Three teams, three revenue figures, and a meeting about which one is right.
Most analytics problems are not visualisation problems. They are definition problems — nobody agreed what counts as a customer, an active user or a closed deal. We fix that first, then build reporting that arrives on its own and dashboards built around a decision instead of a data source.
✓ One agreed definition per metric✓ Reports that arrive without being asked✓ Every number traceable to its source✓ Built on the tools you already pay for
Tell us the three numbers your leadership argues about. That’s usually the whole brief.
In short: we get your operational data into one place, agree what each number means, and put it where the decision gets made — a scheduled report in an inbox, a live figure in the tool people already use, a dashboard someone actually opens. AI comes into it for the parts that genuinely benefit: writing the commentary, flagging what moved, forecasting what’s next. Not for making the chart prettier.
Why the last dashboard went unopened
Almost every business we work with already has one. It was built with enthusiasm, used for a fortnight, and now lives in a bookmark nobody clicks. Three reasons, in roughly this order.
Nobody trusted the numbers. Someone spotted a figure that disagreed with their own spreadsheet, couldn’t find out why, and quietly went back to the spreadsheet. One unexplained discrepancy is enough to kill a dashboard permanently, which is why traceability matters more than presentation.
It answered no question anyone was asking. It was built from what the database contained rather than from a decision someone makes on a Monday. Data-source-shaped dashboards are always complete and rarely useful.
It required remembering to look. Anything that depends on a busy person developing a new habit loses to the thing that arrives in their inbox. Push beats pull, nearly always.
The definitions conversation comes first
It is the least exciting part of an analytics project and the part that determines whether the rest of it survives.
Take “customer”. Is a person who bought once, three years ago, still one? Does a trial count? Is a company with four accounts one customer or four? Sales, finance and support each answer differently, each is right within their own world, and each reports a different total to the same board.
We run that conversation for every metric that matters, write the answers down, and build the calculation from the written version. It usually takes a couple of sessions, it is occasionally uncomfortable, and it is the single highest-return thing in the project. Afterwards, the argument in the meeting is about what to do rather than about whose number is correct.
Three places where it earns its keep, and one where it is mostly decoration.
Writing the commentary
“Revenue is down 6% on last month, driven almost entirely by two accounts that renewed late; volume elsewhere is flat.” Somebody used to write that paragraph every month. It is a good use of a model, because the numbers are already computed and the model is only describing them.
Noticing what moved
Anomaly detection across dozens of metrics that no person has time to watch. The value is in the alert, not the dashboard — you find out that one region’s refund rate tripled without anyone having gone looking.
Asking in plain language
“How many new customers in the northeast last quarter?” typed rather than requested from an analyst. Genuinely useful, with one condition: it must answer using the agreed definitions and show the query it ran, or you have simply automated the production of disputed numbers.
Where it doesn’t help
Choosing chart types, generating “insights” nobody asked for, and summarising a summary. If a feature makes the screen busier without changing what anyone does, we leave it out.
What we won’t do
Move you onto a new platform to make our life easier
If you already pay for a business intelligence tool, we build in it. Migrations should happen because the current tool genuinely can’t do the job, not because a supplier prefers something else.
Build a forty-tile executive dashboard
Forty tiles means nobody decided what mattered. We’d rather deliver six numbers that change behaviour and leave the rest reachable one click deeper.
Report a number we can’t trace
Every figure has to be clickable back to the rows that produced it. If we can’t build that path, the number doesn’t go on the page — because the first time someone disputes it, the whole thing loses credibility.
How the work runs
Week 1 — which decisions, and who makes them
We work backwards from decisions rather than forwards from tables. Every metric that survives has to be attached to something someone does differently depending on its value.
Week 2 — the definitions, written down and signed off
One page per metric: what counts, what doesn’t, which system is authoritative, and how it’s calculated. This document outlives the project and is the thing new staff should be handed.
Weeks 3–5 — pipelines, then the first report
Data flowing on a schedule with failure alerts, then one scheduled report rather than a dashboard — because a report proves the numbers are right while a dashboard only proves they exist.
Week 6 onward — dashboards for the roles that need one
Built per role, once the underlying figures have been trusted for a few weeks. Then we check usage after a month and remove whatever nobody opened, which is always something.
What our clients say
The businesses we’ve built for, in their own words.
★★★★★
“When I approached Abedin Tech with my land share selling plan, I wasn’t sure how it would work. But thanks to their precise strategy and powerful marketing, my business is now thriving. They truly understand their clients’ needs and go above and beyond.”
Owner, Richland Properties
Real Estate
★★★★★
“I approached Abedin Tech to develop my website with several specific functionalities. Their team delivered exactly what I envisioned, creating a beautifully designed website that met all my requirements. I highly recommend Abedin Tech.”
Rohit
Owner, Shop from China
★★★★★
“The decision to partner with Abedin Tech was the best decision we made. Our site looks great, our traffic is through the roof, and our sales are better than ever. Abedin Tech is the perfect digital partner that offers what is beyond your expectations!”
James Anderson
★★★★★
“We had an idea but no sense of direction. With each step of the way, Abedin Tech guided us and turned our vision into a beautiful website with functionality. The outcome is evident by the numbers!”
Isabella Scott
Name the three numbers your team argues about
Plus which systems they come from and who needs to see them. That’s enough for us to say whether this is a definitions problem, a plumbing problem or a reporting problem — and they need very different budgets.
Usually one of three reasons. Someone found a figure that disagreed with their own spreadsheet, could not find out why, and went back to the spreadsheet — one unexplained discrepancy kills a dashboard permanently. Or it was built from what the database contained rather than from a decision someone actually makes. Or it required a busy person to remember to log in, and anything that arrives in an inbox beats anything that waits to be visited.
Why do you start with metric definitions rather than building?
Because most analytics disputes are definition disputes wearing a technical costume. Whether a customer who bought once three years ago still counts, whether a trial counts, whether a company with four accounts is one customer — sales, finance and support each answer differently and each reports a different total. We write one page per metric covering what counts, what does not, which system is authoritative and how it is calculated, then build from that document.
Do we need to move to a new analytics platform?
Usually not. If you already pay for a business intelligence tool we build inside it. A migration should happen because the current tool genuinely cannot do the job, not because a supplier finds something else easier to work with. The pipelines and definitions are the valuable part and they are portable.
Where does AI genuinely help with analytics?
Three places: writing the commentary that explains what moved and why, detecting anomalies across metrics nobody has time to watch, and answering plain-language questions — provided it uses the agreed definitions and shows the query it ran. It does not help by choosing chart types, generating insights nobody asked for, or summarising summaries. If a feature makes the screen busier without changing what anyone does, we leave it out.
Is a scheduled report better than a dashboard?
For most clients, yes, and we build one first. A report that arrives with the commentary written and the exceptions flagged requires no habit change and proves the numbers are correct. A dashboard only proves the numbers exist. Dashboards come afterwards, built per role, once the underlying figures have been trusted for a few weeks.
How long does an analytics project take?
Typically six weeks to something genuinely useful: a week working backwards from decisions, a week agreeing and documenting definitions, three weeks building scheduled pipelines and the first automated report, then dashboards for the roles that need them. We check usage a month later and remove whatever nobody opened — there is always something.
Related work
What analytics projects tend to lead into once the numbers are trusted.
