For AI labs Talk to us

The AI workspace for business teams.

Engineers have Cursor. Designers have Figma. Business teams still piece their work together across dashboards, spreadsheets, documents, and Slack.

Veloccy gives your team one place where people and AI investigate questions, develop changes, and measure results, on your stack and by your rules.

Become a design partner See one workflow, start to finish

One workspace, from question to outcome.

Explore the findings, reviewed change, and results from one commerce growth investigation.

app.veloccy.com/simcorp/arcs/2026-08-18-women-category-list

From observation to shipped test.

01

It notices on its own.

Veloccy keeps track of how your customers move through the app and where each group of them drops off. When one group is doing worse than the rest, it checks the numbers first, then posts what it found in Slack, with the chart.

02

Your team asks the follow-ups. It answers with data.

  • Is it the layout, or the customers?It checks whether the same pattern holds for everyone, and on other devices.
  • How does it know who is who?It uses the profile where it is filled in, and shows how well browsing history predicts the rest.

Every number links to the query that produced it, and it says what it could not check.

03

It sizes the opportunity and drafts the test.

  • 01It works out what the gap is worth in orders and revenue.
  • 02It drafts a few versions of the fix and says which one it expects to win.
  • 03It works out how many people and how many days the test needs.
  • 04It writes the guardrails before asking anyone to approve.
04

It works through your tools, and asks the people who own them.

Changes go through the internal tooling your team already has. Before you approve, Veloccy asks the engineer who owns the tool and the person who owns the category to review, and their conditions are recorded on the change. Write access is limited to this one change.

05

It watches the test and tells you when the result is clear.

Each day it compares every version with the control and checks the guardrails. When there is enough data, it reports the winner, retires the others, and suggests where to look next. If nothing had worked, it would say so and put things back.

That was one question on a demo store. Read-only access is enough for Veloccy to start on yours.

Try Veloccy read-only Become a design partner
Thread# growth
V
Veloccy APP7:04 AM

Women are 54% of daily users but 38% of orders. 62% of their product views start from the Shop tab’s category list, which puts the five women’s categories at positions 8 to 12, below the fold on most phones. 44% of women who open the list leave without opening a category; for men it is 21%. I have opened an arc. @Priya Nair, you own growth, so I will bring it to you.

where-product-views-start.pngPNG
Share of product views by how the customer got there
womenmen

6 queries, every one linked. Open the arc

Priya Nair 9:31 AM

how sure are we it’s the list order and not that women just browse differently? and how do you even know who’s a woman

V
Veloccy APP9:36 AM

Two checks. The chance a category is opened falls with position the same way for men and women; women’s categories are simply lower down. And on tablets, where all fourteen fit on one screen, 27% of women leave without opening a category against 22% of men. On who is a woman: 71% have it on the profile, and for the rest browsing history predicts it 91% of the time.

category-list-open-rate-by-position.pngPNG
Share of list openers who open each position
womenmen
👀 3
V
Veloccy APP11:52 AM

If women reached a category as often as men, that is about 1,200 more orders a week, roughly $300k of GMV a month. I have drafted three versions of the list for women only: women’s categories first, each customer’s own order, and a “for you” row above the standard list. I expect the first to win. 18,000 women open the list a day, so four arms can detect a 5% lift in nine days. Guardrails: men’s list unchanged, Sneakers in the top five for everyone, AOV within 2%. The order lives in the storefront config tool, so @Tomás Herrera @Hannah Lindqvist could you review?

TH
Tomás Herrera Aug 19th at 10:14 AM

per-audience ordering works since 4.10. one thing: the nav config is cached for 6 hours, so the switch takes up to 6h to show on every phone. approved from the app side ✅

HL
Hannah Lindqvist 11:02 AM

fine by me as long as Sneakers stays top 5 for both lists, it’s shared and our best margin. dresses first matches what we’re pushing in sept. approved ✅

Priya Nair 12:31 PM

approved, run all three

V
Veloccy APP12:40 PM

Applied, with Tomás’s and Hannah’s conditions on the change page. The cache means the first full day is tomorrow, so the test counts from then. I check all four arms every morning, and if men’s numbers or AOV move more than 2% I put the list back and tell you here.

