAC.
Meta
$38.10
PMax
$50.40
TikTok
$44.70
VP
Should we rebalance back toward Meta?
O

Campaigns

Campaign intelligence · Allocation vs. value
8 active
30-day allocation
$57,760
8 campaigns active
Portfolio avg LTV
$3,071
90-day avg customer value
Top value campaign
Repurchase Sequence
$7,518 · 2.45x · underinvested
Spend alignment
2 under 6 aligned
Realloc opportunity ~15%
Reallocation signal detected

Shift ~15% from the Seasonal Push toward the Repurchase Sequence to align allocation with the highest-value retention channel. +$6,371 avg LTV per shifted conversion.

Repurchase Sequence
Automated · repeat buyers
Underinvested$7,518LTV 90D16.0%of spend
Follow-up Nurture
Personalized · engaged buyers
Aligned$3,308LTV 90D11.5%of spend
Seasonal Push
Broadcast · full list
Reframe$1,125LTV 90D17.5%of spend
PORTFOLIO BASELINE$230 avg cost/order$3,111 avg LTV 901.01x avg LTV mult

Cohorts

Audience intelligence · Value by behavioral segment
Updated 3 hr ago
Primary segment
High Frequency ICP
53 accounts identified
90-day LTV
$8,426
+174.4% vs portfolio
Avg order value
$1,190
Within ICP cohort
Avg lead time
48.2d
First touch to order
Cohort value index 90d LTV
High Frequency
$8,426 ·
Dormant Repeat
$2,619 · ▲ 12.2%
Active Repeat
$2,290 · flat
One And Done
$1,185 ·
One And Done grew 11% this month · second-purchase conversion is the largest LTV lever
ICP session fingerprint
Device
desktop
Channel
email
Landing
/best-sellers
HF identified
53
ONYX Directive

High Frequency customers represent 53 accounts generating $8,426 avg LTV, and current outreach is not optimized for this profile: it is retaining statistically lower-value customers. Target the fingerprint above as the quality benchmark for every acquisition campaign. View ▸

Forecast

Revenue forecast · 60-day projection
1,215 customers · 12 campaigns
Current value
$3,071
Baseline trajectory $2,941 · −4.2%
Optimized projection
$3,229
+5.1% uplift with recommended moves
Portfolio impact
+$115K
Cost of inaction −$14,434 /week
60-day projection summary 2 actions modeled
Baseline · no action taken$2,941 · −4.2%
Optimized · recommended moves executed$3,229 · +5.1%
Delta captured per customer+$288
Portfolio impact at current volume+$115K
Why is baseline declining?What drives the optimized scenario?Ask about these projections ▸
1
Increase allocation toward the High-Value Upsell Trigger from 8.2% toward 12 to 14% of budget
+2.8% portfolio LTV · 64% confidence · cost of inaction −$8,599 /week
+2.8%
2
Pilot a retention sequence for active repeat customers approaching the purchase 11 cliff
+1.9% portfolio LTV · 56% confidence · 50% dropout observed at purchase 11 without intervention
+1.9%
The intelligence layer

From metrics
to directives.

Weighed against every shift, contextualized to your business, and synthesized into action.

Metrics from your business
AOV ▼ 6.2% Transaction volume ▲ 2.4% Repeat rate day 26 → 19 New-customer CAC ▲ $7 Seattle LTV ▲ 34% Dormant cohort ▲ 11%
ONYX The intelligence layer
Actionable directives
Cohort Directive

Shift $18K/wk to the repeat-purchaser lookalike layer.

+$64K/mo projected · 91% confidence
Verified

Cohort directive measured +$71K against +$58K projected.

Sees every signal Explains every shift Prescribes the next move Proves it worked
Inside the intelligence layer

How signal becomes execution.

Each layer is engineered to understand a different dimension of what drives demand. Compounding continuously, resolved into one decision at a time.

The unit of execution

Behind
every directive.

The cause traced, the action sized, the deadline fixed, the return verified. No interpretation. Just execution.

