Can you just automate the report we already have?

Sometimes, but check it first. Most existing packs are two or three different reports stapled together, carrying metrics nobody has ever defined in writing. Automating that faithfully means automating the arguments too. We usually split it by reader and settle the definitions before anything is wired up.

What is the monthly fee actually for?

Watching it. Sources change without warning — a field gets renamed, an export format shifts, a new location opens and is not in the mapping. The monthly work is a scheduled check that every source is still answering and still means what it meant, plus the short written summary that goes at the top of each send.

Can it go to Slack instead of email?

Yes. Slack takes it through an incoming webhook, driven by Make or n8n, and Looker Studio and Power BI both send on a schedule of their own. What we recommend either way is a short summary in the message and a link to the full pack — people read three lines in a channel and they do not read an attached PDF.

One of our sources is a manual export. Does that kill it?

No, but it is the part that will break, and it is the part that makes the monthly fee worth paying. A human export that stops arriving produces a report that either fails loudly or silently reuses last week — we build it to fail loudly and to say so on the face of the report.

Will the numbers match our accounts?

Not exactly, and you should be suspicious of anyone who says they will. Analytics and ad platforms undercount because of ad blockers, consent refusals and people using several devices. They are reliable for direction and change. For the final money figure your payment system and your accounting are the truth, and we write the size and reason of the gap into the pack so nobody rediscovers it every quarter.

How many numbers should be on it?

Four to eight per reader. Past that, the pack stops being read and starts being filed. If you need twenty numbers, that is two or three reports for two or three different people, which is how we normally build it.

Data & Analytics

Automated KPI reporting

A weekly or monthly pack that builds itself, lands on time, and tells you when one of its own sources has quietly stopped being reliable.

From: $400–$4,000 + $150/mo
Typical time: 1–4 weeks
Usually bought by: Leadership teams

What this actually solves

Somebody in your company builds this report. Ask who, and ask how long it takes them. The answer is usually a name and a number of hours, every week, spent copying figures out of four systems into a spreadsheet that is already stale by the time it is opened. That person is not doing analysis. They are doing data entry against a deadline, and the deadline is exactly why nobody ever gets round to improving the report itself. Multiply their hours by twelve months and put it next to their salary — that is the real comparison, and it is usually not close.

The other version is more common and more expensive. The report simply does not get made. Leadership runs on the last figure somebody remembers, or on the one dashboard a single person knows how to open, and decisions get taken anyway — just without evidence, and without anyone noticing until the quarter closes. What you are buying is not a prettier chart. It is the guarantee that the same agreed numbers land in front of the same people at the same hour, whether or not anybody had time this week.

There is one thing an automated report does that a manual one never does, and it is the reason half of this work exists: it keeps arriving after it has stopped being true. A person building a pack by hand notices that a column is empty. A scheduled job does not. It delivers the same confident layout with a number that quietly went wrong six weeks ago.

What we do, and in what order

Decide who is actually reading it. A managing director's pack and a branch manager's pack are not the same document with different filters — they answer different questions and they trigger different actions. This step usually ends with two or three separate reports rather than one, and that decision is made before anything is built. Half a day to a day.
Write down what every number means, and get it signed off. This is the step people skip and then pay for. Does a new patient mean the appointment was booked or that it was attended? Does revenue include refunds, tax, shipping? Every ambiguous term is resolved in writing and approved by name, because the alternative is a meeting six months from now where two departments defend two numbers. Two to three days.
Connect the sources and build the delivery. Looker Studio for scheduled delivery, Google Sheets with Apps Script where the shaping is fiddly, Make or n8n to push into Slack, Power BI subscriptions where the business already lives there, and Supermetrics to pull advertising data. Format matters less than it seems — a short email with a link beats a large attachment almost every time. Three to five days.
Add the comparison, the alerts and the freshness line. No number ships alone. Each one carries the same period last month or last year and the difference, calculated on matched period lengths. Thresholds are set for the two or three conditions that should wake somebody up. And under every report goes a line stating what time the data runs to and whether every source answered. One to two days.
Run one live cycle, then hand it over. The report goes out for real once, with us watching, before anyone relies on it. The handover includes who to tell when something looks wrong and what the monthly check covers. One send cycle, a few hours of work.

