Forward MetricsBook a call

Data engineering · Analytics · Power BI · AI

Your numbers should end the argument, not start one.

Forward Metrics builds the data foundations, models and reporting that give growing businesses one trusted version of the truth — then teaches your team to work far faster with AI.

Book a discovery call See what we build

30 minutes · No pitch deck · US & UK

August revenue, as reported3 answers
  • Financeboard_pack_v7.xlsx$4.82M
  • Operationsops_dashboard.pbix$5.07M
  • Salescrm_export_0902.csv$4.61M
$460K apart. Each team is sure they’re right.
board_pack_FINAL_v7 (2).xlsx“Which revenue number is this?”ops_export_0831.csv“Let me check the numbers and come back to you”Copy of Copy of KPIs.xlsx“Only one person knows how that report works”manual_refresh_DO_NOT_TOUCH.pbix“Can someone rebuild the deck by Friday?”board_pack_FINAL_v7 (2).xlsx“Which revenue number is this?”ops_export_0831.csv“Let me check the numbers and come back to you”Copy of Copy of KPIs.xlsx“Only one person knows how that report works”manual_refresh_DO_NOT_TOUCH.pbix“Can someone rebuild the deck by Friday?”

Example build · Executive Pulse

Monday, 8:02am.
The numbers are already there.

Leadership opens one page. Nobody prepared it and nobody argues with it. Change the period or the region — it’s live.

Executive PulseExample
Revenue · QTD
$0.00M
+1.2% vs target
Gross margin
0.0%
-0.3 pts vs plan
Orders · QTD
0
avg $1,890 per order
On-time delivery
0.0%
below 95% SLA
Weekly revenue vs target
Last 13 weeks
At or above target Below target Target

The work

Don’t take our word for it. Click around.

Five things we build, working in your browser. They all run on synthetic data for Harbor & Pine Distribution, a made-up building-supplies business with about 180 people across the US and UK. The company is invented; the approach isn’t.

01/05Foundation · Data engineering

Pipeline Control

The problem
Five systems, joined by hand in a spreadsheet every Monday.
What we built
Scheduled pipelines into a Fabric lakehouse, with checks that catch a bad file before a report goes wrong — and tell the person who owns it.
What changes
Adding the next data source is a small job, not another project.
  • Microsoft Fabric
  • Azure Data Factory
  • SQL
  • APIs
Pipeline ControlExample
NetSuiteGL · AR · invoicesSalesforceaccounts · oppsShopify B2Borders · refundsWMS · SQL Servershipments · stockBudget_FY26.xlsxSharePointFabric Lakehousebronzeraw, as landedsilvercleaned, dedupedgoldbusiness-readySemantic model42 measures1 definition eachExecutive PulseMonth-end packOps dashboard
Run loghealthy
  1. yesterdayRun complete · 2m 09s · 1.83M rows · 0 failures
02/05Trust · Analytics engineering

Metric Contracts

The problem
Sales, Finance and Ops each had a correct answer to the same question.
What we built
A semantic model where every metric is defined once, signed off by the people who use it, and reused in every report.
What changes
Decisions get made in the meeting, not deferred to “let me check the numbers”.
  • Semantic model
  • DAX
  • KPI framework
  • Governance
Metric ContractsExample
Question from the Monday meeting
How many active customers do we have?
SalesSalesforce report
0

Any account with an open opportunity or an order in 12 months

FinanceBoard pack, tab 3
0

Invoiced in the last 12 calendar months, credit notes included

OperationsOps dashboard
0

Shipped to in the last 90 days

Three answers, 222 apart. Each one is correct by its own rule — which is why the meeting ends in “let me check the numbers”.
Monday leadership meeting · 9:14am
  • Sales“We're at 1,412 active customers. Best year yet.”
  • Finance“My board pack says 1,284.”
  • Operations“I've got 1,190, and that's from actual shipments.”
  • CEO“So which number goes to the board?”
  • Finance“Let me check and come back to you.”
03/05Decisions · BI & Power BI

Month-End Pack

The problem
The first week of every month went on rebuilding the same board deck.
What we built
A close workflow that reconciles to the ledger, drafts commentary for Finance to edit, and builds the pack straight from the model.
What changes
The deck lands the same day every month.
  • Power BI
  • Semantic model
  • Automation
  • Executive reporting
Month-End PackExample
  1. Pull August GL actualsNetSuite · 18,406 journal lines
  2. Reconcile revenue to GL and ARDifference $0.00
  3. Refresh semantic model42 measures · fiscal calendar
  4. Draft variance commentaryAI draft · reviewed by Finance
  5. Build board pack6 slides, house template
  6. Send to board and leadership7 recipients · PDF + live link
At a glanceBoard pack · August
Revenue
$4.91M
+6.3% vs Jul
Gross margin
31.8%
−0.3 pts vs plan
EBITDA
$0.52M
10.6% of revenue
Cash
$3.40M
41 days of cover
04/05Leverage · AI & analytics training

