In depth

How Smartin calculates, measures and checks numbers.

For those who want to see how it works inside: what the Monday letter and the daily Overview look like, how we measure every change, what exactly we watch on the customer journey, why AI answers reliably through Smartin and how we measured it.

This is what a Monday with us looks like

One thing a week, and why it's that one.

Every Monday you get a short letter of about 200 words. At the top is one thing for this week: why it's that one, who on your side will do it and when we'll know it worked. Below it is one request for you and how last time's agreement turned out. At the end there is a tip for something worth verifying: you start it by sending one sentence to your AI, and if you'd rather not, you don't need to do anything. Everything else is in the Overview, the daily page with the whole customer journey. Sample Overview

Sample company · Monday letter

Hello Tomas,

Most important this weekThere are too few new leads: 9 came in during August and your target for the end of September is 25. Since September you have been reaching out to partners, but nobody can see how many there were or what came of it. This week's step: log every partner contacted in the CRM, a name and a date are enough. Petr will do it, he's copied in. In two weeks, on 12 October, we will check how many partners are logged and how many leads they brought in. You will see the progress every day at the top of the Overview.

One request for youPlease confirm that the target of 25 new leads also applies to October. Just reply "yes", or send a new number.

How last time's agreement turned outLast time we agreed that you would go through the deals that were stuck. Of the 20, 9 now have a next step, 6 are marked as lost and 5 are waiting for an update from the sales team. We don't need anything from you on this.

What has changed since last timeIt has turned out that you haven't spoken with four out of five customers in three months.

The Smartin team

Everything else, every day and in detail, is in the Overview of your customer journey.

A tip for youToday 16 of your 40 open deals have no recent note (40%). If you want to find out whether adding a short note to the CRM after every call helps, Smartin will measure it on its own for six weeks; the target is below 25% by 9 November. Just write to your Claude: "Start a verification: if we add a note after every call, the share of open deals without a recent note will drop below 25% by 9 November." If you'd rather not take this on, that's fine.

A sample for a fictional company, in the form clients receive the letter.

In 6 weeks you know whether a change worked.

We measure every change we agree on. Every day the system goes through your customer journey, finds where you are losing money, calculates what it costs, and every week suggests up to 3 changes based on how much money is at stake. At your meeting the consultant picks with you the one to start on (how the system finds them). You can also enter a change yourself, for example through your AI. Before you even start, we lock the baseline number and agree where it should move, by when and who on your side is responsible, so the result is decided by a number we agreed on in advance. Meanwhile the consultant works with your team to make sure the change really happens. In 6 weeks we show you from the data whether it worked, and then for another 90 days we watch whether the result holds.

You know where to put your money.What didn't work, you have in black and white, and you don't try it twice. And because we remember what worked at similar companies, we suggest that first.

You can show the owners what each change brought.Every number has a baseline, a date and a source, so you can defend the result in front of the board.

What you are trying and whether it worked

Running (2)

Partners contacted bring more leads running

Baseline: 9 new leads in August, locked on 28 Sep
Target: 25 new leads a month
Responsible: Petr
We evaluate: 9 Nov

A next step on every open deal running

responsible Petr · we evaluate 9 Nov

Closed (2)

Alerts on deals without activity worked

responsible Lucie

Discount for an annual subscription didn't work

responsible Petr

Martin Souček, your consultant, goes through every verification with you at your meetings.

Daily Overview

Every red dot has a person, a decision and a date.

The letter tells you one thing for the week. Every working day in the morning, the Overview shows you the whole customer journey from today's data. At the top is the most important thing and who does what about it today and tomorrow. Below it you'll find what we are waiting for from you, the statements for leadership and every point of the journey with a number and a period. The system prepares it, and every letter links to it.

The sample is for the same fictional company as the Monday letter.

we have the data and it fitswe have the data, something isn't going to plandeal with it this weekwe don't know yet, we say why

SOverview for Sample company (B2B software)

OverviewCompassVerification

As of 28 Sep

Overview of your customer journey

Target: 25 new leads by the end of September · Now: log partners contacted in the CRM

From the Monday letter, 28 Sep

There are too few new leads. 9 came in during August, the target for the end of September is 25.

Today14 of 40 open deals are stuck. Step: agree on each what happens next. Who: Petr
Tomorrow5 deals are waiting for an update from the sales team. Step: remind the sales team. Who: Petr

P.S. Please confirm that the target of 25 new leads also applies to October. Just reply "yes", or send a new number.

