The operating system for sales-engineering leaders

Run the Monday review
in fifteen minutes.

Evivant reads Salesforce, Calendar, and Gong, ranks the five deals worth talking about, tells you the next move on each, and remembers what you decided last week — so your SEs never log a thing.

Read-onlyEncrypted at restScoped to your orgZero SE data entry
Apply for Early Access

Live, seeded demo — no signup, no Salesforce connection. You’re inside in one click. Or email hello@evivant.io.

What this replaces

Your Monday call is half storytelling. Evivant makes it evidence.

Presales runs a weekly deal review whether a tool helps or not — a Salesforce scan, a Slack scramble, a POV doc nobody opens. Most of that loop is wasted. Evivant is built around the loop instead of bolted onto it.

The Sunday-night Salesforce scanreplaced by →
A ranked top-five with a one-line reason on each, waiting for you Monday morning.
The Monday “what actually happened?” Slack threadreplaced by →
Recall narrates what moved since last Monday — and whether the call you made was right.
The POV doc nobody opens after week tworeplaced by →
A live POV workspace pinned to the deal, with a buyer-ready scorecard you can hand over.
Post-review amnesiareplaced by →
Every decision becomes a tracked commitment, graded against what actually happened.

Not a forecasting tool — Clari does that. Not conversation intelligence — Gong does that. Evivant is the operating loop that sits on top of both, and turns their raw signal into the five decisions you actually make on Monday.

POV Workspace

Run the POV in the deal, not in a doc nobody opens.

Your SEs stop re-keying status into spreadsheets. Buyers see a live, honest scorecard through one expiring link — and you can see exactly what they engaged with.

POV-in-the-deal

The evaluation is a tab on the Salesforce Opportunity — not a sibling tool. Activity, conversation signals, and stakeholders flow in automatically.

Linked deals + net-new prospective POVs

Import, don’t re-key

Paste the customer requirements spreadsheet. Evivant parses categories, expected results, and SE-note status, then auto-creates the validation items. No re-typing the RFP.

Up to 500 criteria per batch

Buyer-ready Proof Room

A tokenized, read-only, no-login scorecard reachable through one unique link with a default 7-day expiry — rate-limited and non-enumerable. Internal evidence is filtered out at the database boundary.

Forward-only · no path from a note to a link

A structured evaluation tree

Weighted criteria (Must-have / Important / Nice-to-have), checklists, owners, and per-node validation capture: who validated it, when, and by what method.

Validation method
DemoDocumentationVerbalEmailMeeting
Evidence visibility
InternalBuyer-visibleBoth

Six grounded verdicts, never a binary

Every claim a buyer asks about resolves to a graded verdict — with the evidence behind it, or an honest “we don’t know yet.”

VerifiedPartially VerifiedNot VerifiedContradictedInsufficient EvidenceNeeds Human Review

Buyer engagement surfaces as sanitized, PII-stripped signals — a reaction, a question, a re-open — never raw buyer prose.

Artifacts the champion can take internal

Champion Pack
CFO Brief
Tech Evaluator Deep Dive
Kickoff & Closure Plan
Weekly Buyer Update

Each is assembled deterministically from validated outcomes — every section traces to evidence, nothing is fabricated. Citations only render when the source link is HTTPS and on a fixed allowlist of vendor and public-docs domains, so a pasted internal Confluence or S3 link can never reach a buyer PDF.

Leadership Analytics

See which deals to watch — and your real pipeline, not rep self-report.

Half your Monday call is rep storytelling. Evivant discounts raw pipeline down to what the evidence supports, and ranks exactly which at-risk deals to work this week — within a realistic capacity.

Forecast-truth waterfall

Illustrative · demo org

Raw pipeline is discounted by proof posture into disjoint buckets. The protected buyer-validated band at the bottom is the number you can defend to the CRO.