Ask the Warehouse

The problem
Analysts were already pasting data into ChatGPT, with nothing checking the answers.
What we built
AI-assisted SQL inside controlled boundaries. The AI drafts, automatic checks tie every answer back to the model, and an analyst approves it.
What changes
Watch the first question: the draft runs, looks right, and is 3.2% wrong. The checks catch it.
  • Claude & ChatGPT
  • AI-assisted SQL
  • Fabric warehouse
  • Review workflow
Ask the WarehouseExample
You asked
Drafting SQL…
Checks before anyone sees a number
  • Query runs
  • Same period both years
  • Totals tie to the semantic model
05/05Leverage · AI & analytics training

Pricing Sandbox

The problem
Every pricing what-if waited in a queue for someone technical.
What we built
A working app on top of the semantic model — the kind of thing an analyst can build in an afternoon once they know how to use AI tools properly.
What changes
Teams answer their own questions instead of joining a reporting queue.
  • Analytics apps
  • Claude
  • SQL
  • Semantic model
Pricing SandboxExample
Customer segment

Starting sensitivity comes from 24 months of contractors order history. Change it to test a what-if.

Volume
−4.1%
Revenue / month
−$24K
Margin / month
+$33K
Gross margin
32.2%
Monthly margin change across price movesDot = your scenario
more marginless margin
-8%-4%0%+4%+8%

A 3.0% increase loses 4.1% of contractors volume and adds about $33K of margin a month.

What we do

Four capabilities. One continuous chain.

Most firms sell you a single link — a dashboard here, a pipeline there. We build the whole chain from source system to decision, or plug cleanly into the part you’re missing.

  1. Raw dataTrusted data

    01 · Foundation

    Data Engineering

    Get the data out of your systems, into one place, on a schedule, without anyone touching it.

    • Microsoft Fabric
    • Azure Data Factory
    • ETL / ELT
    • SQL
    • APIs
    • Warehousing
    See Pipeline Control working ↑
  2. Trusted dataOne definition

    02 · Trust

    Analytics Engineering

    Turn raw tables into business logic everyone agrees on, so a metric means one thing everywhere.

    • Data modeling
    • Semantic models
    • KPI frameworks
    • Governance
    See Metric Contracts working ↑
  3. One definitionBusiness insight

    03 · Decisions

    BI & Power BI

    Reporting built around the decisions you actually make — fast, trusted, and still used after week one.

    • Power BI
    • Executive dashboards
    • Operational reporting
    • Self-service
    See Executive Pulse working ↑
  4. Business insightBetter decisions

    04 · Leverage

    AI & Analytics Training

    Teach the analysts you already have to do a week's work in a day — safely, with their judgment intact.

    • AI-assisted SQL
    • ChatGPT & Claude
    • Power BI + AI
    • Agents
    • Analytics apps
    See Ask the Warehouse working ↑

Sound familiar?

Most reporting problems aren’t reporting problems.

By the time a dashboard is wrong, the damage was done three layers upstream. Under the hood of a mid-sized business, we usually find some version of these.

  • Two reports, two answers

    Finance says one number, operations says another. The meeting becomes a debate about whose data is right instead of what to do next.

  • The month-end scramble

    Someone loses the first week of every month rebuilding the same board deck by hand, from the same exports, in the same fragile workbook.

  • Systems that don't talk

    CRM, finance, operations and the warehouse each hold part of the picture. The only place they're ever joined is a spreadsheet.

  • Dashboards nobody opens

    You already bought Power BI. What got built is technically impressive and answers questions nobody in the business was asking.

  • One person holds it all

    Reporting works because of one spreadsheet that one person understands. That isn't a system — it's a risk with a vacation policy.

  • AI is happening anyway

    Your analysts are already pasting data into ChatGPT. Nobody has shown them how to do it safely, accurately, or in a way that compounds.

How we work

A short path to something real.

No six-month discovery phase. You should have working, useful output in weeks — and know exactly what it costs before we start.

  1. 1

    Discovery call

    Thirty minutes on your business, your systems and where the reporting hurts. We'll tell you honestly whether we're the right fit.

    You get: A straight answer

  2. 2

    Data health check

    A fixed-scope review of your sources, models and reporting. We map what exists, what's fragile and what's worth fixing first.

    You get: Findings, priorities, costed roadmap

  3. 3

    Build in increments

    Short cycles against the roadmap, starting with whatever removes the most manual work. You see progress every week.

