INVAR：面向产品驱动团队的由内而外分析工具

Why did activation fall for campaign B?

IN YOUR NUMBERS

EXAMPLEPostHog

activation_rate · campaign B

24.0%

▼ 41%

Most accounts stop at Salesforce setup.

In your team's conversations

EXAMPLESlack

Jo, Customer Success #cs-escalations

Another customer stuck mapping Salesforce fields. Who knows this one?

40 Salesforce questions to Eng this month1 person answers most

EXAMPLE

TOP CHANGE THIS WEEKCorrelation

Same problem: Salesforce setup

The accounts that stall at setup are the ones CS keeps escalating. Fix it once and both numbers should move.

9 linked recordsInvestigate

INSIDE-OUT ANALYTICS FOR PRODUCT-LED COMPANIES

A confusing integration shows up as an activation drop in PostHog and as the same question hitting one engineer in Slack. INVAR connects the two, tells you what to fix, and proves whether the fix worked.

PostHog · Stripe · HubSpot · Slack. Connected in under 15 minutes. First finding within 24 hours.

THE PROBLEM

Each tool sees half. Nobody is paid to join them.

Teams dig by hand for weeks, the same expert keeps getting pulled in, and the fix lands on the symptom instead of the cause.

POSTHOG

knows users stall at setup.

STRIPE

knows those accounts never upgraded.

HUBSPOT

knows they came from one campaign.

SLACK

knows one engineer answers every question about it.

INVAR knows it's one problem, and which fix to try first.

HOW IT WORKS

From a metric that moved to a fix that's proven

One loop, start to finish. Every pass writes back what was tried and what worked, so the next investigation starts smarter.

01

Connect

Read-only connections plus your goals and metric definitions.

Under 15 minutes

02

Investigate

Ask why in plain words. Causes ranked, every angle checked, every number linked to its query.

Under 2 minutes

03

Decide

The owners of each angle weigh in next to the evidence. INVAR learns your context.

Owner, target, baseline

04

Act

Approved in INVAR, shipped from Slack to Jira, Linear or GitHub Issues.

Nothing without approval

05

Prove

The target metric and question load measured against baseline, with a recorded verdict.

Both sides, measured

EXAMPLE DATA

INVAR

EXAMPLE CO.

Top changes

Investigations3

Recommendations2

Outcomes2

LENSES

Growth

Product

CS and Support

Engineering

Why did activation fall for campaign B?

Activation fell 41% for campaign B accounts. Most of the drop is associated with one step: Salesforce setup.

LIKELY CAUSES, RANKED

1  Campaign B accounts stall at Salesforce setup

Correlation

71% of the drop · passed seasonality, cohort size and plan mix · 212 accounts

2  Fewer accounts on the Growth plan

Fact

18% · plan mix shifted 6 points

3  Release 4.2 navigation change

Hypothesis

4% · too small to separate from noise

EVERY ANGLE CHECKED

Usage68% of campaign B accounts stop at field mapping

RevenueAccounts that finish setup convert 2.3x more often

AcquisitionConcentrated in LinkedIn campaign B

Team load40 Salesforce questions to Eng; 1 person answers most

KnowledgeSetup guide is 7 months old

CostNothing unusual

Sales contextNothing unusual

GoalsAdd your goals doc to check this angle

CONNECTIONS

11 of 15 min

PostHog

Events, persons, feature flags

Connected

Stripe

Subscriptions and invoices

Connected

HubSpot

Deals, campaigns, owners

Connected

Slack

Question metadata only; text stays in Slack

Installed

Company context

Goals 2026, metric definitions, pricing

3 docs

Plug in what you already use

Your goals and metric definitions go in too, so every answer uses your numbers, not a generic formula. No migration, no tracking plan rewrite.

RECOMMENDATION

Correlation

Simplify Salesforce setup and publish one field-mapping guide

Owner ProductTarget activation_rate 24.0% → 32.0%

CS lead · 2h
Half of these accounts came from the LinkedIn webinar. They expect setup to be done for them.

Engineering · 1h
The field-mapping screen hasn't changed since v3. Happy to pair on the guide.

ApproveChangeDefer

The right people weigh in

Each recommendation opens a short thread with its owner. When someone says "that spike was a one-time migration," INVAR learns it for your company.

Approved in INVAR, shipped from Slack

INVAR drafts the ticket and posts it to the owner with the evidence, a suggested assignee and the target metric. Nothing is created without a person's approval.

INVAR APP 10:42

Drafted a ticket from the approved recommendation. Review it, then create.

Simplify Salesforce field mapping in setup

Campaign B accounts stall at field mapping; 40 related questions reached Engineering this month.

