A familiar situation
The Numbers Don't Agree
The CRM, ad platforms, and reporting each tell a different story.
When marketing systems do not share definitions or data properly, meetings turn into number debates and important handovers depend on manual work.
Same month, three different answers
Open the same month in the ad platform, the CRM, and the board report. The numbers do not match.
Usually nobody is simply wrong. Each system is counting a different event, using a different date, or missing a handover somewhere in the flow.
The meeting becomes a debate about whose number is real instead of a conversation about what to do next.
The cost nobody calculates
Someone exports the ad data, reformats the columns, joins it to the CRM, and updates a spreadsheet before the weekly meeting. The process works because one person remembers how to do it.
That’s the visible cost. The invisible one is worse.
When getting a report requires pulling from four systems and hoping the numbers align, people stop checking. Decisions get made on gut feel. Leads slip through because the handoff between tools is a manual step someone forgot on a busy Friday.
The most expensive integration is the one you’re doing manually every week without realizing it’s an integration problem.
What good actually looks like
Good does not mean every tool contains everything in real time. It means the team knows which system owns each piece of data, how the definitions map, and when the information moves.
The important numbers can be traced back to a source. The expected delays are known. Errors surface somewhere a person can see them. A manual step can be fine when it is deliberate, owned, and documented.
That is how the team gets one usable answer in the pipeline review instead of three unexplained ones.
The patterns I see most
The CRM-to-marketing connection is the classic one. Contacts sync, but lifecycle stages do not. The fields technically map, but the definitions do not. Or a delay nobody planned for makes lead routing unreliable.
Ad platforms to CRM is the second. LinkedIn, Google, Meta — they all generate leads, but attribution dies somewhere between the click and the closed deal. Fixing this single connection often changes how a company thinks about channel spend.
Then there’s the website itself. Form submissions, page views, product usage events — all of it should flow to the systems that need it. Most of the time, it flows to Google Analytics and stops there.
I use whatever gets the job done: native integrations when they fit, automation tools for simpler flows, and custom API work when the standard options fall short. I do not push a particular stack.
How I work on these
I do not follow a numbered playbook. But there is a shape to it.
I start by mapping what exists. What talks to what, what’s manual, where data gets stuck or corrupted. This usually surfaces the real problem, which is rarely what the client thought it was going in.
Then I design the target state and build toward it one connection at a time. Incremental, tested, documented. Because the worst thing you can do with integrations is build five at once and then try to figure out which one broke your data.
Agents are useful here for comparing definitions, checking exports, tracing edge cases, and documenting the flow while it changes. I still decide which discrepancy matters, what is safe to automate, and where a human step is the better answer.
Documentation matters here more than anywhere else. Six months from now, someone will need to understand why lead source maps this particular way. If it’s not written down, it might as well not exist.
If the problem is still unclear, discovery maps the systems, definitions, ownership, and breaks in the flow. If the work needs ongoing ownership, I can join embedded to build and operate what comes next. If the numbers keep winning the meeting, send me the situation.