ONYX Directive
Dormancy watch · Repeat purchaser cohort
122 of your highest-value customers are going dormant. Each carries $4,509 in average LTV, with 41 days remaining until full dormancy.
2Launch the win-back sequence with the replenishment offer, targeted at all 22, ahead of the day-117 threshold. Projected recovery +$42K. 84% confidence
Timeframe within 7 days Owner: Lifecycle Cost of inaction −$7,171 /wk
Issued Mon Assigned Executed Tue Verified Day 30
311 of 22 customers reactivated within 30 days: +$49.6K recovered against $42K projected.
See Full AnalysisView Cohort
Suggested
Why are these customers going dormant?What did past win-backs recover?What happens if we wait? ▸
1

The Why

Built entirely from your numbers: the exact customer segments, campaigns, and timeframe to act. Every directive begins with the reason.

2

The Action

An order of operations for a single outcome. Each directive connecting revenue back to its source and signal.

3

The Outcome

Every result measured against its projection and prior performance, then fed back into the engine, each directive compounding from the knowledge gained.

The operating model

From fragmented data
to execution

One cycle, followed end to end, from the moment your numbers connect to the moment action is taken and verified.

Click each stage to see what your team would see.

01

Connect

Every data source, one model of your business.

02

Understand

Every shift traced to its cause and its cost.

03

Direct

Every directive sequenced by impact.

04

Compound

Every result measured, every cycle smarter.

Stage 01

Connect. Native connections to the tools in your operation, backfilled with up to 24 months of history. Every identity, order, and touchpoint resolved into one model of your business — learning the benchmarks first: seasonality, volatility, and promo rhythm for every channel and segment.

Meta
Google Ads
Shopify
HubSpot
Google Analytics 4
+ more
Customers resolved
1,215
Transactions matched
2,613
Signal types armed
40+
What ONYX learned before its first alert
Your buying rhythm. How often customers come back, by segment and by season, so timing is never guessed.
Your normal. The expected range for every metric on every channel, so an alert only fires when something truly breaks it.
Your value map. Which customers, campaigns, and markets actually carry the LTV, so every finding is priced in dollars.
24 months of history backfilled · one definition of CAC, LTV, and ROAS everywhere
Stage 02

Understand. Nothing is read in isolation. When a number breaks its baseline, every layer reads the same moment: Detection isolates where, Behavior shows whose patterns are bending, Value prices what it costs. Three reads, converging on one cause.

Blended ROAS ▼ 9% Isolated to Meta prospecting One ad set, frequency 6.8 Root cause: creative fatigue
Signal Engine

Blended ROAS broke its expected range on Tuesday. Decomposition isolates 82% of the move to a single ad set: frequency 6.8, CTR down 41% over 9 days.

Behavioral Model

The audience behind that ad set is your repeat-purchaser seed. Their observed patterns say fatigue here reaches second orders within two buying cycles.

Value Engine

At $8,426 average LTV against a $3,071 portfolio average, the exposure is concentrated in your most valuable customers, not your cheapest clicks.

Converged findingHigh
Creative fatigue · "Spring Hero v3" · Core audience
One cause, confirmed three ways, priced in dollars. Two tested variants are already staged, so the fix exists before the finding ever reaches your team.
Stage 03

Direct. Signals become directives: the specific move, the projected return, the owner, and the deadline. They arrive ranked by what they're worth, weighed across everything the model holds at once — timing windows, cohort behavior, funnel structure, and spend.

01 · Creative
Swap "Spring Hero v3" for variant B and cap frequency at 4, before Friday's spend refresh.
Owner: Paid Social · 89% confidence · cost of inaction −$1,285 /day
+$9K/wkprojected
02 · Market timing
47 Miami customers pass the day-25 intervention point this week. Send the repurchase sequence ahead of the day-32 median.
Owner: Lifecycle · 78% confidence · window closes in 6 days
+$21Kprojected
03 · Funnel
41.7% of sessions exit at step 2 with $324K of LTV upside at stake. Rebuild the step before scaling spend into it.
Owner: Growth · 71% confidence · highest-value segment converts lowest at this step
$324Kat stake
Three of seven directives this week, ranked by projected impact · every one carries its forecast, its owner, and its deadline
Stage 04

Compound. Your team executes; ONYX keeps measuring. Every directive is weighed against its own forecast inside a defined window, and the result feeds back into the engine. The next decision is sharper than the last, and the track record lives in your account.

