Data & Integration
Three systems, three answers to the same question, and a meeting that stalls while everyone defends their own spreadsheet. The data is not missing. It is contradicting itself.
Book your Diagnostic ↗The information exists. Nobody agrees on which copy is true.
Data work at Sell with Marketing means getting information out of the system that is holding it hostage, cleaning and reconciling the sources that contradict each other, connecting the systems that should have been talking, and leaving one definition of each number that the whole company works from.
This is the least glamorous line of work we sell and it is the one that quietly blocks the others. A catalogue that cannot be exported blocks the ecommerce build. Customer records with four spellings of the same company block the CRM. A price list that lives in someone's files blocks the quoting engine. Every one of those looks like a technology problem and every one of them is a data problem.
The typical trigger is a system change: a new ERP, a new CRM, a store, software that will not survive another year. The information has to come across, and the vendor being left rarely makes that easy.
Get it out, clean it up, keep it connected.
- Extraction from the system that is holding it: Legacy ERPs, software whose vendor has vanished, databases nobody has the password to, scanned documents and files in formats that stopped being current a decade ago. Where an export exists we use it; where it does not, we build one.
- Cleaning, deduplication and normalisation: One company recorded four ways becomes one record. Units, currencies, codes and categories end up following a single rule, which is the part that makes everything downstream possible.
- Reconciliation between sources that disagree: When the ERP, the CRM and the spreadsheet give three answers, the work is deciding which is authoritative for each field and writing that decision down so the argument does not come back next quarter.
- Integration so it stays connected: A one-off migration decays from the day it finishes. What keeps it alive is the synchronisation between systems, with the direction of truth defined per field.
- Migration with verification, not just transfer: Record counts, totals and spot checks before and after, documented. A migration nobody verified is a migration whose errors surface in six months, in front of a customer.
- The data dictionary: What each field means, where it comes from, who owns it and how it is calculated. This is what stops the same discussion happening again the next time two reports disagree.
When changing systems is blocked by getting the data out.
- Companies changing ERP or CRM with years of history in the old system
- Manufacturers whose catalogue lives in files no current software can read
- Businesses where two reports answer the same question differently
- Teams re-keying the same data into three systems by hand every week
- Companies whose software vendor disappeared and left the data behind
- Operations that cannot start an ecommerce or CRM project until the catalogue is usable
This work is usually the first phase of a bigger project rather than a purchase on its own. It is separated out and priced on its own because it is real work with real hours, and hiding it inside another quote is how a project ends up late and over budget.
Before you sign.
Our old system has no export. Is the data lost?
Almost never, though the route depends on the case: a supported export where one exists, direct database access where the credentials can be recovered, or reading the interface systematically when neither is available. What we do before quoting is confirm which of those applies, because a migration quoted without knowing how the data comes out is a quote for a problem nobody has looked at.
How do we know nothing was lost in the migration?
Because it is verified rather than assumed: record counts, financial totals and sampled records compared before and after, and the comparison handed to you as a document. Ask any supplier for this before signing. A migration nobody verified is one whose gaps appear months later, usually in front of a customer.
Can we clean the data ourselves and save the cost?
The part that is genuinely yours is the decisions: which company name is the real one, which price list is current, which customer is still a customer. Nobody outside can make those calls. What we bring is the extraction, the deduplication rules, the reconciliation across sources and the verification, which is where the hours and the risk actually sit. Teams that try to do all of it in a spreadsheet usually discover the scale of it after the project has already started.
How long does this take?
It scales with volume and with how bad the state is, and the second one matters more than the first. The Diagnostic samples your real data before quoting a timeline, because the difference between a clean catalogue and one with fifteen years of accumulated exceptions is measured in weeks, not in percentages.
What happens after the migration?
Without integration, the copies start drifting apart again from day one, which is why the synchronisation and the data dictionary are part of the scope rather than an upsell. What is proposed is the connection that keeps the systems in step and the written definition of each number, so the next disagreement has somewhere to be settled.
ONE NUMBER.
ONE SOURCE.
It starts with the Diagnostic: what standing still is costing you, what fixing it costs, and what gets fixed first — in your own data. Paid work, credited 100% against the fix if you move within 90 days.
Book your Diagnostic ↗