The Smartin team

What you are waiting for from us

What you asked Smartin for and don't have yet.

we reply by 2 OctSplit leads by source: website, partners, referrals.

What we are waiting for from you

What we need from you to continue.

1 figure1 to close

by 2 OctAre the partners contacted logged in the CRM? Just reply "yes".

Fourteen statements for leadershipholds 6 · doesn't hold 1 · don't know 3 · not measured yet 4

Revenue and plan

Revenue is growing to plan

not measured yet. We are missing invoicing data.

We will hit this quarter's revenue plan

don't know. We are missing the revenue target only leadership knows.

We sell what we forecast

holds. The expected close is right for seven deals out of ten.

Our open deals are enough to hit the plan

don't know. We are missing the revenue target only leadership knows.

Sales and inquiries

New inquiries aren't dropping

We win what we start

holds. Over the last 12 months you won 24% of deals, the year before 22%.

We close deals as fast as before

holds. You close three quarters of won deals within 64 days, last year within 71.

New customers are the ones we target

not measured yet. We are missing industry and company size.

We aren't relying more and more on discounts

holds. The average discount is 6%, last year 7%.

Customers

Customers don't leave us

don't know. You haven't talked to four out of five customers in three months. We will see churn only from invoicing.

Revenue doesn't depend on a few customers

holds. The three largest customers brought 22% of the value of won deals over the last 12 months.

Sales doesn't depend on one salesperson

holds. The strongest salesperson brought 38% of the value of won deals.

Customers pay on time

not measured yet. We are missing invoicing data.

A new customer pays for itself in time

not measured yet. We are missing invoicing data.

New inquiries aren't dropping

Doesn't hold: In the last 90 days 21 new inquiries came in, 34 in the same period last year.
What we are doing about it: Decided with leadership on 28 Sep: log every partner contacted in the CRM and rewrite the website copy for companies with 40+ people.
Who: Petr
When we evaluate it: 9 Nov

Customer journey

Weakest point: new leads are dropping.

Interest1 issue
Inquiryfine
Negotiation1 issue
Signingfine
Launchfine
Usageunknown
Retentionunknown

Your Claude or ChatGPT has everything above at hand. Try asking: "Which leadership statements don't hold, and why?"

Your consultant: Martin Souček. With you for the whole program.

What we watch for you

The whole customer journey, from interest to retention.

Every day we compare what is happening with your customers against the plan and the company's rules. When something isn't working, you hear about it on Monday, before you feel it in revenue.

Interestinterest that doesn't lead to an inquiry
Inquiryan inquiry with no reply or next step
Negotiationdeals stuck for more than 30 days
Signingwin rate dropping, losses with no reason
Launcha new customer with no set process
Usagea customer nobody has talked to in three months
Retentioncustomers leaving, not expanding, not renewing

We tested whether Claude can work out company numbers on its own.

We built an artificial company and planted traps that real data really contains, such as an internal job logged as a deal, or a group of companies that looks like three customers in the CRM. In the first round the company was small; in the second it was the size of a real B2B company: 2,570 deals, 1,680 companies, about 4,500 invoices and credit notes and 990 support tickets. Then Claude got 7 questions a leadership team asks, once with direct access to the data and once through Smartin. We worked out the correct answers in advance with a program, straight from the raw data.

Through Smartin, Claude's answers cost roughly 200 times less and arrived about 20 times faster.Connected directly to the data, in the second round it read 1.3 to 2.2 million tokens for the set of 7 questions and took 4 to 5 minutes. Through Smartin the same set took about 6,700 tokens and 12 seconds, roughly 200 times fewer tokens, which is what you pay for with AI. The numbers were already calculated and Claude only read them.

When Smartin doesn't have a number, Claude says it doesn't know.Through Smartin, Claude did not make up a single number. For questions Smartin doesn't measure yet, it said it didn't know and explained why.

Without agreed definitions, Claude gets company data wrong and doesn't notice.In the first round (5 runs of 7 questions), 18 of 35 answers working directly on the data were wrong, and Claude never once said it didn't know. It counted an internal job as a deal and a group of companies as three customers. In the second round, on a company of real size, it confidently got 4 of 6 scored questions wrong (we didn't count the seventh because of a mistake in the artificial company).

Agreed definitions decide it.When Claude got the definitions written out completely and precisely, it calculated correctly without us too. We ran the test on 25 October 2026 with Claude Sonnet 5.5 on an artificial company with traps based on real data, and we will repeat it after every new model.