V
Veloccy APPSep 4th at 7:03 AM

The results are in. Women’s categories first won: 31% more women open a category from the list, and women’s orders per active user are up 9.2%. The personal order tied it for repeat buyers but lost with new users. The “for you” row was ignored. Men’s numbers and AOV held. Parents show the same pattern with Kids, so I have opened that question.

list-to-category-by-arm.pngPNG
List → category among women, by arm and day
controlA women firstB personalC for-you row
🎉 7🙌 2
Priya Nair 8:12 AM

yes to both. this is great 🙏

Reply…

Your business work in one app.
Your team and AI building on the same context.

  1. Today

    Your team investigates in a notebook, proposes changes in a doc, coordinates in Slack, and tracks results in a dashboard—re-explaining the findings and next steps at every handoff. Even with AI in each tool, connecting the work is still your job.

  2. In Veloccy

    Start an arc around a question or opportunity. Together, you and your team work with AI to investigate questions, develop changes and build dashboards—all inside the same arc.

  3. When the numbers move

    Keep the conversation going. With the live dashboard beside the investigation and the change, it’s easier to answer “did that help?” and “what should we try next?” The results, evidence and reasoning are all in one place.

Findings. Investigate with AI, then dig into the evidence together. Every answer keeps its query, results and caveats, so your team can challenge an assumption or take the analysis further.
Changes. Turn the conversation into a concrete change. See exactly what AI proposes, refine it with your team, and review the expected impact and rollback before approving.
Dashboards. See whether the work paid off, right inside the arc. Follow live results alongside the change you made, ask AI about a movement, and build on the investigation without starting over.
Home. Your starting point across every arc. See what needs your review, what your team is working on and what happened this week—then jump back into the work.
Arrow keys move between screens

Every question, with the evidence that answered it.

A finding is one question, the query that answered it, the table it returned, what Veloccy read from it, and what it could not check. The queries run against your warehouse, and anyone on the team can run them again.

Nothing has to be copied out of a notebook or pasted into a doc. When someone asks how a number was reached, the answer is already there, next to the number.

What changed, who agreed to it, and how to undo it.

A change shows the fields it touches in your tool, before and after. Reviews from the people who own that tool and that part of the business sit above the diff, with their conditions.

The predicted effect, the guardrails and the rollback are written before anyone approves, so a change never depends on someone remembering what was agreed.

Dashboards built from the findings.

Every dashboard uses the same queries as the investigation, so the chart your team checks each morning means the same thing the findings meant.

Arm tiles and guardrails appear when a change is applied, and they stay after the decision, so the result is easy to find months later.

One place to see what needs you and what is moving.

Home shows the decisions waiting on you, the questions Veloccy suggests, the arcs in flight, and how the week went. An approval here and an approval in Slack are the same approval.

Your team can see the whole picture without opening five tools or asking around, and anyone new can read how a decision was reached.

You don’t have to ask. It’s already looking.

Most analytics waits for you to ask a question. Veloccy walks your data all day and all night: customers, cities, pages, coupons, searches, delivery promises. When something could be doing better, it tells you what it’s worth and who should decide, and you decide whether to go ahead.

FIND 0417MON 07:04
Sold out, front and centre.

The men’s page opens on three sold-out products, and new visitors leave at twice the usual rate. Is the stockout what sends them away?

Estimated opportunity$86k/month
FIND 0424WED 23:40
Free shipping, seven dollars out of reach.

Free shipping starts at $75 and the average basket stops at $68. Do carts stall just below the threshold?

Estimated opportunity$36k/month
FIND 0429FRI 06:20
Back in stock. Nobody told the 4,100.

4,100 people asked to be told when the court sneaker came back. It came back, and nobody was told. How many would still buy?

Estimated opportunity$28k/month
FIND 0433SUN 05:31
They asked for it. Search said no.

“Linen set” and 41 other searches return nothing, and every one maps to a product you sell. How many of those shoppers leave?

Estimated opportunity$19k/month
FIND 0438TUE 22:15
Promising three days, delivering two.

Three Houston zip codes are promised three days and get two. Does the slower promise cost orders?

Estimated opportunity$14k/month
FIND 0421TUE 03:12
A discount reaching people who were already buying.

SAVE10 mostly reaches recent buyers. Is the discount generating additional orders, or paying for ones that would have happened anyway?

Estimated opportunity$41k/month
FIND 0426THU 04:55
The fast way to pay is below the fold.

On Android, Google Pay and PayPal sit below the fold at checkout. Are Android buyers dropping out where iOS buyers don’t?

