Big Data LDN filled Olympia London on 23 and 24 September with more than 400 speakers, 16 theatres and over 200 exhibitors. Walk the floor for ten minutes and you notice something: almost every stand promises the same thing. "AI-powered." "One platform for all your data." "Insights in seconds."
If you run a small business, that is confusing. How do you choose between products that describe themselves in identical words? That question, more than any single announcement, is what this post is about.
•What you'll get from this post
A simple way to understand what different data tools actually do, a side-by-side look at the best-known platforms, and five questions you can ask any vendor to cut through the marketing. No technical background needed.
•Follow one small business's data
Let's make this concrete. Oak & Ember (a made-up name) is a six-person online homeware shop in London. It sells through Shopify, advertises on Instagram and Facebook through Meta Ads, does its accounts in Xero and keeps everything else in Google Sheets. Like most small businesses, it has five data problems, and each one is solved by a different type of data tool.
Problem 1 — "Our numbers are spread across five different exports." The fix is one place where all of it lives together. Oak & Ember loads every export into a single store once, then asks one question across all of it: "which Instagram campaign brought customers who ordered twice?" You ask in SQL (a simple language for talking to databases) or through an AI assistant that writes the SQL for you. MotherDuck is built for small and medium data, simple to set up and free to start; BigQuery is Google's version, handy if you already use Google's tools; Snowflake and Databricks are full platforms built for large organisations with data teams; Databricks in particular is strong in data engineering and machine learning. With a few gigabytes of data, MotherDuck or BigQuery fits — the enterprise options would be overkill.
Problem 2 — "Someone spends every Monday morning downloading exports." The fix is exports that download themselves. A connector logs into Shopify, Xero and Meta Ads on a schedule, pulls the new data and drops it into the store from Problem 1. Nothing is analysed here; the job is only moving data reliably. Fivetran offers hundreds of ready-made connectors with a free plan; Airbyte is the open-source alternative technical teams can run themselves. For Oak & Ember, Fivetran's free plan covers it.
Problem 3 — "Our checkout broke and we only found out from an angry customer." The fix is something watching the site around the clock and raising an alarm before a customer does. Datadog is the best-known monitoring tool, built for engineering teams. But Oak & Ember's site runs on Shopify, which monitors its own servers — so Datadog would be money spent on a problem someone else already handles. It only becomes relevant if the business builds its own website or app.
Problem 4 — "I want yesterday's sales and ad spend on one screen every morning." The fix is a dashboard that reads from the store in Problem 1 and refreshes itself. Dashboard tools are built to display what the store already has, not to replace it. Google Data Studio (formerly Looker Studio) is free and connects to Google Sheets, Analytics and Ads; Power BI suits teams living in Excel and Microsoft 365; Hex targets analysts who want code and AI in one place. For Oak & Ember, Google Data Studio is enough.
Problem 5 — "Shopify says we made £42,000 last month. Xero says £38,500." The fix is to agree once on what each number means and write it down. The gap is usually refunds, VAT or shipping counted differently in each system. Atlan is a data catalogue that records what each number means and where it comes from; DQLabs monitors data quality and flags broken data before it reaches a report. For a six-person business, one shared page defining "revenue", "customer" and "repeat customer" solves most of it; Atlan and DQLabs can wait.
•The best-known tools, side by side
This is why names that sound alike on a conference stand are often completely different purchases. Most of these platforms now do more than one job; what matters is the job each is built around and priced for, because that is what you will actually pay for. Put the best-known ones next to each other:
Datadog. Solves Problem 3. Watches your website and servers and alerts engineers when something breaks. Priced by servers monitored and log volume. A company running its own software needs it; Oak & Ember does not.
MotherDuck. Solves Problem 1 for small and medium data. Holds your orders, invoices, ad spend and customers, and lets you ask questions of them. Built on DuckDB, a free open-source engine; charges for storage and compute used. Fits Oak & Ember.
Snowflake. Solves the same Problem 1 for large organisations and has grown into a broader platform with its own AI and application features: hundreds of users, terabytes of data, strict governance. Priced on compute time, which is where bills grow fast. Overkill for a six-person shop.
Databricks. Also solves Problem 1 as a full data platform; its strength is data engineering and machine learning at scale, which is why it is built for teams with data scientists and engineers. Not a fit for Oak & Ember today.
BigQuery. Google's answer to Problem 1. Convenient if the business already lives in Google Ads, Analytics and Sheets. Priced per query, which is cheap at small scale.
Fivetran. Solves Problem 2 only. Moves data, stores nothing, analyses nothing. Often mistaken for a database because it sits on every data diagram.
Power BI, Google Data Studio, Hex. Solve Problem 4. They are built to display data, not to be your main store, even though each can hold some data of its own.
Atlan, DQLabs. Solve Problem 5. They document what numbers mean and catch bad data. Valuable once there are several teams; a shared page does the job before that.
Most of these say "AI-powered" on their website. Only the job each is built around tells you whether you need it.
•Five questions to ask before you choose
This is the part to keep. Whether you are at a stand, on a sales call or reading a pricing page, these questions separate products that sound identical.
1. What job does it actually do? Ask the vendor to describe it in one sentence without the words "AI", "platform" or "insights". If they cannot, be cautious. Then place it in one of the jobs above: storage, moving data, monitoring, dashboards or quality.
2. Is it built for your size? Ask how much data their typical customer has. If your whole business lives in a few spreadsheets and exports, a tool designed for petabytes will be expensive and overcomplicated. Lighter tools like DuckDB (free) and MotherDuck (free starter tier) exist exactly for this.
3. How does the price grow? Products charge very differently: per user, per hour of compute, per server, per gigabyte. Ask what your bill looks like if your data doubles or your team grows. The cheapest starting price is not always the cheapest tool a year later.
4. Can you leave easily? Ask whether your data is stored in an open format and how you export it. Tools built on open-source foundations, like MotherDuck on DuckDB, make it easier to move later. Being locked in is a cost you only notice when you try to leave.
5. Will it work with what you already use, and can they prove it? Check it connects to your real systems: Shopify, Xero, Google Sheets, your CRM. Then ask to see a demo using data like yours, not a perfect sample. Useful follow-ups: how much setup sits behind this demo, and where does a person still need to step in?
•What to try next
Write down your three most important questions about your business — for example, which marketing actually brings repeat customers? Then list where the data to answer them lives today. That one page tells you which job you need a tool for, and protects you from paying for tools built for companies a hundred times your size.
“This is how we approach every project at KMCP Solutions: understand the job first, pick tools that fit the business rather than the brochure, and add AI only where it clearly earns its place. If your data lives in five spreadsheets and a few exports today, that is not a problem. It is a starting point.
KEKerem Ege PaktenFounder & CEO, KMCP Solutions