Raw pipeline
$9.6M
− without a live POV
−$0.9M
− stakeholder unmatched
−$0.7M
− internal-only proof
−$0.6M
= Thin proof (< 2 evidence nodes)
$1.4M
= Buyer-validated (≥ 2 nodes)
$6.0M
Quality-adjusted forecast
$7.4M

Which deals to watch

Recovery moves are ranked by expected value — deal amount × win-probability lift — across at-risk deals, then packed into a capacity-capped week.

2.2h
1.9h
1.4h
.9h
Weekly capacity · 8h default78% of at-risk pipeline covered

A move only appears when it adds ≥2 percentage points of win probability — ranked by lift per hour, grounded in your own won-deal medians.

Pipeline cleanliness

Illustrative

Deals stuck 1.5× (aging) or 2.0× (zombie) the typical time in their stage — using per-stage medians learned from your closed deals.

Clean · 58%
Aging 22%
Zombie
Base
CleanAging (1.5×)Zombie (2.0×)Still baselining

Stages without enough history are reported as still baselining — never silently counted as clean.

Concentration risk

Illustrative

How much of the forecast leans on too few accounts — and what happens if your biggest deal slips.

0.28
Herfindahl index (HHI)
41%
Top-account share of pipeline
−$2.1M
Forecast drop if largest deal slips

Every leadership number self-discloses a cold or building state when the data is thin. Evivant would rather say nothing than fabricate a figure — so the numbers you take into the forecast call are numbers you can stand behind.

Why we lose: closed-lost deals are classified into seven origin categories with per-deal evidence — descriptive, what we observed, never a causal root-cause claim. Capacity is operational planning, never per-rep performance scoring.

How it works

The engine underneath — and how we earn the numbers.

Every formula, threshold, and schema below is the same code that runs in product. The point isn’t the math for its own sake — it’s that you can always ask why and get an answer, not a vibe.

40+
analytics engines · 6 families
12
ML features · fully explainable
84
data models · every record org-scoped
15
minute sync cadence
01

Every prediction comes with its reasons.

You can always ask why a deal is at risk and get a formula, not a feeling. The win-probability score is a real per-customer model — and every score decomposes into the top forces pushing it up and pulling it down, rebuilt live so the explanation always matches the deal’s current state.

68%
P(win) · this deal
Decision friction-0.18
+0.32Health Score
Evaluation gap-0.11
+0.21Champion Health
Timeline slippage-0.07
+0.14Buying-Group Coverage
← pushing toward losspushing toward win →

A per-customer logistic-regression model trained on your own closed deals — plain TypeScript, no external ML library, no opaque service. Twelve deal-shape features, cross-validated, scored on a held-out Brier metric. Every coefficient is auditable.

Under the hood · the model
// logistic regression, batch gradient descent + L2 sigmoid(x) = 1 / (1 + e^(−x)) P(win) = sigmoid( w · x_norm + b ) contribution[i] = weight[i] × (norm[i] − 0.5)

Trains only on ≥30 closed deals with full outcome snapshots, and ≥8 won and ≥8 lost. It rejects its own trained model if it can’t beat the naive majority-class base rate — a model no better than guessing the common case never ships. Each org trains on its own outcomes; no shared weights.

The 12-feature vector
#FeatureMeaning
0healthScoreComposite deal health
1threadingScoreStakeholder coverage
2championHealthKey-contact engagement
3committeeScoreBuying-group breadth
4dysfunctionScoreDecision friction
5technicalDebtUnresolved evaluation gap
6activityCount30dMeetings + calls (30d)
7meetingCount30dMeetings only (30d)
8closeDatePushTimeline slippage
9uniqueStakeholdersDistinct externals
10stageProgressPipeline position
11totalActivitiesLifetime engagement

Each feature is min-max normalized against stats persisted with the model; out-of-range inputs clip to the training range, missing values impute to the training median.

02

We discount raw pipeline to what the evidence supports.

