Someone else's wing

In 1901, two bicycle makers from Dayton, Ohio, built a glider using the best aerodynamic data in the world. It produced about a third of the lift the tables said it would.

So what have the Wright brothers got to do with your CRM?

More than you’d think. Rather than keep flying and hoping, Wilbur and Orville went back to the workshop. They built a wooden wind tunnel about eight feet long, made balances out of hacksaw blades and bits of wire, and between September and December tested more than 200 model wings. By the end of the year they had the best wing data anywhere.

The twist? The published numbers they’d relied on, from the German glider pioneer Otto Lilienthal, weren’t really wrong. As NASA puts it, they were right for the wing Lilienthal had used. They just weren’t right for theirs.

Industry benchmarks are someone else’s wing.

The wind tunnel moved to the cloud

A century later I was marketing the modern version of that wind tunnel. Rescale gives engineers the computing power to run physics simulations in the cloud, so they can test far more designs, far faster. The pitch was simple - you can only innovate as fast as a model you trust.

Which made it a little awkward that marketing had no model at all.

When I joined, there wasn’t really a marketing strategy yet. There was no agreed way leads came in, who owned them, or how quickly anyone followed up. So I started from scratch.

The first thing we did wasn’t a campaign. It was getting leadership aligned on goals, budget, metrics and reporting. Only then did we go looking for quick wins. Within two years marketing’s share of new-logo pipeline went from 10% to over 65%, while revenue more than doubled. The campaigns mattered, of course. But they only worked because we could finally see what they did.

This is the layer most teams want to skip. It’s the least glamorous part of the engine and the one everything else sits on. So I’ve broken it into three parts: standardise, operate, and only then automate.

1. Standardise: one dictionary

If marketing, sales and customer success each have their own definition of a qualified lead, you don’t have one funnel. You have three, and a weekly argument about whose is right.

Standardising means one set of stages and definitions, one owner for each, and one place they live. It sounds like admin (and, to be fair, some of it is). But the definitions decide what you count, and what you count decides what you fund.

It also has to reflect how people actually buy. Forrester found an average of 13 people involved in a B2B buying decision, with 89% of purchases crossing two or more departments. Closer to home, a 2024 study of APAC buyers put it at 12.8 people per enterprise deal. (The groups aren’t getting any smaller, but more of the shortlisting now happens in an LLM before they ever meet - G2 found half of software buyers now start their research in an AI chatbot. That’s a story for another essay, but it’s one reason AEO and GEO should be a top priority for anyone in marketing right now.)

A data model built around single leads will tell you that thirteen people from one company are thirteen weak signals. They’re one strong one.

And then there’s the quality of what goes in. When Tadhg Nagle, Thomas Redman and David Sammon had managers check their own recent records, they found 47% of newly created records had at least one critical error, and only 3% of the data quality scores reached an acceptable level. That’s not old data decaying in a corner. That’s this week’s data, wrong on arrival.

For me, the definitions are a leadership job, not an ops ticket. If the CMO and the head of sales haven’t signed the dictionary, nobody will use it.

2. Operate: don’t drop what you catch

Clean definitions are worthless if the plumbing leaks.

In 2011, Harvard Business Review published an audit of 2,241 US companies that tested how fast they responded to a web enquiry. Only 37% replied within an hour. 23% never replied at all. Firms that made contact within an hour were nearly seven times as likely to qualify the lead as those that waited even an hour longer.

That matters more now than it did then. Buyers turn up later and more decided. 6sense’s 2025 research found buyers make first contact about 61% of the way through their journey, they reach out themselves more than 80% of the time, and the vendor they favoured before that first conversation wins around 80% of the time. You get fewer signals, and each one is worth more. Dropping one is expensive.

Operating the foundation means routing in minutes, a follow-up SLA that sales and marketing both own, and a plan that knows how many conversations the sales team can actually have. I’ll come back to capacity in its own essay, because it’s where a lot of good demand quietly dies.

3. Automate: garbage in, garbage out at machine speed

So why has this gone from important to urgent?

AI is now doing the work that used to sit with people: scoring accounts, routing leads, writing follow-ups, summarising pipeline for the board. And more of the buying journey happens without a person at all - Gartner found 61% of B2B buyers would rather buy without a rep.

But an agent does exactly what the data tells it. If close to half your new records carry an error, it will act on them confidently, at scale, all day. Garbage in, garbage out has always been true. AI just runs it faster.

We already know where this leads. When Gartner asked sales leaders and sellers about forecasting, only 45% had high confidence in their forecast accuracy, and poor data quality was the reason given. Bolt AI onto that and you don’t get a better forecast. You get a faster wrong one.

I think the “AI dividend” will go to the teams with boring, standard data first. Not the ones with the most tools.

What good looks like

A few golden rules I use to check whether the foundation is in place:

• One dictionary. Every stage has a written definition, an owner and an entry rule, signed off by marketing and sales.
• Buying groups, not just leads. The data model can see several people from one account as one opportunity.
• Routing in minutes. High-intent enquiries reach a named person fast, and someone checks that it happens.
• Capacity in the plan. Demand targets are set against what sales can actually follow up.
• A named owner for data quality. Not a project. A job.
• Reports that match. The board deck, the CRM and finance agree without anyone reconciling them by hand.

For me, yes and no

There’s a fair challenge to all of this. You can’t wait for perfect data before you do anything. The critics who say growth is a flywheel, not a staircase, have a point - standing still while you clean the CRM costs more than acting on imperfect numbers.

I agree, up to a point. The Wrights didn’t wait for perfect data either. They flew at Kitty Hawk in 1903 on numbers that were good enough, not complete. At Rescale we went after quick wins while we were still fixing the plumbing.

So the bar isn’t clean data. It’s data good enough to make a decision you’d defend. Below that line, every layer you build on top is guesswork with a dashboard.

Where to start on Monday

1. Pick one definition and get it signed. Start with what counts as a qualified opportunity. Get marketing and sales leadership to agree it in writing.
2. Test your own speed. Fill in your own contact form and time how long it takes for a real person to respond.
3. Run one reconciliation. Take last quarter’s pipeline number from three places - the CRM, the board deck and finance - and see if they match. If they don’t, you’ve found your first project.

How do you know your numbers are trustworthy? I’d love to hear the tests other people use.

Next in the series: once you trust the numbers, you can start testing what moves them. That’s predictability - and it looks a lot more like the scientific method than a spreadsheet.