1 / 12 ×

Spotonix

FIRST MEETING · 30 MIN

INTELLIGENCE ON YOUR OBSERVED WORKLOAD

From the queries you already run.
A faster, cheaper, safer warehouse.

Builds the case. Lands the change. Measures predicted vs realized.

LEARNS the workload

Six months captured. The Profile.

BUILDS the case

Ranked, falsifiable. Evidence + ROI + blast radius.

LANDS the change

A dbt pull request. Dry-run, undo, approval.

PROVES the forecast

Cohort vs control. The Health Record.

SPEAKER NOTES — SLIDE 1 · 2 MIN · NEEDLE: BUDGET HOLDER COMMITMENT

"Thanks for the time. We'll cover three things in 30 minutes — why your warehouse bill keeps growing and nobody can fully fix it, what we built differently, and a live look at recommendations on a real query history. The last seven minutes are yours."

"What you'll see: an Advisor that reads every query that actually ran, drafts the fix as a dbt pull request, and grades its own forecast against a control cohort of similar untouched queries."

"The Context Graph is the substrate; the Recommendation Card is the artifact your team reviews; the Health Record is the proof your CFO reads."

"What I want you to walk away with: not a buy decision. A conviction this is a category worth a thirty-day preview — read-only credentials, six months of query history in, ranked backlog out."

Spotonix

THREE OUTCOMES A LEADER CARES ABOUT

Spend less. Go faster. Stay in control.

SPEND LESS

Find the waste no one has time to hunt.

Idle warehouses, redundant scans, materializations that cost more to refresh than they return — each a dollar-quantified case.

GO FASTER

Speed up the queries that matter.

Ranked by impact, so dashboards and decisions stop waiting on data.

STAY IN CONTROL

Nothing changes without approval.

Every change is a reviewable dbt PR with an undo window — graded predicted vs realized.

Meaning, turned into action. Spend less, go faster, stay in control.

SPEAKER NOTES — SLIDE 2 (OUTCOMES) · 45 SEC · NEEDLE: ALL THREE BUYERS

"Before we get into the architecture, here is what changes for the people in your org if Advisor works the way I'll describe."

"Spend less. Advisor finds the waste no one has time to hunt — unused materializations refreshing thousands of times a month, warehouses sitting idle, redundant scans. Each surfaces as a dollar-quantified Case."

"Go faster. Advisor ranks recommendations by impact on the queries that actually matter, so dashboards and decisions stop waiting on data."

"Stay in control. Every change is a reviewable dbt PR with an undo window, and Advisor grades predicted vs realized after the change merges. Your team approves; Advisor measures."

"The line I want you to walk away with: meaning, turned into action."

Spotonix

THE PROBLEM

Your warehouse bill is a
“we’re not sure” tax.

FinOps dashboards

Continuous · Passive

Show you the bill, not the fix

Consultant audits

Rigorous · One-shot

Stale before the invoice clears

Auto-tuners

Fast · Opaque

Change things in the dark — one vendor, no undo

Every team overspends on purpose. Because overspending feels safer than breaking something.

Most data teams pay for materializations they don't use, refresh pipelines nobody reads, and right-size their warehouses by guessing. The data exists. Nobody has time to read it.

SPEAKER NOTES — SLIDE 2 · 90 SEC · NEEDLES: BUDGET CONCERN + ENGINEERING SKEPTICISM

"Four ways teams try to bring the warehouse bill down today. Each one solves part of the problem and leaves the rest unsolved."

"Consultant audits deliver a real backlog, but they're weeks, typically six figures, and stale before the invoice clears. FinOps dashboards show you what's expensive, not what to change about it — reading a chart of last month's spend doesn't tell you which materialization to retire. Engineer time is the right answer in theory; in practice your senior engineers don't have a quiet Friday to read query logs. Opaque auto-tuners change your warehouse for you in the dark and only work on one vendor's stack — great until something breaks and there's no undo."

"Pause."