Creative swap · Verification window 14 days
Projected
+$9.0K/wk
Measured
+$10.7K/wk
Measured on your revenue, not on platform-reported numbers
Verified · 119% of forecast
Recovery beat the projection by 19%. Your team sees the result on the directive itself, the day the window closes: no status meeting required.
What the engine learned back
The fatigue threshold moved. This audience breaks earlier than the old model assumed: 6.8 frequency recalibrated to 5.9, so the next alert fires sooner.
The confidence scoring sharpened. This outcome joins 214 similar calls that inform how strongly the engine backs its next creative directive.
The track record grew. Portfolio-wide, verified outcomes hold 87% within forecast, and that record lives in your account, not in a case study.
Get started

Deploy your
intelligence layer.

Built around your business. Redefining how your team operates.

Platform

The system built to prioritize.

ONYX sits underneath your existing operation, connecting every source into one intelligence layer built for execution.

Click through the modules to explore each layer.

Prioritize

Directives, not dashboards.

Every week arrives already decided: each movement traced, each shift priced, each directive ready to be executed. Nothing left to interpret.

  • Top signals flagged by urgency, sized in dollars
  • The reasoning connected under every signal
  • Signals converted into a numbered queue of actions
  • Each action tied to its projected outcome
Understand

Every metric shift, traced to its source.

Behavior, attribution, channels, and segments, brought together and evaluated continuously. Every movement connected to its source, its size, and its meaning.

  • Real-time detection across every input in the model
  • Patterns recognized at the cohort and segment level
  • Every break in behavior classified by what drove it
  • Ranked by severity and the revenue at risk
Interact

The context behind every signal.

Ask why a metric shifted, where a trend started, or what to do about a segment. Every answer arrives grounded in your connected data, carrying its source, its reasoning, and the next action to take.

  • Answers built from your live, connected data
  • Each one traceable to the metrics and queries behind it
  • Each answer carries the full chain: what moved, why, and what it cost
  • Every answer from the same source of truth your directives run on
Measure

Campaigns, judged by long-term value.

ONYX measures your portfolio against the long-term value each campaign produces, following customers well past their first purchase. Each verdict is built on more than just performance: how those customers behave, what patterns they hold, and what each signal means for allocation.

  • Portfolio verdicts built on realized customer value
  • Circular spend caught: budget that recycles customers instead of creating them
  • Contribution read independently across the full journey
  • Budget scenarios priced before a dollar moves
Interactive. Click the modules in the sidebar.
app.onyxintel.co/weekly-brief

Weekly Brief

This week's signals and the queue they created
4 signals · 5 actions
Top signals this week
Budget misallocation: 32.5% of spend producing below-average LTV customersACT
THE SIGNAL: two campaigns hold $17,225/month while the portfolio's best performer stays underfunded at a 16% share…
Budget share vs. LTV index vs. dormancy rate within its own cohortHIGH
THE SIGNAL: 16% of budget, 2.45x portfolio LTV, and 33.9% of the cohort already dormant…
Session fingerprint: the 53 highest-LTV customers share a distinct behavioral profileHIGH
THE SIGNAL: a repeatable pattern across device, entry path, and pre-purchase behavior…
Priority actions
01Reallocate budget from the two lowest-LTV campaigns toward the proven retention pairAllocation
02Adopt the high-frequency session fingerprint as the targeting and quality benchmarkTargeting
03Deploy reactivation for the highest-probability dormant customers before the window closesLifecycle + Ops

Segments

Displacement signals inside your audience
● Live
Competitor displacement
57
established customers displaced
Signal strength
22 · 35
high signal · moderate signal
The pattern

Customers in this segment held established repurchase rhythms, then broke them in a way that aligns with competitor displacement rather than passive drift. Roughly two in five show a break strong enough to suggest an active comparative choice was made.

Why it matters

Standard reactivation messaging is calibrated for customers who lapsed through inertia, not customers who evaluated alternatives and chose one. If the displacement hypothesis holds, discount-led outreach signals weakness at exactly the moment operational confidence is being judged.