Estimated opportunity$31k/month
FIND 0431SAT 01:08
No size guide, twice the returns.

Sixty items have no size guide, and they come back twice as often. Is it the missing guide, or the items themselves?

Estimated opportunity$23k/month
FIND 0435MON 02:47
The reminder arrives a day late.

The abandoned-cart email goes out after 24 hours, and most people who come back do so within two. Would an earlier email reach more of them?

Estimated opportunity$17k/month
FIND 0440WED 03:59
A push at 2 AM.

Pacific-time customers get the push at 2 AM and open it a quarter as often. Would a local send time change that?

Estimated opportunity$9k/month

It only needs to read your data to start finding things like these.

Try Veloccy read-only

No migration. Keep your dashboards. We connect to what you already run.

We connect to your warehouse and the tools behind one workflow, and map the definitions and approvals with your team before anything runs. Nothing moves, nothing gets re-tagged. Where a tool has no connector, your internal tooling for instance, Veloccy builds its own.

SnowflakeSnowflakeAWSAWSSAPSAPPostgreSQLPostgreSQLDremioDremioPower BIPower BIMicroStrategyMicroStrategyDomoDomoHexHexIBM CognosIBM CognosDataikuDataikuFivetranFivetranTalendTalendmParticlemParticleRudderStackRudderStackKafkaKafkaMixpanelMixpanelHeapHeapPendoPendoQuantum MetricQuantum MetricStatsigStatsigDynamic YieldDynamic YieldKameleoonKameleoonBrazeBrazeKlaviyoKlaviyoEmarsysEmarsysMoEngageMoEngageShopifyShopifyVTEXVTEXStripeStripeGorgiasGorgiasGoogle AdsGoogle AdsCriteoCriteoAdjustAdjustSnapchatSnapchatMicrosoft TeamsMicrosoft TeamsAsanaAsanaMicrosoft ExcelMicrosoft ExcelGitLabGitLabWorkdayWorkdayZapierZapierAmazon RedshiftAmazon RedshiftConfluenceConfluence
DatabricksDatabricksMicrosoft AzureMicrosoft AzureTeradataTeradataMySQLMySQLStarburstStarburstLookerLookerThoughtSpotThoughtSpotModeModeSisenseSisenseSASSASKNIMEKNIMEAirflowAirflowMatillionMatillionTealiumTealiumHightouchHightouchSparkSparkAdobe AnalyticsAdobe AnalyticsContentsquareContentsquareHotjarHotjarOptimizelyOptimizelySplitSplitVWOVWOSalesforceSalesforceIterableIterableAttentiveAttentiveBloomreachBloomreachCleverTapCleverTapcommercetoolscommercetoolsContentfulContentfulAdyenAdyenYotpoYotpoMetaMetaThe Trade DeskThe Trade DeskBranchBranchLinkedInLinkedInJiraJiramonday.commonday.comAirtableAirtableDatadogDatadogNetSuiteNetSuiteRetoolRetoolMicrosoft FabricMicrosoft Fabric
Google CloudGoogle CloudOracleOracleClickHouseClickHouseMongoDBMongoDBTableauTableauQlikQlikSigmaSigmaMetabaseMetabaseSupersetSupersetAlteryxAlteryxdbtdbtInformaticaInformaticaSegmentSegmentSnowplowSnowplowCensusCensusAmplitudeAmplitudeGoogle AnalyticsGoogle AnalyticsFullStoryFullStoryPostHogPostHogLaunchDarklyLaunchDarklyEppoEppoAB TastyAB TastyHubSpotHubSpotMarketoMarketoCustomer.ioCustomer.ioInsiderInsiderSprinklrSprinklrBigCommerceBigCommerceAlgoliaAlgoliaZendeskZendeskBazaarvoiceBazaarvoiceTikTokTikTokAppsFlyerAppsFlyerPinterestPinterestSlackSlackNotionNotionGoogle SheetsGoogle SheetsGitHubGitHubServiceNowServiceNowOktaOktaBigQueryBigQuerySQL ServerSQL Server

And the tools your own engineers built.

FIG 0.1 Keep everything exactly where it is

Your warehouse, your store, your messaging, your dashboards. Nothing migrates, nothing gets re-tagged, nothing is rebuilt. It reads what’s there.

FIG 0.2 Read-only until you say otherwise

Information flows one way, into Veloccy. It can’t change anything until the person who owns a tool turns write access on, and every change still needs a yes.

