What B2B data unification actually involves, and what it fixes

Marketing data unification is the work of getting your CRM, advertising platforms, website analytics and partner feeds to describe the same events the same way, in one place you control. It is mostly not a technology problem. Roughly a third of it is engineering and two thirds is getting people to agree on definitions they have been quietly disagreeing about for years. Teams that treat it as a tooling purchase buy a platform and end up with a fourth number that disagrees with the other three.

Why do your numbers not match?

Because each system is answering a slightly different question, correctly.

Your advertising platform counts a conversion within its own attribution window, using its own definition, attributed to itself. Your analytics counts a session that reached a thank you page. Your CRM counts a record that a salesperson accepted. Your partner counts a registration passed through a portal.

None of these is wrong. They are four different measurements of four different things, presented in four dashboards with similar labels. So the Monday meeting becomes an argument about whose number is right, which is unwinnable, because all of them are.

There is a second layer underneath. Run three advertising platforms and you will routinely find they claim more conversions between them than you actually had, because each credits itself for a buyer all three touched. Nobody is lying. Each one is scoring its own contribution with a ruler it made itself, and you are the only person holding all three rulers.

Unification does not make the systems agree. It creates one model, above them, with one set of definitions, that everybody argues from instead.

What does the work actually involve?

Five stages. The order matters, and skipping the first is why most attempts fail.

Stage one: settle the definitions. Before anything is connected, you need written agreement on what a lead is, what makes it qualified, what counts as an opportunity, and who decides. This is a negotiation between marketing and sales, not a data exercise, and it is where the real value is created. A single agreed definition of a sales ready lead, signed off by the person who owns the number, is worth more than any platform you could buy.

Stage two: audit the sources. List every system holding customer or campaign data, who owns it, what it is authoritative for, and how it identifies a person and a company. Most businesses find between eight and fifteen. Most also find at least two that nobody has looked at in a year and one that a single person maintains by hand.

Stage three: resolve identity. This is the genuinely hard engineering. The same person appears as a website visitor, a form fill, a CRM contact, a webinar registration and an email subscriber, under three spellings and two email addresses. In B2B you also need company level resolution, because the buying group matters more than any individual in it, and a lead from a subsidiary needs to roll up to the parent.

Stage four: build the model. One place where the resolved data lands, with the agreed definitions applied. What that place is matters far less than the fact that it is yours: a warehouse, a properly configured analytics setup, or a hosted database are all workable. What matters is that when you stop paying a vendor, the history stays.

Stage five: reporting people will actually read. A model nobody looks at is an expensive filing cabinet. The output has to answer the question a board asks, which is why the number moved, not merely that it did.

How long does it take, and what does it cost?

Longer than the vendor demo suggests, and less than an enterprise integrator quotes.

Realistically, for a mid sized B2B business with a normal amount of mess: the definitions work takes two to four weeks of meetings, and it will feel slow because it is genuinely contested. The source audit takes a week. Identity resolution and the model take six to twelve weeks depending on how many systems and how bad the historical data is. Reporting takes a further two to four.

So a quarter, roughly, before you have something trustworthy. Anyone promising a live single view in a fortnight is either connecting two systems and calling it unification, or has not looked at your data.

The costs that surprise people are not the licences. They are the time of the person who knows why the CRM is set up the way it is, and the cleanup of historical records that nobody wants to own.

Why does owning the infrastructure matter?

Because the log is the asset, and rented logs disappear.

On the HMS Networks programme, measurement runs on infrastructure HMS owns, using Matomo alongside their CRM, rather than reporting held inside the platforms buying the media. Three things follow from that.

The first is that there is one number. Website sales ready leads, confirmed CRM leads and media cost sit in one model with one set of definitions, which is why it is possible to state plainly that the programme delivered 1,588 website sales ready leads and 201 confirmed CRM leads, and that cost per qualified lead fell from around £760 through trade events to around £72. Those two figures come from different systems. They only sit in the same sentence because the definitions were settled first.

The second is that the history compounds. Two years of clean records tells you which early signals in your business actually preceded revenue. A competitor can buy the same platform tomorrow. They cannot buy your two years.

The third is commercial leverage. When your measurement is inside a vendor's platform, switching costs include your entire history, and the vendor knows it.

The programme returned 35:1 on media. That number is only defensible because everyone involved agreed in advance what was being counted.

What does good look like once it is done?

Four practical tests. If your setup passes all four, the work is finished.

One number, one definition. Marketing, sales and finance quote the same figure for the same thing without checking with each other first.

The report explains why, not just what. When pipeline drops 15%, the reporting shows which channel, which segment and which week, without anyone opening a spreadsheet.

It survives a supplier change. Switch an advertising platform and your historical record is intact.

It answers the board's question directly. What is marketing returning, and how confident are we. If your model cannot produce that in one view, it is not finished, however clean the data is.

Most businesses reach the first two and stop, which is a reasonable place to stop. The third and fourth are what turn a data project into an asset.

Where to start

Do not start with a platform evaluation. Start with a definition.

Get marketing and sales in a room and write down, in one sentence each, what a lead is and what makes it qualified. If the two answers differ, you have found the reason your numbers do not match, and you have found it for free.

Then list every system holding customer data and mark which one is authoritative for each field. Most teams cannot complete that grid on the first attempt, and the gaps are your scope.

Our B2B data unification work starts with exactly those two exercises, and our marketing data integration programmes build the model on infrastructure you own. If what you actually need is the view at the end of it, that is unified marketing dashboards.

Frequently asked questions

What is marketing data unification? Marketing data unification is the process of bringing CRM, advertising platform, website analytics and partner data into one model with one agreed set of definitions, held somewhere you control. The aim is a single number that marketing, sales and finance all recognise, rather than several dashboards that disagree.

Why do my marketing platforms report different numbers? Because each measures a different event with its own definition and attribution window, and each credits itself for buyers that several platforms touched. Run three advertising platforms and they will often claim more conversions between them than you actually had. Unification does not make them agree; it creates one model above them that everybody uses instead.

How long does a data unification project take? For a mid sized B2B business, expect around a quarter. Two to four weeks settling definitions, a week auditing sources, six to twelve weeks on identity resolution and building the model, and two to four weeks on reporting. Timelines that promise a live single view in a fortnight usually mean connecting two systems rather than unifying all of them.

Do I need a customer data platform? Not necessarily. A warehouse, a properly configured analytics setup or a hosted database will all hold the model. What matters more than the category of tool is that the infrastructure is yours, so the history survives when you change supplier, and that the definitions were agreed before anything was connected.

What is the hardest part of data unification? Agreeing definitions. Identity resolution is the harder engineering, but the definitions work is where projects stall, because it requires marketing and sales to settle a disagreement they have been avoiding. Teams that treat unification as a purchase rather than a negotiation tend to end up with an additional dashboard that disagrees with the others.