The action

Outreach for this cohort leads with operational proof points, not promotional framing. The 22 high-signal customers come first, before consolidation with a competitor deepens.

Ask ONYX

Answers with evidence
VP
Why did blended CAC rise last week even though Meta CAC improved?
O
Blended CAC rose 6.3% because your channel mix shifted. Meta improved 4%, but its share of new customers fell from 58% to 46% as spend rotated into PMax, where CAC runs $12 higher.
Meta
$38.10
PMax
$50.40
VP
Should we rebalance?
O

Campaigns

Every campaign, given a verdict
8 active
Repurchase Sequence
Automated · repeat buyers · LTV 2.45x avg
Scale$7,518
High-Value Upsell Trigger
Trigger-based · high-frequency · LTV 1.43x avg
Scale$4,382
Win-back: Lapsed 60D
Scheduled · dormant cohort · LTV 1.07x avg
Watch$3,300
Welcome Series
Always-on · new customers · LTV 0.42x avg
Reframe$1,297
Circular allocation signal

The top campaign carries a 2.45x LTV index, yet part of the customers it engages cycle straight into the dormancy queue, creating an estimated $10,112 in double-spend across initial engagement and reactivation on the same customer base. The friction sits between conversion and the customer's next independent action. The fix is a follow-through sequence timed inside the first window after engagement, before dormancy onset rather than after it is detected.

The model

How a directive
earns the budget.

Four mechanisms separate raw data from a directive: learned baselines, isolated causes, priced risk, verified outcomes.

01Baselines

Learned from your operation.

ONYX starts by learning how your operation actually behaves: seasonality, day-of-week rhythm, promo cycles, and the natural volatility of every channel and metric. The engine is measured against the context of that starting baseline.

Timing intelligence · Miami cohort238 transactions
Avg lead time
55d
Repurchase interval
32d
Danger threshold
117d
Intervene by
Day 25
0d25d32d117d
Learned from 238 observed transactions in this market, not an assumed benchmark. The intervene point sits at 80% of the median repurchase interval, positioning outreach ahead of the natural decision window. Customers past day 117 cross the 90th percentile of observed intervals and fall outside the recoverable range.
02Isolation

Cause, not correlation.

When a metric breaks its baseline, the engine splits the move across every dimension, tracking channel, campaign, audience, creative, geo, and cohort. The search narrows until the shift is pinned to the smallest segment that explains it, with each factor's contribution quantified.

Engagement-to-repurchase gapJourney by stage
01 · Landing
$544,058
02 · Browse
$323,943
03 · Detail
$226,156
04 · Checkout
$68,715
05 · Confirmed
$0
The problem

At step 2, 2,059 sessions are present and 41.7% exit before reaching step 3, with $323,943 of LTV upside at stake. Notably, the highest-value customer segment drops here at a conversion rate of 53.1%, the lowest of any segment at this step, pointing to a structural step issue rather than an audience problem.

03Risk signals

The cost of waiting.

Not every risk is a sudden anomaly. ONYX watches for slow-building exposure: segments drifting dormant, value concentrating in fragile cohorts, markets losing velocity. Every risk signal carries its LTV impact, a confidence level, and a time window, visible before the exposure compounds.

Risk signals2 active
HIGH · 60 days −3.1%LTV impact

The dormant repeat segment holds 28.3% of the customer base, and 112 customers within it represent a large pool of value that requires active reactivation to convert. The drop-off cliff shows the highest-LTV customers are among those churning, meaning passive approaches will not recover the most valuable accounts in this segment.

Impact
−3.1%
Confidence
53%
Combined exposure−4.9% LTV
04Verification

Verified against its own forecast.

After a directive is executed, ONYX weighs the actual result against the projection inside a defined verification window. The outcome is stamped onto the directive, logged to the model, and used to recalibrate the thresholds that produced it. Every decision makes the next one sharper.

The outcomeMeasured at 30 days
Success is measured by

Portfolio-wide 90-day LTV rising above $3,071, conversion rate holding above 1.0% at increased volume, and cost per order remaining below $120.