The forecast-truth waterfall (shown above) and pipeline-cleanliness bands run on per-stage velocity medians learned from your own closed deals — with a measurable-only denominator, so a stage you haven’t run enough of is reported as baselining rather than quietly counted as healthy.

03

The product refuses to state a number it hasn’t earned.

This is the differentiator. Predictions are logged automatically, graded against real outcomes when deals close, and thresholds recalibrate weekly from your own graded deals. Auto-detected actions stay guesses until a human confirms them — operator clicks never silently feed the accuracy numbers. And when you flag a deal, act, the health recovers, and it wins, that’s graded a save, not a false positive.

12345closedloop
1
Predict
Logged with type, value, and the exact feature vector that produced it.
2
Expose
Recorded each time a prediction is actually shown to an operator.
3
Act
The move the operator made, with the observed result.
Confirmation gate
An inferred action→outcome link only counts toward learning once a human confirms it. Auto-grading noise can never pollute the truth.
4
Grade
correct · correct-but-mitigated · wrong · partial · not-judgeable.
5
Recalibrate
A weekly sweep retunes each threshold within governed bounds — every attempt written to an append-only ledger.

Sample floors, stated plainly

Below these thresholds, Evivant shows a transparent cold state — never a fabricated number.

ClaimFloor before it publishes
Accuracy shown at all20 resolved predictions
Published rate (±15pp band)40 graded outcomes
Tight hero metric (±10pp band)96 graded outcomes
Leader-grade accuracy claim150 graded predictions
Win-probability model trains30+ closed · ≥8 won · ≥8 lost
Browse the engine catalog · six families, 40+ engines

The part that makes people say “wait — this isn’t a CRM skin.” Each engine is a mostly-pure function that reads a deal plus its activity graph, computes a value, and writes it back so the UI renders from a single query. Hover, focus, or tap any tile to read its method.

Health & scoringhow alive each deal is5 engines
Risk & forecastingwhat's slipping and why6 engines
POC / POV lifecyclethe evaluation engine room5 engines
Engagement & behavioralwho's actually in the room4 engines
Prescriptive & signalthe next move5 engines
Portfolio & governancethe leadership & honesty layer7 engines

Built on 84 multi-tenant data models, every record scoped to your org. Engine names are drawn directly from the live codebase.

Every formula, threshold, and schema shown here is from the product. Sample dashboards on this page use illustrative data from the live demo org — not customer results.

Who it’s for

Built for the VP/Director of SE — and the SEs who never log in.

VP & Director of Sales Engineering · 5–30 SEs · B2B SaaS · running 2–10 concurrent POVs.

For the VP / Director
  • Which five deals to watch this Monday, with the reason on each
  • True, proof-discounted pipeline health — not rep self-report
  • The forecast-truth waterfall you can defend to the CRO
  • Decision Recall — what moved against last week’s calls
  • Concentration risk before too much rides on one account
For your SEs
  • Zero data entry — required fields are forbidden by design
  • The POV lives in the deal, not in a doc that goes stale
  • A buyer-ready scorecard you hand over and can track
  • The Fix-This next move, phrased once from the deal’s own signals
  • No login required to get value — it runs on what already exists
Not a forecasting tool — Clari does thatNot conversation intelligence — Gong does thatNot a CRM replacement — and it never writes to your CRMNot a buyer portal
Security & trust

We read. We never write. Every byte is scoped to your org — enforced in code, not policy.

The trust story a VP SE’s security team needs to greenlight access — stated honestly, including what we don’t do yet.

Read-only access model

Read-only OAuth on Salesforce, Google Calendar, and Gong. We never write a record back to your CRM. Each integration is a separate, revocable grant — disconnect anytime and access stops immediately.

Data we never store

We do not store raw call recordings or audio, transcript bodies, or LLM prompt and response text. Behavioral engines run on auto-matched activity metadata, not message content.

Encryption at rest

OAuth tokens are encrypted at rest with AES-256-GCM; in production the app fails closed if the key is missing. Tokens are never logged and never exposed to the browser.