The part clients read most is not automated at all. Two sentences written by a human at the top of each send — what changed, and what it probably means — is what turns a pack that gets filed into a pack that gets opened. We build the frame for it and it stays part of the monthly arrangement.

What moves the price

The range is $400–$4,000 + $150/mo, delivered over 1–4 weeks, and the spread comes from three places.

How many distinct reports. One pack for one reader is a short build. Three audiences with genuinely different questions is three definitions exercises, three layouts and three delivery schedules.

How many sources, and whether any of them are manual. An automated pull from a platform API is a fixed cost. A source that depends on somebody exporting a CSV every Monday is a permanent monthly obligation, and it should be priced as one rather than absorbed quietly.

Whether the written summary and the monthly check are included. They should be, and they should be named explicitly rather than buried, because that is the difference between a system that is maintained and a system that is merely still running.

What usually goes wrong

It is switched on and walked away from. Six weeks later a field gets renamed in the source system, or a new location opens and is not in the mapping, and the report keeps arriving on schedule with the same confident formatting and wrong figures. Nobody catches it, because arriving on time is what everyone was checking for. The remedy is unglamorous — a stated data-freshness and source-health line on every report, plus a diary check that takes a quarter of an hour a month. That check is what the monthly fee buys.

A number with nothing beside it. A single figure says nothing at all; it only becomes information when it is next to the previous period and the difference. The trap underneath that is period length. Open a monthly pack on the twelfth and it compares twelve days against a full thirty-one and shows a collapse that is not happening. February is nine percent shorter than January and looks like a bad month every single year. A retail month counted as five weeks beats a four-week one without anything improving at all. Comparisons run on matched day counts, and today's partial day is either excluded or marked incomplete.

It is built for everyone. A pack that tries to serve the board, the marketing lead and the operations manager at once ends up with twenty tiles, and twenty tiles is a document people scroll past. Four to eight numbers per reader, each one attached to a decision that reader actually makes, is what gets read on a Monday morning.

See where the site stands before you brief anyone.

If the site is already live, start from what it scores rather than from a guess. The audit reports what it checked, what it found, and — the part most tools leave out — what it could not check, and why.

See the website audit Talk about this work

Questions people ask before they commit.

Can you just automate the report we already have?: Sometimes, but check it first. Most existing packs are two or three different reports stapled together, carrying metrics nobody has ever defined in writing. Automating that faithfully means automating the arguments too. We usually split it by reader and settle the definitions before anything is wired up.
What is the monthly fee actually for?: Watching it. Sources change without warning — a field gets renamed, an export format shifts, a new location opens and is not in the mapping. The monthly work is a scheduled check that every source is still answering and still means what it meant, plus the short written summary that goes at the top of each send.
Can it go to Slack instead of email?: Yes. Slack takes it through an incoming webhook, driven by Make or n8n, and Looker Studio and Power BI both send on a schedule of their own. What we recommend either way is a short summary in the message and a link to the full pack — people read three lines in a channel and they do not read an attached PDF.
One of our sources is a manual export. Does that kill it?: No, but it is the part that will break, and it is the part that makes the monthly fee worth paying. A human export that stops arriving produces a report that either fails loudly or silently reuses last week — we build it to fail loudly and to say so on the face of the report.
Will the numbers match our accounts?: Not exactly, and you should be suspicious of anyone who says they will. Analytics and ad platforms undercount because of ad blockers, consent refusals and people using several devices. They are reliable for direction and change. For the final money figure your payment system and your accounting are the truth, and we write the size and reason of the gap into the pack so nobody rediscovers it every quarter.
How many numbers should be on it?: Four to eight per reader. Past that, the pack stops being read and starts being filed. If you need twenty numbers, that is two or three reports for two or three different people, which is how we normally build it.

Often taken together.

Start with the brief, not the invoice.

Send what you are trying to achieve. You get a written scope, a price and a date — or an honest “this is not the right service for that”.

Talk about Automated KPI reporting All Data & Analytics services