Your AI with Smartin

At Smartin, AI does whatever AI can already do. It is kept in check by code that calculates every number the same way every time, and a person works alongside it. When you ask your own AI about the company, through Smartin it answers roughly 200 times cheaper and about 20 times faster than with direct access to the data, because the numbers are already calculated.

Why your AI answers differently every time

Your AI→web · CRM · marketing · customer care

Each tool sees only its own part of the customer journey.
AI connects them anew every time, and differently every time.
It doesn't know where an inquiry ends and a deal begins in your company, so it guesses.
The result can't be repeated or checked.

Why it always gives the same answer

Your AI→Smartin→web · CRM · marketing · customer care

Between your AI and your data sits our own system. In one place it holds the customer journey, Compass, the go-to-market plan, the playbooks, the area map and the decision memory. Everything is connected and every rule has the date of its last check.
We connect the whole journey from the website to retention once, following your rules.
Each step has a clear definition and is measured the same way every time.
Every day, code recalculates where the journey isn't working. Every sentence and every number has a source and a date.
AI reads the finished state of the journey that the system calculated.
The same data always gives the same result.

Ask your AI

You connect Claude, ChatGPT or any other AI that speaks MCP to the Smartin MCP. Then you talk to your customer journey whenever you like: at your desk, before a meeting or in the car.

What you ask

"What did we decide about pricing?"
"What is overdue?"
"Which leadership statements don't hold, and why?"
"How is my idea doing in Compass?"
"Prepare me for the meeting with the sales team."

What you tell it

"Record that we decided to raise prices." It is saved as a proposal and takes effect once we confirm it.
"Add an idea for a new segment to Compass."
"Our revenue target for this quarter is eight million."
"We've completed this point on the area map."
"Tell the Smartin team to split our inquiries by source."

Any question the data can't answer yet, your AI sends to our team as a message. You learn straight away who is handling it and by when we will reply, and then you find the answer in the Overview and your AI tells you too.

The numbers are calculated before you ask. AI only puts the answer together from them.

How Smartin finds what to improve on its own.

The system also measures every change, and the consultant is there the whole time.

Tracking.Every day, code reads the customer journey from the CRM and other sources: from interest and inquiries through negotiation and signing to launch, usage and retention.
Signal.It finds where the company is losing money, such as deals with no next step, customers who pay less or nothing, or losses with no reason. Code calculates the count and the amount.
Suggestion.From a catalogue of proven interventions it picks what to do and every week suggests up to 3, based on how much money is at stake. It takes into account what came up in meetings, and the consultant picks with you which one to start on.
Verification.Each suggestion becomes a verification with a baseline and a deadline in 6 weeks. Code measures the number before and after, and a person confirms the verdict.
Learning.The result is recorded and the system suggests better next time. The credit goes to your team.

We don't trust anything until we've checked it.

Every sentence and every number can be traced.For everything Smartin writes in the letter and in the Overview, you can see where it comes from: which data and which day. You, your team and your auditor can check it back at any time. What isn't in the data, we don't fill in with a guess. We say we don't know, and what it would cost to find out.

We test what leadership claims.What holds up becomes a rule. What is disproved, we write down, so the company doesn't keep coming back to it.

We keep everything up to date.Every rule has the date of its last check. When it stops holding, we fix it.

How we measure results

Every claim about impact belongs to one of three levels and has a source.

Level	Example
1. What was built	written processes for sales, marketing and customer care; new deal stages in the CRM; a new offer; hiring a salesperson in three rounds
2. What changed in how people work	leadership decides every two weeks, on three things at most; marketing moved its money to where real prospects come from
3. What moved in the numbers	in 10 weeks the share of deals stuck for more than 3 weeks dropped from over half to a third, mainly by closing dead deals (one company); we measure every further change within 6 weeks

No integration, and your team doesn't have to enter anything by hand.

Smartin doesn't need any particular system. It reads what your company already has, and if you have no data, the changes you try are enough: we record and measure their results together. It calculates every number the same way, whoever it suits, so it is impartial.

Read access is enough.Your CRM, analytics and shared files. With read access we recalculate every day. Without access, an Excel file, a CSV or even a photo of the whiteboard in your office will do, and we calculate as often as you send the files.

Bad data is a management problem.When people don't log things in the CRM or everyone logs them differently, leadership doesn't know what's going on with customers. We find it and see you through the change: we write the playbooks with your team, what gets logged, by whom and how, and then check every week that people work by it and that it works.

30 minutes on what is hurting you today.

Book 30 minutes