Product · How Bicycle works

What happens when a KPI moves?

Bicycle detects the move, explains the business and technical causes, and recommends a safe next step. It plugs into the data stack you already use.

No credit card required

01 · Detect
Detected
Approval rate · cross-border
−6.7ppdetected early
02 · Explain
Explained
Likely cause · confidence high
3DS frictionprimary
Issuer downgradescontributing
Routing changecontributing
Latency · Fraud rules · Holiday window ruled out
03 · Act
Scoped
Recommended next step
Route low-risk traffic → backup gateway
24h · scoped · reversible
scopeTTLrollbackaudit
04 · Learn
Banked
Outcome banked
Approval rate
+5.9pp
Recovery
37 min
Cause accepted. Feeds the next Detect.
01

Movement is the signal

A KPI at rest carries no new information. Attention should start the moment it shifts, and only then.

02

A signal is not a decision

Seeing the dip is the easy part. Knowing why it moved and what to do next is the hard 95%.

03

The gap is the job

Carry a signal all the way to a confident decision. That is what analytics should be about.

The idea behind Bicycle

Data isn't the signal. The change is.

A number sitting still tells you nothing new. And a number that just moved is still only a signal, not a decision. The real work is the distance between "something changed" and "here is what to do about it." Closing that gap is what analytics should be about. It is what Bicycle is about.

So we built the loop that closes it

Detect → Explain → Act → Learn.

Every time a revenue-critical KPI moves, Bicycle runs one loop end to end. Not an agent for the sake of being an agent. An agent loop with a single job: turn a signal into a decision, and get sharper at it every cycle.

DDetect
EExplain
AAct
LLearn
Explain · the widest and hardest step

Detecting movement is table stakes. Explaining it is where analytics actually lives.

A KPI rarely moves for one tidy reason. The cause can sit in the business or buried deep in the stack, so Explain has to test both at once, then prove which one it was. This is the step where every other approach quietly hands the work back to you.

The Steering

Business performance

Governed by KPIs like auth rate, approval rate, payment completion.

Influenced by business factors like pricing, routing rules, fraud policy settings.

The Power

Technical performance

Governed by metrics like gateway latency, processor uptime, error rates.

Influenced by technical factors like routing changes, feature flags, network health.

Bicycle tests business and technical drivers in parallel, powered by the multi factor cause analysis. Deterministic statistics, not a language model's guess. That engine is the hardest part to build, and the reason every answer arrives with evidence instead of a confident-sounding sentence.

Compare Bicycle to your current tools.

Every team already has dashboards, data pipelines, or an AI chatbot. Bicycle works on top of all of them. See exactly how it differs.

How it runs

You just saw the four steps.
Here is how they actually run.

Two things you set up once, on the stack you already have, and then the same Detect → Explain → Act → Learn cycle runs every time a revenue-critical KPI moves.

Setup · one time, before it starts
1Activate
On your stack

Connect existing systems. No migration, no rebuild. Bicycle reads what you already have and proposes a reviewable starting point.

Typical signal connectors
Warehouse Event streams dbt models BI assets Operational DBs Logs + tickets Docs + runbooks
Output: events, dimensions, KPIs, and a first analytics agent ready for D&A review.
2Contextualize
In your business language

Apply the vertical pack and the context that makes outputs read like a peer wrote them, not a system. Override anything.

Context Bicycle brings + you add
Vertical pack Business glossary KPI definitions Driver templates Action templates Role defaults
Output: KPIs, slices, patterns, drivers, and action templates that speak your team's language.

Once setup is done, the agent runs every day.

Runtime · runs automatically, every day
D
Detect

Watch every KPI and slice. Surface meaningful movement before anyone has to ask.

Powered by Always on KPI Intelligence
E
Explain

Test business, technical, partner, and external drivers in parallel. Return the likely cause with evidence.

Powered by Multi factor cause Analysis + Defensible answers and data validation
A
Act

Recommend the next step. Route the owner. Preview, scope, approve, log. Every action inside governance.

Powered by Governed self service and actions with guardrails
L
Learn

Capture accepted causes, outcomes, and reusable context. The next explanation is faster and more relevant.

Powered by Decision trace, outcome tracking, reusable context
↻ Every cycle improves the next
Vertical
View as

Same setup. Same sequence. The block sits on persona pages and vertical pages with the right lens preselected.

One capability system · many ways to deliver value

The six capabilities that make it work.

Bicycle combines fast activation, vertical context, continuous KPI intelligence, cause analysis, defensible answers, and governed action paths into one operating system.

Capabilities work together continuously to drive better outcomes every time.

Explore the playbook
Two vantage points · one system

One operating system.
D&A keeps it trusted; business teams act on it.

Business teams can self-serve because D&A controls the foundation, the intelligence layer, and the action boundaries. Neither side learns the other's job.

Data & Analytics · The Control Layer
What D&A keeps in control
Build the trusted foundation

Review the events, dimensions, KPIs, joins, and context Bicycle proposes. Approve the model before the business sees it.

E.G. Approve the definition of approval rate, map issuer and BIN dimensions, confirm which slices matter.
Tune the intelligence layer

Tune pattern detection, driver trees, cause connectors, thresholds, and recommended actions as the system learns.

E.G. Add issuer downgrades and 3DS challenge rate as drivers for payment approval drops.
Govern self-service and safe actions

Set roles, access, approvals, action permissions, audit trails, and rollback paths.

