Business Analytics Without a Data Team: Playbook

Small teams answer cross-system questions today with exports, a shared spreadsheet, and one person who knows where everything lives. A durable path connects the systems read-only, resolves customers once, and returns answers with source rows, so the knowledge stops living in one head.

TL;DR

How do small teams get answers today?

A 40-person company runs a CRM, an accounting ledger, a payment processor, a support desk, a project or inventory tool, and somewhere between five and fifteen other systems. Nobody on staff is an analyst. When the CEO asks which customers are unprofitable, the process is roughly this.

Someone exports three reports to CSV. Someone pastes them into a shared spreadsheet. Someone matches company names by eye, notices that Acme Corp and Acme Corporation are the same account, and merges them by hand. Someone corrects two figures that looked wrong, without pushing the correction back to the source. The answer arrives two days later, and it is usually close enough to act on.

That is not incompetence. It is a rational response to systems that were never connected, and the numbers disagree for structural reasons covered in why your business systems do not agree on the numbers.

Why does the spreadsheet approach break?

It breaks on repetition, not on accuracy.

The matching work is thrown away. Next month the same names are matched by eye again from scratch.
The corrections live only in the sheet, so the source systems stay wrong and the fix has to be remembered.
Nothing is traceable. When a figure is challenged, nobody can name the rows behind it.
The sheet accumulates logic in formulas that only its author can read.
Each new question means a new export and a new tab rather than a follow-up.

The result is a reporting function whose cost per answer never falls. Month twelve takes as long as month one, and by then the file has 14 tabs.

What is the risk of one person holding all the context?

Every small company has this person. They know that the revenue report has to exclude the internal orders, that one region's data lags a day, and that the December figures were restated. None of it is documented.

Most of what they produce is a one-off answer rather than a report, which is why how to answer business questions without building a dashboard is the closest description of the work.

Three things follow. Answers stop when they are away. Onboarding a replacement takes months because the knowledge was never written down. And nobody can audit their work, which means an honest mistake can persist for a year unnoticed.

The goal is not to replace that person. It is to move what they know out of their head and into a system that any leader can query directly.

What does a better path look like?

Four steps, none of which require a hire or a warehouse project.

Connect the systems read-only, including the spreadsheets, so nothing is excluded because exporting it is inconvenient
Resolve entities once, so one customer, product or vendor is recognized across every system and reused by every later question
Name each metric definition, so booked revenue and recognized revenue are separate labelled measures
Require citations, so every figure names the system, table and rows it came from

SIGNLD, a decision intelligence platform by Inzata Analytics, connects to 800+ business systems and runs entity resolution automatically on connect, with no schema to author first. The homepage shows the question box a CEO, CFO or COO types into, and connectors lists the accounting, CRM, payment, support and spreadsheet sources most of these questions run through.

Connections are read-only, data is encrypted with AES-256 at rest and TLS 1.2 or higher in transit, inference runs on a single-tenant AWS Bedrock instance, a private LLM powered by AWS Bedrock, your data stays yours, and we never train on it.

Which three questions should you start with?

Do not try to model the business. Start with the questions you already ask every month.

Most small teams pick some version of these three: which customers or jobs are unprofitable after cost of delivery, why cash collection is lagging revenue, and which accounts are at risk based on activity across support and billing. Each one spans at least two systems, which is precisely why they take two days today.

In construction those questions have a concrete shape, worked through in connecting Procore and QuickBooks for construction analytics.

Ask them, open the cited rows, and check the answers against what your spreadsheet said. Agreement builds trust. Disagreement is more useful, because the citation shows exactly which record explains the difference.

What does this cost in time and attention?

Connecting the first system takes about 15 minutes and the first answer comes back in minutes. There is no modelling phase, so the work is measured in a first afternoon rather than a first quarter.

The ongoing cost is review rather than construction. Entity matches are worth checking, especially parent and subsidiary groupings, because how to group a corporate family is a business decision rather than a technical one. Pricing is listed at /pricing, and it is worth comparing against the fully loaded cost of a hire covered in do you need to hire a data analyst.

Key takeaways

In most companies under 100 people the reporting function is one spreadsheet and one person who knows which export to pull.
That arrangement works until the person is on holiday, changes role, or the spreadsheet grows past anyone's ability to audit it.
The expensive part of every question is not the analysis, it is matching customers and reconciling definitions across systems.
Doing the entity matching once and storing it is what stops each month restarting from an empty sheet.
Start with the three questions you ask every month rather than trying to model the whole business.

FAQ

Can a company do analytics without a data team?

Yes, for the class of question whose answer already sits in connected records. Cross-system questions about what happened and why are answerable once entities are resolved and definitions are named. Building new statistical models, forecasts and experiment designs is different work that still needs a trained analyst.

Is a spreadsheet good enough for a 40-person company?

It gets answers, and for a single month it is often the fastest route. It breaks on repetition, because the customer matching and the manual corrections are discarded each cycle and nothing is traceable when a figure is challenged. Connecting the spreadsheet as a source keeps the corrections visible instead of losing them.

What if our data is messy?

Messy data is the normal starting condition, and duplicate customer records are the most common form of it. Entity resolution is designed for exactly that case: it matches records by names, domains, addresses, identifiers and transaction patterns rather than requiring a clean shared key. Cleaning the source systems first is not a prerequisite.

Who in a small company owns this?

Usually the person who already answers the questions, often a finance lead, an operations manager or the founder. No engineering skill is required to connect a system or ask a question. What helps is business context, because deciding which revenue definition matters is a judgment about the company, not about the data.

Do we need to replace any of our systems?

No. Connections are read-only and nothing is migrated or written back, so your CRM, ledger, payment processor and support desk keep operating exactly as they do now. The joined view lives in your private Knowledge Graph, and the source systems remain the systems of record.

Try SIGNLD free

Pick the question your spreadsheet answers every month and see how long it takes connected. Try SIGNLD free or browse all articles.

How small teams get answers without a data team