Skip to content
The Growth Spire
Client panel Book a call

Count. Verify. Decide.

Your company seen in numbers every day: a management report computed live from the sources, a forecast checked against the past, and one decision every week that follows from those numbers.

For e-commerce companies with PLN 5 to 50 million in revenue.

Book a call See what the report looks like

The first call is free and ends with a list of things to check in your own business.

Figures from a deployment

By this hour the report already knows yesterday
03:30
Monthly forecast error in the backtest
4.41%
Markets and currencies in one profit and loss account
6
Of sales recovered after a single payment fix
PLN 0.95m

Figures from a 2026 deployment at an e-commerce client, unnamed until the case study is approved. Each one has a source in the report that recomputes it every night.

What it looks like

Five numbers with thresholds and three alerts, instead of two hundred cards.

In the morning you know what strayed from the plan, what it costs and who should check it. The report lives at your own address, behind a login, and computes itself from the store, ads, payments and the bank.

What we build

Four things, and each has to deliver a number, not a presentation.

Management report

Net sales, margin by channel, companies, payments. Computed live from the source, available at your own address and behind a login.

A current view for the board

Data warehouse

Store, ads, costs and stock in one place, with one definition of every metric. No more arguing about whose spreadsheet is right.

One source of truth

Sales forecast

A model trained on your seasonality and checked against closed months. A change goes in only when it improves the historical error.

Model error verified

Internal applications

Panels and tools cut for one process: order flow, work with creators, margin calculators. Where a spreadsheet stopped being enough.

Tools for one process
Platforms that run the business

Not just a report about the company. Also the system it runs on.

A report says what happened. But customer service, bookings, contracts and payments happen elsewhere, usually in several tools at once that do not talk to each other. We build platforms that tie them together: an app for the end customer and a panel for the team, on one set of data.

An app for your customer

Web and phone from one codebase. The customer books, pays, signs and contacts you without email and without a phone call. From the browser or the app, whichever is more convenient.

CRM: from enquiry to contract

Sales funnel, contact history, offers and renewals in one place. The team sees who needs a reaction today, not who spoke last.

Bookings, access, contracts

Resource calendar, building entry from the phone, contracts signed electronically and invoices issued automatically. Less paper and fewer questions at reception.

The same data in the report

If the system runs the sales, it is the source of the numbers. Occupancy, revenue and churn land in the management report without retyping and without definitions drifting apart.

The first platform of this kind is being deployed in a coworking space right now

A tenant app together with a sales panel. It is built by the same team that counts the data, so the tool and the report speak one language from day one, not after a year of integration.

Deployment in progress
How we work with AI

AI does the work that used to eat weeks. Decisions stay with people.

We do not sell “AI solutions”. We use it every day as a working tool: to write and check queries, comb through data for anomalies, build models and prototypes that once took weeks. That is why an analysis that would be a quarter-long project in the classic mode takes a few days.

Architecture

Building the warehouse and queries

Data models, transformations and queries come together faster because AI writes the first version and we verify it on real data. That shortens implementation from months to weeks.

Instead of weeks: 2-3 days
Anomaly detection

Looking for what nobody looked for

Instead of checking ten hypotheses, we check hundreds: anomalies in payments, markets, channels and cohorts. The machine combs, a person judges what is real.

Hundreds of vectors, continuously
Forecasting

Models and forecasts

Model variants are built and backtested in series, not one by one. Only the one that really reduces the error on closed months goes in.

A hard historical test
Fast delivery

Applications in days, not quarters

You get a panel or an internal tool as a working prototype right away, not as a mock-up to approve. Fixes go in on the live screen.

Working code from day 1
Where AI does not decide: every number that reaches the report is checked against its source.

The model may propose a hypothesis, but it does not settle the budget, the discount or the forecast. A person who signs off on them is responsible for that.

What usually surfaces

The money is usually where the report is not looking.

The kinds of things we find in the first weeks. Not because someone is at fault, but because a standard report shows sales, not what happens to them afterwards.

Profit bridge · margin, PLN thousand (sample)Profit audit

Sales grew by 12.4%. Margin in PLN only by 6.0%. This chart shows where the growth was lost along the way.

The axis starts at PLN 1,900 thousand, not at zero, so the changes are visible.
  • adds to the result
  • takes from the result
  • closing balance
The board asks:Are we growing or just turning over bigger numbers?

Turnover grows faster than the result: sales +12.4%, margin in PLN +6.0%. Every zloty of sales brings less today than a year ago.