E.G. Let regional managers request a routing change, but require payments-ops approval before execution.
Explore the D&A playbook
Business Teams · The Solution
Know what changed. Understand why. Act on it.
Know

See which KPI moved, where it is concentrated, and how big the impact is.

E.G. Search conversion dropped 18% for Delhi grocery searches.
Understand

See the likely cause, supporting evidence, and what Bicycle checked but ruled out.

E.G. Top contributors: inventory availability, price gap, and search relevance.
Act

Take or request the next step inside approved guardrails.

E.G. Notify category and replenishment teams with affected SKUs, stores, and expected impact.
See how business teams use it
Architecture & stack fit

One system, on top of the stack you already run.

Bicycle is the agentic layer that turns business signals into alerts, stories, dashboards, chat answers, triage, and governed actions. It reads your stack; it does not replace it.

Activate on your stack.

Connect existing systems. No migration. Bicycle reads what you already have and proposes a reviewable starting point.

Contextualize in your language.

Apply the vertical pack and business context so outputs read like a peer wrote them, not a system.

Bicycle the agentic layer
Surfaceswhat business teams touch
AlertsTriageStoriesDashboardsChatActions
Governancethe self-serve safety rail
access · approved definitions · evidence · audit · rollback
AgentsAI · your business analyst

Frames the question, orchestrates the investigation, writes the story, recommends the next step. It reasons.

AlertsTriageStoryChatOnboarding
IntelligenceAutoML · your data analyst

Detects movement, ranks drivers, computes confidence, forecasts. The numbers are calculated, not generated.

Pattern EngineCause EngineDriver TreesForecastingImpact Ranking
Business modelthe semantic layer
OntologyKPIsDimensionsJourneysCohortsPoliciesPlaybooks
Connectorsthe bridge, in and out
Signal Cause Action Knowledge
reads ↑ · acts ↓ · Bicycle sits on top
Your stack · system of record
SnowflakeBigQueryDatabricksRedshiftdbtKafkaDatadogSplunkGitHubSlackJiraNotion+ more

And the rest of what you run: warehouses · streams · observability · payments · tickets · docs · BI · operational systems.

Bicycle runs on top of the systems you already use, and your data can stay where it already lives.

Pick your starting point

Bicycle meets you where your KPI moves.

For the team that owns the number

You own the KPI.

Bicycle detects the move, finds the cause, and recommends the next step on the revenue KPIs critical to your business.

Category managersRevOpsEcommerce opsPayments opsSupplier ops
See the business view
Alerts inbox
Revenue Risk · top focus set
$182Kat risk · 8 cities
Search conversion dropacting
Cart abandonmentwatch
For the team that answers for the data

You govern the answers.

Bicycle handles the recurring investigations so your team works on what matters. Defensible chains, audit logs, governed self-serve.

Heads of dataAnalytics leadersBI teamsData product owners
See the Data & Analytics view
Data stories · catalog
Recurring investigations
Shopper weekly · 18 runs
Subscribed: marketing, ops
Conversion health · daily
Approved · Data & Analytics signed off
Pattern Analysis · weekly
3 sites · 1 owner
For the team that decides what to do

You shorten the time to a solution.

Bicycle gives leadership signal-to-decision in one read: the move, the cause, what's being done, what changed since last week.

CEOsCOOsGMsBU leadersStrategy leaders
Book a demo
Story · APAC
APAC Revenue · 18% GMV decline · March
−18%MoM · mid-market
Leading hypothesis: pricing change in week 2. Causes ranked. Action recommended.
Detecting KPI moves across
Self-serve on-ramp

Start with Vibe Analytics.

Try the self-service entry point first. Ask a business question, connect a trusted data source, and see how Bicycle turns prompt plus data into an analytics agent.

Prompt + data
"Why did approval rate drop for cross-border?"
Snowflake · connected
Analytics agent
Builds the model, detects movement, tests causes.
Outputs
Answer Story Alert

No credit card required

Walkthrough · 30 minutesSee it on a use case that mirrors yours. A use case in your vertical, pre-loaded. Bring the question. We bring the example.

Common questions

Before you ask.

BI shows what happened. Bicycle explains why and what to do next. It sits on top of the BI you already trust, runs the investigation across business and technical signals, and recommends the next step.

Most AI analytics starts when someone asks. Bicycle starts when the business changes. Chat is one surface inside Bicycle, not the product. The Analytics Agent watches the KPI before you knew to ask.

No. Bicycle works on top of the stack you already use. Data & Analytics teams stay in control of definitions, evidence, publishing, and access. Business users self-serve through that governed layer, not around it.

Warehouses, event streams, BI assets, dbt models, payment systems, supplier systems, catalog and inventory, logs, incidents, tickets, runbooks, Slack, and email. No migration required. Data can stay where it already lives.

First useful answer in days, not months. The two-week trial delivers one working agent for one KPI by day 14: connect by day 2, first proposed agent by day 4, review and publish by day 10, measure value by day 12.

No. Your raw data is never sent to the AI model and never leaves Bicycle's secure runtime. AI generates structured artifacts (mappings, rule-sets, KPI logic); those are evaluated inside the runtime against your data, and Bicycle does not train any model on your data. OpenAI runs with Zero Data Retention. SOC 2 Type II and GDPR.

Test ride Bicycle on a revenue-critical use case.

Bring one KPI that matters. We'll show how Bicycle detects movement, explains the cause, and recommends the next step.

No credit card required

No credit card required Start for free