If not acted on within 7 days

$17,225 in monthly spend continues producing customers at $1,332 LTV while the highest-value dormant customers advance further past their repurchase windows.

LTV multiplier on reallocated spend4.47x
Monthly budget reallocated$17,225 (32.5%)
Source campaigns · avg LTV 90$1,332
Destination campaigns · avg LTV 90$5,950
Every outcome logged to the model. Thresholds recalibrated after each call, including the misses.
Start here

See it inside
your own data.

We'll connect your sources and walk you through what ONYX finds: the baselines it learns, the causes it isolates, the directives it builds.

Solutions

One model.
Built for every function of growth.

The best consumer brands run marketing as one connected operation. ONYX gives every part of yours the same intelligence, framed for the decisions each one owns.

For those setting the direction

Every question, already answered.

Owning the number means owning the explanation. ONYX briefs you before anyone asks: what moved, why, what's being done, and what it should return. Visibility into everything the operation is doing.

  • What's moving, why, and who's on it, at a glance
  • Every directive assigned, tracked, and closed
  • Priorities set once, carried through the whole operation
  • Every figure defensible down to the source query
Weekly Brief 7findings7actions3watches
ONYX InsightCritical

The Repurchase Sequence produces customers at $7,518 avg LTV 90, 2.45x the portfolio average, while receiving only 16% of the $53,000 monthly budget. 33.9% of its own 65-customer cohort has already gone dormant.

The findingsThe actionsThe watchlist

01Reallocate budget from the two lowest-LTV campaigns toward the Repurchase Sequence and the High-Value Upsell TriggerAllocation

Impact$17,225/month redirected from campaigns producing $1,332 avg LTV to campaigns producing $5,950: a 4.47x LTV multiplier on the same spend

TimeframeBudget reallocation within 7 days; dormant cohort analysis within 48 hours

The action

Reduce send volume on the two underperformers and redirect the freed $17,225 per month. In parallel, deploy a targeted reactivation campaign for the 22 dormant repeat buyers, carrying $4,509 avg LTV each, before they cross the irreversible dormancy threshold.

The justification

The destination campaigns convert at 1.2% and 1.5% with cost per order near $100 and produce $7,518 and $4,382 avg LTV. The source campaigns produce a combined $1,332 at more than double the cost per order. Every dollar moved buys 4.47x the customer lifetime value.

The outcome

Success at 30 days: portfolio-wide LTV rising above $3,071, conversion holding above 1.0% at increased volume, and cost per order below $120. If not acted on within 7 days, $17,225 in monthly spend keeps producing customers at $1,332 LTV.

LTV multiplier4.47x
Monthly budget reallocated$17,225 (32.5% of $53,000)
Source campaigns · avg LTV 90$1,332
Destination campaigns · avg LTV 90$5,950

02Scale outreach toward the High-Value Upsell Trigger and monitor cost per order as volume increasesRetention

Impact+$25,744 of spend realigned to the segment converting at 2.24x the portfolio average

For those allocating budget

Built to execute.

The hours between a break and its cause are the most expensive hours in performance. ONYX closes them before anything reaches you, allowing you to spend the day executing instead of searching.

  • Reallocation specified to the dollar, from source to destination
  • Saturation flagged before marginal ROAS crosses breakeven
  • Scaling decisions priced on marginal return, not blended averages
  • Allocation concentrated by LTV, measured from behavior at the segment level
Risk signals4 active
ACT NOW32.5%of spend misallocated

32.5% of monthly spend is producing customers 56.6% below average LTV. The two campaigns responsible hold $17,225 of budget while the portfolio's best performer remains underfunded at a 16% share. Reallocation plan generated and ready to review.

Impact
−4.9% LTV
Confidence
83%
HIGH · 60 days−1.8%LTV impact

Orders in the New York segment have fallen to zero in the most recent 7-day window, against 4 in the prior 30 days and 12 in the prior 90. Its 8.1% cancellation rate is the highest among tracked markets, and an estimated $18,000 to $22,000 in annualized revenue is at risk if the segment does not recover.