Marketing asks:Has advertising stopped paying back?

Not only that. Volume and price earned 263 thousand, while discounts and media gave back 123 thousand, almost half.

Finance asks:Where exactly does the zloty leak?

Margin fell from 24.5% to 23.1%. Each bar is a specific line in the books, so the conversation is about numbers, not impressions.

Orders counted, but never paid

A payment method quietly stops settling. Sales in the report grow, money does not come in, and it only shows when reconciled with the bank statement.

A discount deeper than anyone assumed

A price promotion and a discount code stack and multiply instead of adding up. The real markdown is often a dozen points higher than the policy.

A comparison base that understates growth

A year earlier the data was incomplete or counted with a different key. Year on year looks great, but it compares against a hole, not reality.

A channel living off past glory

Revenue attributed to a partnership from half a year ago looks like a current return on budget. Once separated, it turns out the new spend does not earn.

A return booked in a different month than the sale

Sales count on the order day, returns and cancellations a month later. Every month looks better than it was, and the correction lands where nobody looks for it.

A forecast nobody checked against the past

A model fitted to the last quarter hits the past and misses the future. Without a test on closed months, the annual plan is a wish, not a number.

How we work

From diagnosis to a decision rhythm, in this order.

Every stage ends with a number that can be checked and a decision whether to go on. None of it requires a hire on your side.

  1. Read-only access, a consistency audit against the bank, one definition of the metrics. Ends with a list of money to recover and a decision whether to build further.

  2. Warehouse, a report at your own address, a backtested forecast, five numbers with thresholds and three alerts. Refreshes every night, without asking anyone for a summary.

  3. Forecast calibration, an anomaly scan of the last weeks, a brief with three decisions to make. Once a month a plan-versus-actual review with you.

  4. More markets, channels and companies in the same profit and loss account. Internal applications where a spreadsheet stopped being enough.

Questions before the start

What owners ask before they show their data.

Your question is not here? Write to me, I usually reply within one working day.

Ask a question
How is this different from a Power BI or Looker Studio dashboard?

A dashboard shows numbers someone loaded into it. The management report computes them from the sources every night, reconciles them with the bank statement, keeps one definition of every metric and tells you what crossed a threshold. On top of that: a backtested forecast and a person who signs off on a decision every week.

How much does it cost?

Three levels: Diagnosis billed once, Report as a monthly subscription, and Staff, which is the Report plus weekly work on decisions. The amount depends on scale, the number of markets and sources, so I quote it after the first call. The Diagnosis is credited towards the Report implementation.

What data do I need to share?

Read-only access: the store, ad platforms, analytics, mailing, payments and the bank statement. We sign a data processing agreement before the first access. Data stays in the European Union and never reaches the code contractor.

How long does implementation take?

The Diagnosis takes two to three weeks. The first live report appears three to four weeks after the decision to implement. The weekly rhythm starts on the day of the first report.

Does AI make the decisions here?

No. AI writes queries, combs through data for anomalies and builds model variants. Every number that reaches the report is checked against its source, and the budget, the discount and the forecast are approved by a person who signs off on them.

Which stores and systems does it work with?

Any store with an order export or an API: WooCommerce, Shopify, IdoSell, Shoper, BaseLinker and Allegro. Ads from Google, Meta and TikTok, analytics from GA4, payments from PayU, Przelewy24 or Stripe, plus the bank statement and an accounting export.

Rafał Bartoszewicz in a conversation at a laptop with a client's team
Who is behind itRafał BartoszewiczFounder & Lead Intelligence Operator

Day to day a fractional marketing director for companies that have outgrown their agency but not yet grown into a full-time CMO. The Growth Spire is the execution workshop of that work: this is where the tools, reports and applications that decisions rest on are built.

One person is responsible both for the numbers and for what follows from them. There is no handing the topic over between an analytics department and strategy.

1:1 advisory at bartoszewicz.co Fractional CMO & Strategy
Direct contact

Let us start with a question you cannot answer today.

Write what you want to know about your company that the current report does not show. I will reply whether it can be counted and what is needed for that.

Where do we start? Pick oneThe topic goes into the invitation
  • I reply personally, usually within one working day.
  • No data to send. For the first call, what you already see in your own report is enough.
  • You leave with a list of 3-5 things to check, whether or not we work together afterwards.
Write directlyrafal@bartoszewicz.co Book an intro call

The booking page is in Polish; the call can be in English.