The 27-Point Gap Between Attribution and Reality
Freshworks CMO Kady Srinivasan on the 27-point gap between attribution models and real pipeline impact — and what AI success should actually measure.
Topics
Kady Srinivasan has spent her career sitting at the uncomfortable edge of marketing and measurement. At You.com, she coined the term “forward-deployed marketers” to describe a role that didn’t yet have a name, helping define an entirely new category from scratch. Today, as Chief Marketing Officer at Freshworks, she is tackling a different kind of challenge: convincing a market that already understands IT and customer service software to rethink what AI allows that category to become.
In this conversation, Srinivasan is candid about one of the thornier problems in modern marketing — the gap between what attribution dashboards claim and what actually moves a business. She recounts pausing a channel for six weeks, watching an attribution model predict a 31% pipeline loss, only to see the actual impact come in at just 4%. That 27-point gap, she says, is less a data error than a warning about mistaking proxies for outcomes.
She also discusses Freshworks‘ partnership with McLaren, the always-on demands AI support now has to meet, and why she’s sceptical of AI “productivity wins” that hollow out an organisation’s ability to teach its own people.
Full interview;
You’ve written that pausing a channel for six weeks reduced the actual pipeline by just 4%, a 27-point gap relative to your attribution model. What did those six weeks feel like internally? Was anyone pushing to turn it back on?
It was actually an important learning moment for all of us. The internal reaction wasn’t simply, “Turn it back on.” The attribution model indicated we were giving up 31% of the pipeline, whereas the actual pipeline impact was only 4%. That’s a pretty significant gap.
Of course, there was pressure to turn the channel back on. That’s natural when a dashboard tells you you’re losing something. But I wanted us to pause and understand why the two numbers were so different.
For me, that’s one of the biggest challenges in modern marketing. We have incredibly sophisticated measurement systems, but sophistication doesn’t necessarily mean accuracy. If a dashboard shows something is working but the business doesn’t experience the impact, the answer isn’t to automatically spend more. Sometimes the lesson is that your measurement model needs to change.
At You.com, you coined the idea of “forward-deployed marketers” for a category that was still being defined. At Freshworks, AI is transforming a category everyone already understands. Which is harder, creating a category or reinventing one?
I actually think they’re hard in completely different ways.
Creating a category is probably harder intellectually, because you’re starting with a blank sheet. At You.com, when we talked about “forward-deployed marketers,” we were essentially trying to give a name and a framework to something that wasn’t yet fully defined. You’re asking the market to learn a new language.
At Freshworks, AI is entering categories that people already understand. So the problem isn’t “Do people understand this category?” They do, and they’re already experts in their field. The question is, “Do they understand what this category can now become?”
That’s a different challenge. You’re asking customers to change their expectations and, sometimes, how they work. That means the product, the customer experience, the proof points and the marketing all have to reinforce the same change.
If you had to guess, what percentage of marketing dashboards are currently telling a story that isn’t true?
I wouldn’t put a precise percentage on it, because the problem isn’t that the data is wrong. It’s that we often measure proxies for business impact and treat them as the impact itself.
A dashboard can tell you whether a campaign generated engagement, leads, or pipeline attribution. That doesn’t necessarily tell you whether it changed the business’s trajectory.
The best marketers I know are increasingly asking a simpler question: What actually changed for the business because of this campaign?
Freshworks partnered with McLaren to connect operational efficiency with performance. What can marketers learn from a partnership where the brand promise is demonstrated through the customer’s actual experience, rather than just through sponsorship visibility?
In my opinion, the strongest brand partnerships don’t need to be explained; they can be demonstrated.
McLaren is a performance-driven organisation where technology has to work at the highest level, under extreme pressure and with very little room for error. Freshworks’ presence in that operational environment makes the partnership more than a visibility exercise.
For marketers, that’s the distinction between putting your brand next to a story and having your product actually be part of the story. The latter creates credibility because customers can see the promise being lived out, not simply advertised.
Freshworks found that 47% of IT tickets now arrive outside standard business hours. Does that change the case for always-on AI support, and where should companies draw the line between automation and expecting technology to compensate for a broken work model?
It strengthens the case for always-on support, but it also exposes a bigger question: What are we actually asking technology to solve?
If employees work across time zones and the business doesn’t stop at 6 p.m., IT support shouldn’t necessarily stop either. AI can resolve routine issues instantly and give people access to help when humans aren’t available.
But AI shouldn’t become an excuse for a broken work model. The goal should be to remove friction from work, not make people permanently available. Automation should handle predictable, repetitive work so humans can focus on the decisions and problems that genuinely require them.
If AI is changing workflows and decision-making, not just automating tasks, are companies making the mistake of redesigning technology without redesigning how people work?
Yes, and I think this is one of the biggest risks with AI. It’s very easy to add AI to a workflow without first asking whether it still makes sense. If an AI agent is now doing something that previously required three people, who owns the outcome? If an AI is making a recommendation, who is accountable for the decision?
Those aren’t technology questions. They’re organisational questions.
So I think the companies that win with AI will be the ones willing to redesign the work itself. Don’t just take the existing process and make it 20% faster by using AI and automation.
Ask whether AI allows you to create an entirely different process — there has been no better time for marketing to reimagine the way it operates.
If AI makes a senior employee 1.4 times more productive today but leaves the organisation less capable tomorrow, is that really an AI success?
Not necessarily. Productivity is only one side of the equation. A 1.4-times productivity improvement sounds great, but I’d ask, “At what cost?”
What happens to the junior employees who were supposed to learn by doing that work? What happens when the senior person leaves? Are we actually building organisational knowledge, or are we making one individual incredibly effective while making everyone else more dependent on them?
The goal shouldn’t simply be to make today’s employee faster. It should be to make the organisation more capable over time.
The real measure of AI success should be whether the organisation becomes more capable, not simply whether individuals become faster. The best use of AI should amplify human judgment while creating a stronger system around it.
ALSO READ: Beyond Downloads: What Sustainable App Growth Looks Like