FIG 0.3 Internal tooling? It builds the adapter.

For the tools your engineers built and nobody wrote a connector for, Veloccy writes its own: it works out the API and forms, drives them with code, and tests that code against a copy first.

Company brain

A shared understanding of your business that grows as your team works.

What “active user” means here, and which team counts it differently. Who owns the Shop tab. Which data problems keep setting off false alarms. Which tables to trust. What the last test taught you.

All of it lives in plain notes your team and the AI can read and correct. Leave, and the notes leave with you.

# growth-pod·Sundays, 9:40 PM

Your Monday review, posted Sunday night.

Every number against last week, each move labelled real, expected, noise or data problem. The meeting starts at “what do we do”, not “is this even right”.

# growth-pod12 members
Sunday, 6 September
V
Veloccy APP9:40 PM

Week of 1 Sep. 14 numbers checked against last week. Two moved for real, one was a data problem, the rest are noise or expected.

Orders +2.1% noise
Orders per active user, men −6.1% real Men’s page rows
Average order value −3.4% expected Sale ended 29 Aug
Cancellations, Texas +9.0% data problem Duplicate rows, fixed
Repeat rate, 30 days +0.8% noise
Searches with no results +4.2% real Missing synonyms

Full review with the work behind each line: veloccy/review/2026-w36

🙏 4👀 2
Riya Menon 9:52 PM

dropping this straight into the Monday deck, thanks

You’ve built a way of working. Veloccy works with it.

Your team has tools it trusts, people who own decisions, and checks before anything goes live. Veloccy works with those tools and follows your rules for testing, review and approval. We map your setup with you before anything runs, so it fits how your team operates from the start.

  1. 01More questions than your analysts can get to?

    Work through them with AI, using the data and definitions your team trusts.

  2. 02Every change involves three teams and five tools?

    Bring the investigation, reviews and results together in one arc.

  3. 03Different teams keep relearning the same things?

    Give your team and AI shared context that grows with the work.

  4. 04You already have a process that works?

    Keep your tools, testing standards and approvals. Veloccy works with them.

See what Veloccy finds in your business.

Try Veloccy in read-only mode with your team’s data. Get insights and concrete suggestions, with the evidence behind them. Act on them with Veloccy or take them back to your team. If you don’t continue, the findings are still yours.

Try Veloccy read-only No changes to your systems. No strings attached.
+0.0%orders per active user
demo store, 30 days
CHANGE, KEPTDATA PROBLEMEXPECTED MOVEIDEA, NOT TESTED

Eleven findings in thirty days. Four changes kept, three data problems caught, four that needed nothing more than a note. That’s one month of watching every day.

EVERY DAY, 07:00Yesterday’s numbers against their normal range.
ALL THE TIMELooking for what could be doing better, priced.
ANY TIMETag it in Slack. @Veloccy why is Texas down?

Am I still in control? Yes, completely.

01Nothing leaves your account.

It runs inside your cloud. The models run through your own Bedrock or Vertex. We keep notes about your data, never the data.

02It can’t change anything until a human says so.

It starts read-only everywhere. Each tool gets write access only when its owner turns it on, and every change still needs approval.

03Everything it knows is a file you can read.

Notes, findings, changes and decisions are plain files in a git repository. Your team can edit them. Leaving means taking a copy.

Design partners

Help shape Veloccy around one real growth workflow.

We’re looking for commerce teams to work with us on investigating metric changes and turning findings into reviewed interventions. We start read-only, with a named workflow and your team’s definitions.

THE FIRST CONVERSATIONBring one growth question your team keeps coming back to.

We’ll show you the workflow, discuss how your team handles it today, and identify a useful place to start. Access and approvals come afterwards, once we both think it fits.

THEN, TO STARTOne workflow and the person who owns it.

Read-only access to your warehouse and the tools behind that workflow. Someone on your team who can say yes or no to a change. Your definitions: what counts as an active user, which numbers matter, who owns which page.

WHAT YOU GETVeloccy on that workflow, with your team in the loop.

Findings in your Slack, with the query behind each number. Changes drafted for your review and run through your own tools. A direct line to us, and a say in what we build next.

Become a design partner Or email us

Somewhere in your numbers right now,
something is quietly costing you.
Only one way to find out how much.

Talk to us

Become a design partner

We’re working with a small number of teams before a wider release. Tell us a little about yours and we’ll reply personally, usually the same day.