"So most teams overspend on purpose. They keep MVs alive that nobody uses, leave warehouses idle most of the day, and right-size by guessing — because overspending feels safer than breaking something. That's the ‘we're not sure’ tax. It's on every line of your warehouse bill."

"The wall isn't observability. Observability is solved. The wall is action — specific, evidence-backed, reversible action. That's what Advisor builds."

Spotonix

A CASE · ILLUSTRATIVE

One materialized view.
~$4,200/month. Three queries served. Six months observed.

ASSET

`mv_daily_sales_rollup`

materialized view · hourly refresh

QUERIES SERVED

3

over 6 months of observed workload

REFRESH COUNT

~4,320

hourly × ~180 days

REFRESH-TO-SERVE

~1,440 : 1

refresh frequency vs serving frequency

Every Case has this shape. Asset · workload · predicted impact · blast radius · measurement plan — in one artifact.

Spotonix

THE ECONOMICS

Continuous & self-grading. Not one-shot & stale.

	Consultant audit one-shot	FinOps dashboard monitoring	Spotonix Advisor
Time to first ranked fix	weeks	never — you read the bill	days (pilot-dependent)
Cost of finding the next win	typically six figures	license + your team's time	Advisor run + bounded review
Evidence per recommendation	a slide	a hover tooltip	a one-page Case Card
Grades its own forecasts?	No	No	Yes — before/after baseline

Read every query. Apply through dbt. Measure the outcome.

Illustrative comparison based on the Advisor preview operating model vs typical consulting and FinOps engagements. Methodology and customer-specific figures in the pilot proposal.

Spotonix

THREE BUYERS

One product.
Three unlocks.

CFO · FINANCE

Warehouse spend, itemized.

Every recommendation a one-line dollar claim.
Realized savings measured, not assumed.
Quarterly proof, measured before vs after.

VP DATA PLATFORM

Your best engineer, continuous.

Team focuses on novel problems.
Advisor handles the boring wins.
Every change lands as a dbt PR.

SI PARTNER

A clean consulting wedge.

Query history → ranked backlog.
Land with measured savings in 30 days.
Expand into governance + first-cycle execution.

Itemized spend for finance. Continuous leverage for platform. A clean wedge for the SI.

SPEAKER NOTES — SLIDE 5 · 60 SEC · NEEDLE: ALL THREE BUYERS

"Three buyers. One product. Each one unlocks something different."

"The CFO — itemized spend. Every recommendation reads like a line a finance leader can approve. Realized vs predicted, every quarter, measured before vs after. No more vendor-claim savings."

"The VP of data platform — continuous leverage. Your senior engineers spend their time on novel problems; Advisor handles the boring wins. And every change lands as a dbt PR your team reviews like any other — no new console, no new workflow."

"The SI partner — a clean consulting wedge. Query history in, ranked backlog out, measurable results inside 30 days. The SI adds governance, approval flows, first-cycle execution — and gets a natural land-and-expand."

"One product. Three unlocks. Now let me show you what Advisor actually delivers."

Spotonix

WHAT WE DELIVER

A Case Card.
Not another dashboard.

EVIDENCE

Which queries. How often. What % scanned.

The evidence trail a senior engineer would walk.

PROPOSED CHANGE

A dbt pull request.

Drafted, reviewable, reversible — just like every other PR.

HEALTH RECORD

Predicted vs realized.

Cohort baseline. CFO-legible. Filed back to the graph.

One page per fix. Evidence. Change. Proof.

SPEAKER NOTES — SLIDE 6 · 90 SEC · NEEDLES: PLATFORM TRUST + CFO CONFIDENCE

"This is the deliverable. Not another dashboard, not another console. A Recommendation Card — one page per fix. Three sections."

"Evidence. Which queries triggered the rule, how often they ran, which tables they touched, what % of partitions they scanned. The same trail your most senior engineer would walk before signing off — only it's already walked, every query, every table."