Integrations teamactivation_rate → 32.0%9 linked records

Create in LinearEdit first

EXAMPLE

Simplify Salesforce setup

Watching

24.0% → 29.5% · target 32.0%

Day 18 of 30 · Salesforce questions 40 → 22 a month

Write the field-mapping guide

Hit

40 → 14 questions / mo · target 20

Closed Sep 30 · before and after, no holdout

Know whether it worked, on both sides

That history of what you tried and what worked stays with your company.

ASKRIGHT · FOR YOUR TEAM

Ask once. Reach the person who knows.

AskRight is INVAR's Slack app. DM it or mention it in a channel.

# cs-teamSlack

Sam 10:02

@AskRight does the Salesforce sync support custom objects?

AskRight 10:02

Yes, from version 4.1. Custom objects need the "Advanced sync" toggle in Settings.

Salesforce setup guide · Sep 30#eng-integrations thread

Jo 10:15

@AskRight Halden Co. sees duplicate contacts after a re-sync. Known issue?

AskRight 10:15

No documented answer yet. Routing to Priya on Integrations, who has answered 12 sync questions this month. Fallback: the Integrations team.

Salesforce integrationHalden Co.

Repeat questions, answered

From your approved docs and past answers, with sources.

New ones go to the right person

With a fallback owner per topic, and load spread so one expert isn't buried.

Gaps turn into docs

Topics with no doc or a single answerer show up in a weekly gap report.

Reports show teams and topics, never individual scores. An expert is named only to route a question to them.

WHAT ONLY INVAR SEES

Answers that need both sides of your company

Any analytics tool reports activation and churn. These need customer data and team conversations joined at the account and feature level.

True cost to serve

Plans that look profitable but quietly consume support and engineering time.

Stripe + AskRight

Revenue at knowledge risk

Paying accounts whose issues only one person knows how to solve.

Stripe + AskRight routing

Silent regressions

Releases that left product metrics flat but spiked internal questions.

PostHog releases + AskRight

Confusion or failure

Whether users can't find a feature, or find it and get stuck.

PostHog usage + question topics

Expansion hidden in questions

Accounts outgrowing their plan before they ever ask Sales.

AskRight + PostHog limits + Stripe

Slow answers that cost revenue

Accounts whose escalations took longest before they shrank or left.

AskRight response time + Stripe

ONE BRAIN, A LENS PER TEAM

Every team starts from its own question

Lenses are entry points, not walls. Each gets saved questions and a weekly digest, and every investigation still checks every angle.

TOP CHANGE THIS WEEK{{lens.claimName}}

{{lens.top}}

{{lens.sources}}

SAVED QUESTION

{{lens.question}}

WHAT INVAR ADDS

{{lens.adds}}

People Ops lens, for new-hire questions in the first 30 days, joins during the pilot.

TRUST AND PRIVACY

It sees a lot of your company. So it touches as little as possible.

Slack stays in Slack

Message text is fetched to answer a question, then discarded. INVAR stores only metadata.

Your own key

A separate warehouse and encryption key for each company. Delete it and the data is unreadable, with a signed certificate.

Kept inside your company

Chat content, context and everything INVAR learns stay in your company's own space. Nothing crosses to another customer or trains a model.

Your permissions, kept

Every answer respects what the asker can already see. Private channels stay private.

Not a monitoring tool

INVAR doesn't track individual messages or report on anyone's performance. No scores, no rankings. Reports show teams and topics, and an expert is named only to route a question to them.

Approval and audit

Nothing is written to another tool without a person's approval, and every action is logged.

Security questions: security@invarhq.com

HONEST BY DEFAULT

INVAR only says "caused" when a test backs it

Numbers come from queries on your data, never from the model. Every claim carries one of four labels, so you know how much weight it can bear.

Fact

Seen directly in your data.

Correlation

Moves together. Not proven to be the cause.

Hypothesis

Plausible. Needs more evidence before you act on it.

Tested

Backed by a holdout or experiment. The only label that says "caused".

WHERE INVAR FITS

Keep your tools. INVAR does the part between them.

Built to work alongside your analytics tools, not replace them.

DESIGN PARTNER PILOT

First finding in a day. A verdict in 60.

For product-led companies of 30 to 150 people. Best fit today is teams on PostHog, Stripe, HubSpot and Slack. On a different stack? Reach out anyway. We're choosing the next integrations based on what pilot teams use. We're taking 10 to 12 teams.

1. Day 1Connect your tools and install AskRight in Slack.
2. Day 2Your first finding, with the evidence behind it.
3. Week 2We agree one target metric and set its baseline.
4. WeeklyA digest with what changed and what to try next.
5. Day 30Checkpoint. Keep going or stop, either side can call it.
6. Day 60Review the target together, then decide on conversion. Still measuring a fix? We extend that one outcome, not the whole pilot.

Unlimited viewers and askers. No per-seat pricing.

Thanks, we got it.

A founder will email you within two working days.