    You get: Working pipelines, models, reports

  4. 4

    Handover & enablement

    Documentation, training and — if you want it — AI enablement for your team. The aim is that you need us less over time, not more.

    You get: A platform your team owns

The difference

We’d rather make your analysts twice as fast than make you twice as dependent.

Every consultancy now has an AI page. Most mean “let us build you something with AI in it.” We mean something more useful.

The people who know your business best are already on your payroll. What they don’t have is leverage — so the analysis that would move the business sits behind three days of manual SQL and data cleaning. Used well, modern AI tooling closes that gap.

This is not “AI will replace your analytics team.” It’s the opposite. We teach them how, on your data, with your tools — including where AI is confidently wrong and how to catch it.

See a wrong answer get caught ↑

Why Forward Metrics

You work with the people who do the work

No account manager, no bench, no handover to juniors. The person on your discovery call is the person writing the SQL.

We fix the cause, not the symptom

If the numbers behind the dashboard you asked for can't be trusted, we'll say so and fix that first.

Built for teams without a data department

Everything is designed to be run and extended by your one analyst or finance lead, not a platform team of eight.

Fixed scope before we start

A defined piece of work at a defined price. If scope changes, we talk before the work happens — not on the invoice.

We’re probably not for you if…

…you want the cheapest available Power BI contractor, a single dashboard with no interest in what’s underneath it, or a partner who won’t push back. We’d rather tell you that now than three weeks in.

Where to start

The Data Health Check

Most people don’t need a proposal. They need someone senior to look at what they’ve got and tell them the truth about it. It’s how nearly every relationship here begins.

Fixed scope · Fixed price · Typically 1–2 weeks

  • A map of every system your reporting depends on
  • Where your numbers diverge, and why
  • The manual work to automate first
  • An honest read on your existing Power BI
  • A prioritized roadmap with a cost for each stage
  • A recommendation you can act on without us

You own the output whether or not you work with us.

Book a discovery call →

What the report looks like

Data Health CheckExample
0
out of 100
  • Sources & pipelines64
  • Models & definitions41
  • Reporting55
  • Access & governance72
5 findings · 3 stages
  • HighRevenue is defined three ways across 11 reportsSales, Finance and Ops each filter credit notes and intercompany differently.
  • HighMonth-end pack is rebuilt by hand from 6 exportsAbout 38 hours of Finance time a month, and a single point of failure.
  • MediumPower BI refresh fails about twice a weekThe data gateway runs on one person's laptop.
  • Medium212 duplicate customer accountsNo owner for the customer master between CRM and ERP.
  • Low27 reports unopened in 90 daysRetire them; they slow refresh and confuse new starters.

Questions

Before you get in touch

We already have Power BI and it isn't working. Can you fix it rather than start again?

Usually, yes — and that's normally the cheaper answer. Failing Power BI estates tend to fail for one of three reasons: the model's logic doesn't match how your business defines things (your active customer, your revenue recognition, your financial calendar), the refresh is unreliable, or the reports were built around the data that happened to be available rather than the decisions people make. All three are fixable in place. We'll tell you if a rebuild is genuinely better, but it's not our default.

Do we need Microsoft Fabric?

Often not. Fabric is excellent and we build on it regularly, but plenty of mid-sized businesses are better served by a well-designed SQL warehouse and a properly built semantic model at a fraction of the cost. We'll recommend the smallest architecture that solves your problem and scales for the next few years.

Is our business the right fit?

If you have data in more than one system and someone spending real time each month assembling reports by hand, then yes. Most clients sit between 20 and 500 people, with one analyst or none. If you're smaller or not sure, ask anyway — sometimes the right answer is a focused week of work rather than a project.

Will you replace our analyst?

No. We make them considerably more effective. In most engagements your analyst is our closest collaborator — they know the business, we know the platform — and a large part of what we deliver is transferring capability to them, including how to use AI tools properly.

What does it cost?

Engagements are fixed-scope and quoted up front. The Data Health Check has its own fixed price; build work is quoted against the roadmap it produces. We'll give you a realistic range on the discovery call rather than making you wait for a document.

Is our data safe if we start using AI tools?

It is if you set it up correctly, and that's part of what we teach. There's a real difference between pasting customer records into a consumer chatbot and running AI-assisted analysis inside controlled boundaries. Our training covers what may go where, which tools suit which class of data, and how to verify output before anyone acts on it.

Are the examples on this page real clients?

No. Harbor & Pine Distribution is a made-up company and all the data is synthetic, so we can show the work in full without exposing anyone's numbers. The techniques — pipelines, semantic models, reconciliation checks, reviewed AI-assisted SQL — are the ones we use.