"Proposed change. A dbt pull request your team reviews like every other PR. Dry-run output, blast radius, rollback — all in the PR description. No separate console, no new workflow, no black box."

"Health Record. After you merge, Advisor measures the exact queries the change was meant to help against a comparison group of similar untouched queries. Predicted vs realized. Filed back to the Context Graph so the next recommendation starts from what actually happened."

"The line: one page per fix. Evidence. Change. Proof."

Spotonix

THE SIX MOMENTS

A workload, becoming
an operating loop.

PROFILE

Reads the workload, not the bill.

Six months captured. Schema, dbt lineage, BI assets.

LEDGER

Every asset, paired with the queries it serves.

Tables, MVs, warehouses ↔ the queries that use them. The hard part.

CASE

A ranked, falsifiable business case.

Evidence + ROI + confidence + blast radius + measurement plan. Not a tip; a contract.

CHANGE

Lands as a dbt PR.

Dry-run, ownership detection, undo window. Trust ladder you control.

PROOF

Measures the cohort it promised to move.

Predicted vs realized, against a comparison group. Nets out workload drift.

PLAYBOOK

Confirmed outcomes compound.

Successful patterns climb the trust ladder — earned per play, never switched on globally.

Profile → Ledger → Case → Change → Proof → Playbook.

SPEAKER NOTES — SLIDE 7 · 90 SEC · NEEDLE: STICKINESS + TRUST

"This is the operating loop that separates Advisor from every other tool you've evaluated. Six moments, each one a thing a dashboard, catalog, audit, or auto-tuner structurally can't do."

"Profile. Reads how the warehouse is actually used — not what last month's invoice said. Six months of query history captured and resolved against the Context Graph."

"Ledger. The hard part. Maps every asset to the queries that actually use it — tables, materialized views, warehouses, all paired with the workload they serve, how often, and at what filters. A catalog can list your tables. Only the Ledger can pair each one with the queries that touch it and at what frequency — like the materialized view that served 3 queries in 6 months but refreshed ~4,320 times."

"Case. A ranked, falsifiable business case for each change. Evidence, ROI, confidence, blast radius, reversibility, measurement plan — in one artifact. Not a tip. A contract."

"Change. Lands as a dbt PR your team reviews like any other. Dry-run, ownership detection, undo. No separate console."

"Proof. The evidence layer. Measures the exact queries the Case targeted, against a comparison group of similar untouched queries. Predicted vs realized. The before/after baseline nets out workload drift."

"Playbook. Confirmed outcomes become operating memory. Repeated successes climb the trust ladder — earned per play, never globally switched on. Autopilot is a rung you earn, not the destination."

"The full technical setup — how attribution and cohort measurement actually work — is Appendix A. The trust ladder is Appendix B."

Spotonix

WHAT COMPOUNDS

A tool on Day 1.
Infrastructure by Quarter 2.

DAY 1

The first decommission lands. Proof path attached.

MONTH 1

First Health Records arrive. Predicted vs realized.

QUARTER 2

Patterns earn autopilot. Same graph powers Analyst.

One Context Graph. Two products. Each one makes the next sharper.

SPEAKER NOTES — SLIDE 8 · 60 SEC · NEEDLE: STRATEGIC COMMITMENT

"This is the part most prospects don't see until quarter 2. By then they can't unsee it."

"Day 1. Tool. Warehouse + dbt connected, six months of query history captured, the first decommission recommendations on the table within hours. Proof path attached — you save real money in week one."

"Month 1. First Health Records. Predicted savings vs realized, measured before vs after against a comparison group. Your CFO stops asking 'is this real?' because the before/after baseline is in the report."

"Quarter 2. No longer a tool. Infrastructure. Classes of recommendations have earned autopilot. Your team's attention moved up the stack. And the Context Graph — the same one that powers Spotonix Analyst — is now sharper for both products because every Health Record compounds the graph."

"The line: one graph, two products. Each one makes the next sharper."

Spotonix

THE ASK

Two phases.
Read-only first. Change second — only if you opt in.

