How Kleene.ai work with agencies as a technical partner

Agencies work with multiple clients, and they usually don't have very technical teams – it's just not what they're built for. So what we do at Kleene.ai is help their customers be more technical: we become an extension of the agency's team. Sometimes it's white-labelled, sometimes it's not, and we deliver data projects for them.

Right now we're mostly working with media and creative buying agencies, which means the projects are the four things that really matter in media: MMM (media mix modeling), MTA (digital attribution – increasingly AI attribution, as the models get smarter), segmentation and pricing.

How it used to work

Imagine an agency is working with a brand like Evian (it isn't, but imagine). Evian would have a data team that pulls all their reports, gets it together, and sends it to the agency – usually in an aggregated, not very detailed format. The agency would then try to analyze it, or outsource it to a smaller analytics agency, and go back to Evian with "this is how we think you're performing."

There are two problems with that. One: the data arrives aggregated and pre-filtered by the brand, so you never get the full picture. Two: agencies don't have the right skills to analyze that data to the fullest.

Sometimes they'd commission a one-off MMM project instead. That typically means one single consultant who takes a static view of the data and comes back with one model, based on that snapshot, with no external inputs. They're only looking at the data they've been provided – not the weather, not exchange rates changing, not holidays. A lot of agencies have gone for these pseudo-MMM models because they're cheap, but they're one-off, and not very indicative of what's actually going on.

That's where we come in. We have access to all the different data sets, including publicly available data, and we can put it all together to give agencies a more accurate and alive view.

Side by side, from the agency's POV, the difference looks like this:

One model per brand, not one model for everyone

So, if a large agency has 20+ clients, do we build one model they roll out to everyone?

No, we partner with each brand specifically and create a bespoke model for each one. There's an underlying logic behind MMM that's the same for everybody, but the inputs are very different for each brand, so every model is tailored.

For the agency, this is far easier than the old way of trying to solve it individually for every single company. They now understand the strategic benefit, they know how to roll it out per customer, and it's actually helping them win business: one agency we work with has already signed three new deals because of this offering – an offering they simply didn't have before us.

What the agency gets out of it

The way I'd describe it is that the agency gets to focus on two things.

First, they're still influencing the strategy, because we're providing the detailed view underneath it. They keep full access to the data and full access to the pipeline – proper self-service analytics – so they can do their own analysis too. That matters, because previously they weren't getting access to all the data at all.

Second, we get to be the experts. They don't need to hire a whole team of data scientists, or build and maintain a modern data stack of their own, because that's us. We take all of that overhead off their hands. And the customer gets more out of it, because there are two teams working at the same time on separate things that come together into one solution.

It also helps agencies build their own internal case studies, which they can roll out to more of their partners.

Why marketing first (and what's next)

We mostly deliver the marketing models for agencies today, but we could use any of them from the KAI Analytics Suite. There's actually quite a lot within media buying around inventory, and we're already working with some agencies on that.

The reason customers care about marketing first is simple: it's where they spend most of their money, so it's where they most want to see results. A company selling physical products will care more about inventory, because that's where their money is held up. But for brands where the product isn't the biggest line, marketing spend is the largest percentage of what they spend – so that's where measurement matters most.

If an agency is reading this

If your clients are still getting one-off, static models with no external inputs, it's worth a conversation. Bring us one brand and one question you can't currently answer; worst case, you'll leave knowing your current setup holds up. Get in touch – or read how we think about AI-ready data platforms first.