Tenant isolation

Every database query is scoped to your organization at the database layer — and enforced by an automated test that fails our build, not by convention. Raw SQL is banned across the server codebase.

Operational posture

Sync runs every 15 minutes with an audit log per run. Background jobs are authenticated with a shared secret and per-org checks. Logs auto-redact secrets. Sessions are stateless JWTs with a 7-day lifetime.

Report an issue

Found something? Email security@evivant.io and we’ll respond quickly. Read the full posture on our Security page.

Subprocessors · what each one receives

SalesforceCRM read source
GoogleCalendar read source
Gongconversation metadata
Slackbriefing delivery (opt-in)
OpenAItext generation · no stored prompts
AWS Bedrockgrounded claim verification
NeonPostgres database
Vercelapplication hosting
Sentryerror reporting (redacted)
Resendtransactional email

Compliance posture, kept honest.We hold no formal certifications yet. We do not claim SOC 2, ISO 27001, HIPAA, GDPR controls, SSO/SAML/SCIM, or data-residency guarantees. What we can tell you is exactly how your data is handled today — and we’d rather state that plainly than imply a badge we haven’t earned.

Built on a current mainstream stack — Next.js 16, React 19, Prisma 7 on Neon Postgres — backed by a deep data model (84 models) and over 1,400 automated test files, most of them architecture-enforcing guards rather than happy-path checks.

Company

Why we built Evivant.

Presales runs a weekly deal review whether a tool helps or not. Reps self-report, and the number lies. We built Evivant so a sales-engineering leader can see the truth — which deals to watch, what’s actually been validated, what to do next — and so SEs never have to log a thing.

Honesty is a product principle, not a tagline. Evivant underclaims when data is thin, grades its own predictions against real outcomes, and tells you when it isn’t sure. The numbers it shows are numbers you can defend.

We’re in early access with design-partner SE teams. We don’t show logos we don’t have — and we won’t pretend otherwise to look bigger than we are.

Talk to us: hello@evivant.io

Early Access

Early Access is a partnership, not a trial.

A small cohort of SE leaders shaping the Monday operating loop with us — with the full product, real influence on the roadmap, and a direct line to the founders.

Who qualifies

VP / Director of Sales Engineering · 5–30 SEs · B2B SaaS · running 2–10 concurrent POVs.

What’s included

The full product against your Salesforce (read-only), white-glove onboarding, and a direct line to the founding team.

The posture

No credit card. Salesforce stays read-only. Salesforce alone is enough to start — Calendar and Gong sharpen the scoring.

Apply for Early Access

Demo: explore now, no signup.   Early Access: shape the roadmap with us → hello@evivant.io

FAQ

The questions a VP SE asks before connecting Salesforce.

Does it write to my CRM?
No. Read-only, always — we never write a record back to Salesforce.
Do my SEs have to log in or enter data?
No. Required fields are forbidden by design; Evivant runs on activity that already exists. Your SEs get value without ever logging in.
What if we don't use Gong?
Salesforce alone is enough to start. Calendar and Gong sharpen the scoring, but they're not required to get value on day one.
How is my data isolated from other customers?
Every query is scoped to your organization at the database layer, and that scoping is enforced by automated tests that fail our build — not by convention.
What happens to my data if we leave?
You can disconnect anytime. Revoking the OAuth grant stops all access immediately.
Is the AI making things up?
Predictions are explainable and evidence-typed. The product underclaims when data is thin and grades itself against real outcomes — it would rather show a cold state than a fabricated number.
Are you SOC 2 or HIPAA certified?
Not yet — we hold no formal certifications and we say so. Here’s exactly how your data is handled today: read the Security page.
Try the loop before you commit

Walk the Monday Ritual, open a real POV, and see how Recall reads — before connecting Salesforce.

Apply for Early Access

No credit card. Salesforce stays read-only. Most operators start with the demo.

Demo: instant, no signup.   Early Access: talk to us → hello@evivant.io