PHASE 1 · ASSESSMENT · 30 DAYS · READ-ONLY

Read-only creds: one warehouse + dbt project
Six months of query history captured
Top 10 ranked Cases with evidence + predicted impact
Executive readout & go / no-go on Phase 2

PHASE 2 · CHANGE SPRINT · OPTIONAL · ~30 DAYS

You pick one or two low-blast-radius Cases
Each Change drafted as a dbt PR — you approve
First Health Record after merge: predicted vs realized
Measured outcomes per Case + ongoing engagement model

“A FinOps dashboard tells you the bill.
Advisor tells you what to do about it — and measures the result.”

founders@spotonix.com · advisor brief

SPEAKER NOTES — SLIDE 9 · 60 SEC · NEEDLES: BUDGET HOLDER + PLATFORM LEADER

"Pick the warehouse. Connect to your dbt project. We'll be back in this room with your ranked backlog and the first measured savings inside thirty days."

"The preview is read-only. No mutations. The one applied change we propose is your team's call, through a dbt PR you review like any other. The cost of saying yes is bounded; the cost of saying maybe is another quarter on a bill you can't itemize."

The memorized line (15 sec, eye contact). "A FinOps dashboard tells you the bill. Advisor tells you what to do about it — and measures the result."

The 'no' question (45 sec). "Before I ask for a yes — what would make this not a fit for you? Wrong engine? dbt isn't your workflow? Different priority this quarter? I'd rather know now."

The cash-test close (30 sec). "If Advisor worked the way you saw on real query history — ranked backlog, dbt PR, Health Record — what would have to be true for your team to start a 30-day preview in the next 30 days?" Note the answer. Specific conditions = hot. Engine hedges = check the matrix. Workflow hedges = explore the dbt boundary."

The champion question (30 sec). "Who else needs to see this before you can move forward? I'll send a 5-line follow-up within the hour with the deck, the Advisor brief, and the Recommendation Card walkthrough."

Reserve 2 min for Q&A. Log questions we couldn't answer well, especially around the before/after baseline + attribution — those are the technical pillars in Appendix A.

Spotonix

APPENDIX A · HOW IT WORKS

Fingerprint. Attribute. Verify.

FINGERPRINT

Every query, normalized.

Parameter-stripped fingerprint per query; grouped by shape; resolved to tables, models, dashboards, owners via the Context Graph.

ATTRIBUTE

Work allocated to assets.

Plan parsing and sampling allocate scans, bytes, and joins to participating tables and materializations. Every Case's evidence depends on this.

VERIFY

Cohort vs control.

The exact fingerprint cohort the change targets, measured before and after, against a control cohort of similar untouched queries. Diff-in-diff nets out workload drift.

Deterministic recommendations. Before/after readout per Change in preview; productized cohort automation in v2.

SPEAKER NOTES — SLIDE 10 · 90 SEC · NEEDLE: ENGINEERING TRUST

Preview honesty (open with this if asked): "Every Case includes a measurement plan and a manual before/after readout. Fully productized cohort/control automation is the v2 milestone. We will not headline guaranteed savings before that ships."

"Three steps under the hood. This is the appendix because it's how we earn the trust on the prior slides — not where we lead."

"Step 1, Fingerprint. Every query that ran is parameter-stripped into a fingerprint and grouped by shape — so ‘sales last 30 days for region X’ collapses into one workload pattern even when 4,000 queries hit it. Then those fingerprints get resolved against the Context Graph — back to tables, dbt models, dashboards, owners."

"Step 2, Attribute. This is the hardest engineering problem. A multi-table query's work — the scans, bytes, joins — has to be allocated to the participating tables and materializations. Otherwise we can't say 'this asset serves these fingerprints, this often, at this filter selectivity'. We do it by parsing and sampling query plans, not on every query. Attribution is on the critical path because every Case's evidence depends on it."