Impact
−1.8%
Confidence
58%
WATCH · 60 days$15.8Kdemand at risk

The seasonal bundle segment is accelerating across three regions while inventory gaps persist in two of them. If capacity does not expand within 60 days, an estimated $15,834 in unfulfilled demand redirects permanently to competitors during peak season.

WATCH · 14 days$68.7Krevenue exposed

Three market-segment combinations are declining simultaneously, with $68,681 in combined 90-day revenue at risk. A second consecutive zero-order week in any of them escalates to a pricing and demand diagnostic.

Combined cost of inaction−$8,599 /week
For those proving what drives revenue

Attribution that can explain itself.

Proving what drives revenue means standing behind the number. ONYX reads contribution independently across every touchpoint, grounds each figure in observed behavior, and carries its reasoning end to end, so every answer holds up to the question.

  • One governed model, defining every metric once
  • Ask ONYX absorbs the ad-hoc queue, grounded in your data
  • Full lineage from any directive back to the source query
  • Warehouse-native: works with Snowflake and BigQuery
Segment CAC analysisCAC vs. LTV 90
SegmentCACAvg LTV 90LTV:CACPayback
high_frequencyICP$816$8,42610.3xTXN 1 · ~6d
dormant_repeat$851$2,6193.1xTXN 2 · ~47d
active_repeat$782$2,2902.9xTXN 1 · ~34d
one_and_done$696$1,1851.7xTXN 3+ · unrecovered
ONYX directive

The 34-day payback on a $782 CAC reflects customers who arrive with a defined use case and convert without extended evaluation. At 2.9x LTV:CAC, this segment recovers its acquisition cost at first order, and the campaign sourcing it should be scaled: a 20% growth in this segment adds $481,590 in 90-day LTV to the portfolio, while a 20% contraction removes $385,380. Increase allocation to its primary acquisition source.

Trajectory

Dividing $2,290 LTV by $1,092 average order value yields roughly 2.1 orders per customer within 90 days, with the first purchase landing at day 34 and the remainder following inside the window.

For those who own the customer lifecycle

Win them back while they're winnable.

The margin lives in the customers you already have. ONYX reads every cohort in real time, catches repeat behavior across segments the moment it bends, and delivers the win-back with the revenue already sized.

  • Cohort health tracked daily: repeat rate, LTV, and purchase windows
  • Replenishment and flow timing signals before revenue slips
  • Dormant high-LTV segments surfaced and ranked by projected recovery
  • Every win-back verified: projected lift vs. measured lift
Cohort behaviorPurchase progression by segment
High Frequency$8,426
No drop-off detected. Retention is near-perfect across all purchase stages.
Dormant Repeat$2,619
Drops 15% at purchase 3. Engagement fails to survive reactivation; the intervention window sits right after purchase 2.
Active Repeat$2,290
Drops 23% at purchase 4. Value truncates late in the cycle; replenishment timing runs a week behind the buying rhythm.
One And Done$1,185
Drops 91% at purchase 2. A single-point failure; the second-purchase decision is won or lost inside 30 days.
Directive

The progression for this segment collapses entirely after the first transaction: 91% of customers never return for a second purchase, and zero reach a third. This is not gradual attrition; it is a single-point failure at purchase 2, which means the retention window is narrow and the intervention must arrive before the second-purchase decision is made, not after it lapses. Deploy a post-first-purchase re-engagement sequence timed 7 to 14 days after the initial transaction, targeted at all 44 customers in this cohort, with the explicit goal of a second confirmed purchase before day 30.

Segment avg LTV 90$1,185
Portfolio avg LTV 90$3,071
Reactivation targets in window44 customers
Recoverable revenue · 83.5% weighting$517,243
ICP connection ▸ the high_frequency purchase fingerprint is available as the targeting baseline for this sequence
One model, every function

See how ONYX works
across every function of your operation.

Integrations

Connected natively.

Native connections to the platforms where B2C marketing actually happens. We handle the connection, backfill your history, and unify everything into one governed model.

Zero-lift setup

Forward-deployed engineering.

Every connection is handled by ONYX engineers, working inside your setup from day one. The engineering conforms to your current architecture, and the operation runs uninterrupted as the model is deployed.

