Why B2B attribution breaks, specifically
Every company says attribution is hard. In B2B it fails for four reasons that are structural rather than technical, and knowing which one you have determines whether the fix is a tool, a process, or an argument nobody wants to have.
The cycle outlasts the cookie. A B2B purchase takes nine to eighteen months. Browser storage lasts a fraction of that, first-party cookies are capped at seven days in Safari, and people change devices twice in a year. Any model that depends on the browser remembering the first visit has already lost. What survives is identity resolution at the CRM, which means capturing source at the moment someone gives you an email address and never relying on the browser to hold it afterwards. Where that capture happens on the website, it has to be built in rather than retrofitted.
The buyer is a committee, and only one of them fills in the form. Six or eight people touch the decision. One converts. Attribution credits that one, so the content that actually moved the technical evaluator or reassured the security reviewer looks worthless in every report. This is the most common reason genuinely effective content gets cut.
The last touch is usually a conversation. A referral, a conference, a Slack message between two people who used to work together. It is unmeasurable, it is frequently decisive, and any model that assigns 100% of credit to whatever digital touch happened nearest the close is describing something that did not occur.
Nobody agrees what a lead is. Marketing counts form fills, sales counts conversations they would take, finance counts closed revenue. All three are correct within their own frame, and the argument that follows is usually treated as a data problem when it is a definitions problem.
The practical position we take: stop trying to build a model that assigns exact credit. Build one that answers narrower questions reliably, which channels produce opportunities that close, which content appears in the path of deals that win, and what the cost of an opportunity is by source. Those are answerable. Perfect multi-touch attribution in a committee-driven eighteen-month sale is not, and pursuing it burns budget that should go to the offline conversion work below.
Offline conversions, the highest-leverage single fix
If a company has ad spend and a CRM and we can only do one thing, it is this.
Ad platforms optimise toward whatever you tell them counts. Almost every B2B account tells them a form submission counts. Form submissions include students, competitors, job seekers, and people who will never buy, and Smart Bidding notices they are cheap and plentiful and buys more of them. Cost per conversion falls, the dashboard improves, and pipeline does not move.
The fix is mechanical. When a lead reaches a qualified stage in the CRM, that outcome is sent back to Google and LinkedIn as the conversion event, with the click identifier captured at form submission so the platform can match it to the original click. Bidding then optimises toward the leads your sales team accepts.
The parts that go wrong in practice: the click identifier is not captured or not stored on the record, so nothing can be matched. The conversion window is shorter than the sales cycle, so qualified leads arrive after the platform stops listening. Or the CRM stage that triggers the import is one sales does not actually maintain, so the signal is noise. All three are fixable and all three are invisible until someone checks.
The data model decides everything downstream
The least glamorous work and the one that determines whether any of the rest survives.
Properties multiply because creating a new field is easier than finding the existing one. Three fields end up meaning company size. Two mean lead source, populated by different systems with different values. Picklists accumulate options nobody uses. Within two years reports disagree, and the disagreement is blamed on the reporting tool.
Fixing it is not a migration, it is a set of decisions: which system is authoritative for each field, what the permitted values are, what populates it and when, and what happens to the historical records that do not conform. Those are business decisions that need someone senior to make, which is why this work stalls in companies that treat it as an IT task.
For HubSpot specifically, this is the first phase of any implementation. We do this before building automation, not after. Workflows written against fields that later change is how a portal ends up with dozens of active automations nobody will turn off because nobody knows what depends on what.
Consent, and what still works after it
Increasingly the constraint, and frequently ignored until a legal review forces it.
Consent mode, cookie banners, and regional privacy rules mean a meaningful share of visitors are never measured client-side. Depending on your audience and geography that can be a third or more. Teams respond either by ignoring it, which makes every report quietly wrong, or by concluding measurement is impossible, which is worse.
What actually works: server-side tracking for the events that matter, so measurement does not depend entirely on a browser executing a script; modelling the gap rather than pretending it does not exist; and first-party data collection at the points where someone is giving you information anyway. The goal is not to evade consent. It is to build measurement that behaves predictably when a large share of traffic declines it, which is now the normal case rather than the edge case.
Where we are not the right fit
We are not a data warehouse or BI practice. We will get marketing and sales data into a state where it can be trusted and connected, and if you need dimensional modelling, transformation pipelines, and a semantic layer across the whole business, that is a data engineering team.
We do not do enterprise Salesforce development. Apex, Lightning components, and multi-org architecture are their own discipline with their own specialists. We integrate with Salesforce and we work inside it; we are not a Salesforce dev shop.
We are not privacy counsel. We implement consent mechanics and we do not advise on what your obligations are.
And if the honest problem is that nobody has decided what the marketing strategy is, better instrumentation will produce clearer pictures of an unclear plan. That is worth saying before the invoice rather than after.