The Journey: #79 The Data Dilemma
The platform itself wasn't the problem. Plenty of companies run on a network of tools held together with duct tape and good intentions.
The problem was that this particular platform was disconnected from the rest of the business, and the numbers inside it were being updated manually by a team doing their best with what they had.
The platform had weak and limited integrations, there was no validation and no formal governance. Honestly, this is the challenge with a lot of CSPs when only one team is using them and the integrations never went past the bare minimum.
So when I pulled the data from it to build projections for the next two quarters, I was building on a foundation nobody had actually checked in a while. The inputs looked fine. They had the right columns, the right format, the right level of detail you'd want to see. What they didn't have was accuracy, because the manual updates were inconsistent, some fields hadn't been touched in months, and nobody owned making sure any of it still reflected reality.
But I didn't second guess it. I assumed it was good data, appropriate to use for my ELT reports. I'd been asked to build out retention, churn, and growth projections for H2, Q3 and Q4, to present at the next ELT meeting before our board meeting. The idea was simple enough, each leader covers their own story, reviews it with the group, and it makes its way into the board deck.
Beyond the data being wrong, our CFO let me go through my entire deck, every slide, every anecdote, before he said a word. Then he told me none of it matched their data. He started asking where I pulled the numbers from. Questioned my formulas and my math. Then he asked about the stories behind the specific customers I'd called out.
I was mortified. It was an honest mistake, I was only two months into the role. And in that moment I felt like the victim, because why hadn't anyone told me which dataset was the actual source of truth?
But the reality was, I never asked, and I never validated what I was using. I assumed that because it lived in our CSP, it had to be accurate.
That's the moment I stopped trusting data by default.
I built a framework to audit it before I ever used it again, is it correct, current, complete, and consistent. Then I didn't just start auditing my own data, I got my entire team governing it with me.
The Takeaway: Your model is only as good as the system feeding it, and a system nobody actively owns will quietly go stale, siloed, and wrong, long before anyone notices.
The Data Audit: Your 4 C's
That experience is the reason I don't just ask CS orgs if they have data. Everyone has data. I ask if they trust it. When I present, I'll ask a room to raise their hands if they trust their data, and I'm lucky if I get one or two hands. Everyone else just kind of looks at the floor, because they know. They know the dashboard is lying a little. They know someone updates that one field by hand every other Friday if they remember. They know the number in the exec deck and the number in the CRM don't actually match, and nobody's gone back to figure out why.
That hesitation, the one that stops you from raising your hand, is data. Trust me, it's worth listening to. So instead of a vague gut check, run your data through four questions. I call them the 4 C's.
Correct. Is the data actually accurate, or has it just been accurate for long enough that everyone stopped questioning it? Manual fields drift. Definitions change. A health score built two reorgs ago is still scoring people today like nothing moved.
Current. When was this last updated, and by what, a person or a system? If the honest answer involves someone's memory and a spreadsheet, you don't have current data. You have a snapshot of whenever they last had time.
Complete. What's missing that you don't even know is missing? Gaps don't announce themselves. They just quietly shrink the picture until you're making decisions off half the story and calling it the whole thing.
Consistent. Does this number match the same number somewhere else? If your CRM, your CS platform, and your finance system all tell a different story about the same account, you don't have three data points. You have three guesses, and you're the one who has to pick which guess to defend in front of your CFO.
Run every important number you rely on through those four, your health scores, your usage data, your revenue projections, your churn reasons. Wherever the answer to any of them makes you flinch, that's exactly where to start digging.
AI in CS
AI will not fix bad data. It will just help you find it faster, and honestly, that's the real value right now, not automation for its own sake.
The teams who are smart are using it in their data audits. Feed it your CRM export and your CS platform export for the same accounts and ask it to flag every mismatch. Point it at a health score model and ask what inputs haven't changed in six months, that's your current problem showing up in plain sight. Ask it to scan for fields that are populated in some records and blank in others, that's complete breaking down at the account level, not just the aggregate.
What AI is genuinely good at here is pattern spotting across volume no human wants to manually reconcile. What it cannot do is decide which system is telling the truth when two sources disagree, or judge whether a definition still makes sense for the business you are today. That part still needs a human who knows the business, asking why, and being willing to sit with an answer they don't like. Use AI to surface the gaps. Use your own judgment to decide what they mean.
Closing with Kristi
Nobody builds a CS org planning to run on bad data. It just happens … one manual update at a time, one siloed system at a time, until the gap between what you're reporting and what's true is wide enough for your CFO to catch it before you do. Go find your weakest C. And if this hit a little too close to home, that's exactly the kind of thing I help CS leaders fix. Whether it's an Advisory engagement to audit and rebuild your data governance, Coaching to help you navigate the fallout of a moment like mine, or Education & Enablement to get your whole team fluent in what "good data" actually means, I'd rather help you catch this before the board meeting than after.
See you next Tuesday,
Customer Success. Revenue Follows.