No engineering ticket Read-only by default Nothing replaced Your team keeps its tools
Day 0

Discovery

Discovery begins by learning your business: the revenue, the customers, and the goals the model will measure against.

Day 1

Connection

Every source is brought into the model, natively connected and backfilled with your full history.

Day 7

Calibration

The model calibrates to your operation: baselines established, learned from your history, validated against your trajectory.

Day 14

Delivery

The model goes live: the first directives land, sized in dollars, and the engine starts compounding.

The catalog

Connections.

Advertising platforms

Spend, delivery, creative, and audience data, read continuously from every channel you buy on.

M
Meta Ads
G
Google Ads
T
TikTok Ads
Y
YouTube
A
Amazon Ads
P
Pinterest
S
Snapchat
A
AppLovin
+
+ more

Commerce and revenue

Orders, subscriptions, and payments: the ground truth your ad platforms never see.

S
Shopify
S
Stripe
R
Recharge
A
Amazon Seller
W
WooCommerce
B
BigCommerce
+
+ more

Lifecycle and CRM

Email, SMS, and customer records, joined to what acquired them and what they became.

K
Klaviyo
A
Attentive
H
HubSpot
S
Salesforce
B
Braze
I
Iterable
+
+ more

Analytics and data infrastructure

Warehouse, product, and event data, read directly from the source.

G
GA4
S
Snowflake
B
BigQuery
S
Segment
A
Amplitude
+
40+ more
Into the model

Every connection, unified.

The inputs
Meta Ads Google Ads TikTok Ads Shopify Klaviyo GA4 Snowflake 40+ more

One governed model

Behavior resolved · Segments modeled · Compounding intelligence

The outputs
Directives, sized Signals, explained Actions, assigned Projections, verified
Get started

Connect your tools
and see what ONYX finds.

Security & trust

How ONYX handles your data.

ONYX is built for companies where marketing data is revenue data. Security, governance, and auditability are architecture, not an enterprise add-on.

01

Overview

Every customer runs on the same security posture: encryption everywhere, least-privilege access, isolated workspaces, and full auditability. None of it is an upgrade tier, and none of it is optional.

This page summarizes how ONYX handles your data, who can access it, how the platform stays available, and what we provide your security and legal teams during review.

Last reviewed · July 2026
02

How your data is handled

Your data stays yours. ONYX reads what it needs, protects it end to end, and removes it when you say so.

Encrypted everywhere

All data is encrypted in transit (TLS 1.2+) and at rest (AES-256). Keys are rotated and managed under strict access controls.

Read-only by default

Platform connections request least-privilege, read-only scopes. ONYX analyzes your operation; it never has write access to your ad accounts unless you grant it.

You own your data

Your data is never sold, never shared across customers, and never used to train models for anyone else. Workspaces are logically isolated per customer.

Deletion on request

Offboarding includes full deletion of your data within a defined window, with written confirmation when it completes.

Data minimization

ONYX ingests the marketing and commerce data required for the model and nothing else. PII handling follows documented minimization policies.

Residency options

Regional data-residency options are available for teams with jurisdictional requirements, scoped during onboarding.

03

Access and governance

Enterprise identity, granular permissions, and a full record of who did what.

SSO / SAML

Single sign-on via SAML 2.0 and OIDC with your identity provider: Okta, Entra ID, Google Workspace.

SCIM provisioning

Automated user provisioning and deprovisioning, so access ends the moment someone leaves the team.

Role-based access

Granular roles by workspace, brand, and module. Analysts, operators, and executives each see exactly what they should.

Audit logs

Every login, permission change, and data access is logged and exportable for your security team.

04

Reliability

A decision engine is only useful if it's there when the decision is due.

Uptime commitment

Contractual uptime commitments with public status reporting and defined service credits.

24/7 monitoring

Continuous infrastructure and pipeline monitoring with automated alerting and on-call response.

Backups and DR

Automated backups with tested disaster-recovery procedures and defined recovery-time objectives.

Incident response

Documented incident-response plan with defined notification timelines to affected customers.