"Step 3, Verify. When we apply a change, we measure the exact fingerprint cohort the change was meant to help — before and after — against a control cohort of similar untouched queries. A naive before/after on total spend is confounded by workload drift; the diff-in-diff against the control cohort is the defensible savings number."

"Recommendations themselves are deterministic — rule-based decision trees, same evidence in produces the same recommendation out. Natural-language input and labeling assists are non-decisional. The optimization decision is rule-based, never delegated."

Spotonix

APPENDIX B · THE TRUST LADDER

Four rungs.
Each one earned.

OBSERVE

Read-only profiling.

Six months of query history captured. No recommendations surfaced.

ADVISE

Ranked Cases with evidence.

Most teams stay here the first quarter. Every Change gets human review.

PROPOSE

A dbt PR per Change.

Drafted, reviewable, mergeable. Your team approves every one.

AUTOPILOT

Earned classes graduate.

Six landings inside predicted band → the class delegates. With undo. With an audit log.

You decide how much you delegate. Rungs are earned by measured outcomes, never assumed.

SPEAKER NOTES — SLIDE 11 · 60 SEC · NEEDLE: GOVERNANCE COMFORT

"The trust ladder is how Advisor goes from a tool you observe to infrastructure you delegate — without ever giving up undo."

"Rung 1, Observe. We connect, capture every query, build the graph. No recommendations leave the system. This is where every customer starts."

"Rung 2, Advise. Ranked recommendations with evidence cards. Most teams stay here for the first quarter — every fix gets human review. You build trust in the recommendation quality, you build trust in the Health Records, and you learn which classes of recommendation land reliably."

"Rung 3, Propose. Advisor drafts each change as a dbt pull request. Same review workflow your team already uses. You approve every PR before it merges. Nothing autonomous yet — just less manual drafting on your team's side."

"Rung 4, Autopilot. A specific class — say, decommission MVs unused for 60 days, or auto-suspend warehouses idle past a threshold — graduates only after it has landed inside its predicted band six times running. When it does, you can promote that class. With undo. With audit log. Every change still reversible."

"The honest framing. Vendors that sell autopilot before you've earned trust in the recommendations are selling you a black box. We refuse to do that. Trust is earned per rung, by measured outcomes."

From the queries you already run. A faster, cheaper, safer warehouse.

Spend less. Go faster. Stay in control.

Find the waste no one has time to hunt.

Speed up the queries that matter.

Nothing changes without approval.

Your warehouse bill is a “we’re not sure” tax.

One materialized view. ~$4,200/month. Three queries served. Six months observed.

mv_daily_sales_rollup

3

~4,320

~1,440 : 1

Continuous & self-grading. Not one-shot & stale.

One product. Three unlocks.

Warehouse spend, itemized.

Your best engineer, continuous.

A clean consulting wedge.

A Case Card. Not another dashboard.

Which queries. How often. What % scanned.

A dbt pull request.

Predicted vs realized.

A workload, becoming an operating loop.

Reads the workload, not the bill.

Every asset, paired with the queries it serves.

A ranked, falsifiable business case.

Lands as a dbt PR.

Measures the cohort it promised to move.

Confirmed outcomes compound.

A tool on Day 1. Infrastructure by Quarter 2.

Two phases. Read-only first. Change second — only if you opt in.

Fingerprint. Attribute. Verify.

Every query, normalized.

Work allocated to assets.

Cohort vs control.

Four rungs. Each one earned.

Read-only profiling.

Ranked Cases with evidence.

A dbt PR per Change.

Earned classes graduate.

From the queries you already run.
A faster, cheaper, safer warehouse.

Your warehouse bill is a
“we’re not sure” tax.

One materialized view.
~$4,200/month. Three queries served. Six months observed.

`mv_daily_sales_rollup`

One product.
Three unlocks.

A Case Card.
Not another dashboard.

A workload, becoming
an operating loop.

A tool on Day 1.
Infrastructure by Quarter 2.

Two phases.
Read-only first. Change second — only if you opt in.

Four rungs.
Each one earned.