05

Compliance and procurement

Everything your security and legal teams will ask for, ready before they ask.

Certifications

Independent security certifications and audit reports, shared with customers under NDA as they are completed.

DPA and GDPR/CCPA

Standard DPA available for signature, with GDPR and CCPA-aligned data-subject processes documented.

Security reviews

Completed questionnaires (SIG, CAIQ) and architecture documentation available under NDA to accelerate vendor review.

Penetration testing

Regular third-party penetration tests, with executive summaries available to customers under NDA.

Start here

See ONYX inside
your own data.

FAQ

What leadership teams ask.

Jump to
The short version

ONYX is a layer, not a replacement.

It connects to the tools you already use, unifies the data into one governed model, and turns it into directives your team executes.

Still have a question?

Talk with our team and see what ONYX looks like inside your business.

Schedule a demo
The product

Dashboards visualize what happened; ONYX decides what to do about it. Every finding ships as a specific action with the reasoning behind it and a projected outcome, and ONYX verifies the impact of that action once it concludes.

No. ONYX is a layer, not a replacement. It connects to your existing ad platforms, analytics, storefront, and CRM, and unifies them into one decision model. Most teams keep everything they have and simply stop doing the manual synthesis work.

Every directive is backed by the source data behind it: the numbers it was built from, the reasoning it runs on, and the projected outcome it carries. And nothing is take-it-or-leave-it — Ask ONYX lets your team question any directive directly, push into the data underneath it, and get answers with the data attached. After execution, the actual impact is measured against the projection, so the track record remains visible.

B2C marketing organizations with a large existing customer base, meaningful multi-channel spend, or an operation where retention and repeat purchase drive the margin. If the growth questions have outgrown the tools answering them, ONYX is built for that team.

ONYX vs. the field

Attribution answers one question: which channel deserves credit. ONYX takes that as a starting point and builds the full picture around it: how those customers behave, what they're worth long-term, and which move that justifies. The output is a directive, sized and delivered to execute.

MMM produces strategic, quarterly-scale reallocation guidance from aggregate models. ONYX operates at the speed your team actually works: daily signals, segment-level causes, weekly directives, and verified outcomes. Many teams run both — ONYX is the layer the team executes from between planning cycles.

A copilot can only read what your dashboards already show, so it inherits their limits: no baselines, no causes, no accountability for being right. ONYX runs its own model underneath — learned baselines, isolated causes, dollar-sized projections, and verified outcomes. Ask ONYX is one interface into that engine, not the engine itself.

Getting started

Forward-deployed engineering means the engineers who build the product also deploy it — working inside your environment, on your systems, until it runs. Instead of a self-serve setup where your team does the integration work, our team comes to you: discovery, connection, calibration, and delivery, all handled by the people who built the model.

The first two weeks. ONYX engineers handle the full deployment: discovery on day zero, every source connected and backfilled on day one, the model calibrated against your history by day seven, and the first directives live by day fourteen.

It starts with the business, not the data. Our team maps your operation first: the revenue, the customers, and the goals the model will measure against. Then every source is connected, the model is validated against your own numbers, and thresholds are tuned with your team before anything ships.

Encrypted in transit and at rest, isolated to your workspace, never shared across customers, and never used to train models for anyone else. Connections are read-only by default, and full deletion on offboarding is standard. The Security page covers the complete picture.

Pricing runs on three, six, and twelve-month engagements, scoped to each operation: the sources connected, the size of the model, and the depth of the deployment. Every engagement is priced to the business it serves.

Talk to our team
Still have a question?

Ask our team
directly.

Schedule a demo

Built on your operation.

Everything your data has been trying to tell you, turned into action.

What to expect

  • First, your business.Understanding your operation in depth: where revenue comes from, your customer base, your segments, your channel mix.
  • Then, ONYX through your lens.A walkthrough of how ONYX works, framed around your operation and how it would look for your team.
  • Where ONYX would deploy first.You leave knowing the first thing ONYX would target in your business: the segments, the influence on revenue, and why it comes first.

Still have a question?

Talk with our team and see what ONYX looks like inside your business.

Talk to us
Pick a time that works.