Turning automated reporting into time for strategy
Generated by Klaviyo AI
Dagne Dover centralised its marketing data and automated performance analysis with Claude and Klaviyo’s MCP server, saving about six hours each Monday. The post explains how connected data, scheduled reporting and human validation help the team spend less time gathering numbers and more time planning.
- Connected data sources: Dubay brought together Klaviyo data, Shopify data and planning documents in Google Sheets to support plain-language analysis through Claude.
- Scheduled performance analysis: Automated analysis delivers daily findings through Slack and saves about six hours of manual reporting each Monday.
- Human validation: Clean data inputs and Dubay’s review of every report help catch errors before findings are shared across the company.
Dagne Dover’s functional bags are built to keep everything organised and within reach. Caitlin Dubay, director of marketing, wanted the same for the brand’s data.
Dubay runs marketing and analytics for the 50-person brand where everyone owns a piece of the revenue. She was previously working with a decentralised set-up many marketers will recognise: 60% of marketing teams have 6–15 tools in their tech stack, according to Klaviyo’s 2025 State of B2C Marketing Report.
Recently, Dubay consolidated all of Dagne Dover’s data sources to one place, so she can get answers on marketing performance via AI conversations rather than solely relying on reports.
In our latest episode of Klaviyo Originals, Dubay explains how she uses AI to reclaim time and get performance data she can act on right away.
The first step: centralising data with an MCP server
With Claude and Klaviyo’s MCP server, Dubay pulled Dagne Dover’s Klaviyo data, Shopify data, and revenue plans and marketing calendar in Google Sheets into one place.
‘We've basically brought in all of our data sources into our own version of a data warehouse, and now we're able to have almost human-like, analyst-level conversations with AI’, Dubay says.
MCP is an open protocol that helps AI models securely interface with data sources and tools. This is what makes it possible for marketers like Dubay to interact with their data warehouse analytics through Claude and get answers about performance in plain language.
But for that to work, all customer and performance data need to be centralised at the source. An LLM can only answer questions about the performance data that’s visible to it, which is why Dubay’s initial centralisation efforts were so crucial to her success.
Reclaiming report Mondays: ‘Instead of gathering data, we can come up with our weekly game plan’
Dubay’s analysis runs on its own schedule, so she doesn’t need to manually build reports. ‘The AI workflow is based on a trigger’, Dubay says. ‘I have schedules in place, so that each morning it runs the analysis and then Slacks me, as if I had a data analyst on my team telling me exactly what happened the day before.’
These conversations keep Dubay in the loop, and now her job is to check the output and make it useful for everyone else. ‘From there, I’m reviewing that and making sure it makes sense, and translating it in a way that’s valuable for our team’, she explains.
Before the switch, Dubay needed most of Monday to open 10 tabs and manually pull data for the team's weekly report, which she presented at a 4 p.m. meeting. Automating that process gave her back about 6 hours each Monday.
The meeting now happens at 10 a.m. ‘It really gives us all of Monday back’, Dubay says. ‘Instead of gathering data, we can come up with our game plan for the week and be strategic.’
The extra time also changed how the team works together. ‘We work less in silos’, Dubay points out. ‘So instead of everyone on Monday morning crunching the numbers for the channels they own, we're all able to come together and have a team discussion about the marketing at Dagne Dover as a whole.’
Humans in the loop: ‘AI is only as good as its inputs’
Most marketers have seen AI deliver an answer that turns out to be wrong. Dubay guards against that with clean inputs and a final human review.
‘There’s a lot of fine-tuning that goes into it, and I think the AI is only going to be as good as its inputs’, Dubay says. ‘So, the first step is making sure you have clean sources, and that the data itself is clean to begin with.’
Checking inputs doesn’t catch everything, though. Sometimes a report comes back with a number that doesn’t add up. That’s why every report goes to Dubay before it goes to anyone else.
‘I would say we’re about 95% there, but I think that’s where a human is super important’, Dubay says. ‘I’ll read through it first before I distribute it anywhere in the company, and do some of those data validation checks.’
How Dagne Dover turns data into marketing decisions
With automated reporting, Dubay’s team spends more of the week acting on their data. ‘We always want to have brand at the forefront and remain creative’, Dubay says. ‘But at the same time, we always look to the data to support our theories that our brains are coming up with.’
The same habit shapes how Dagne Dover is preparing for shoppers who use AI. To learn how customers describe their needs, the brand launched an AI customer agent on their website.
‘I think we see humans being a little bit more brutally honest when they don't see the person behind the screen’, Dubay says. ‘I'm really excited to see how people interact with it, because our theory is that customers will tell us exactly what they need.’
Recent Klaviyo research supports that theory: 78% of consumers include emotional or personal context at least some of the time when using AI, according to Klaviyo's 2026 AI Consumer Trends Report.
Dubay is also making sure Dagne Dover shows up in AI search results outside the brand’s owned channels. For AI shopping visibility, she mentions the importance of press, ‘making sure you're mentioned in those top bags for back-to-school roundups’, as well as community threads like Reddit, customer reviews, and product page metadata that matches how customers describe their needs.
From analysing performance to briefing the next campaign
As an analytics expert, ‘sometimes it’s really hard for me, as a non-creative, to thoughtfully communicate what I’m asking them for’, Dubay points out. ‘If you asked me to put together any design, it would never get approved.’
So, she tests her briefs in Composer, Klaviyo’s built-in AI marketing agent, before passing them to her creative team. ‘I put my prompt into Composer, and have it build the email for me’, she explains. ‘I can start to visualise what the outcome of my words is going to look like, and get it to a place that I feel good about, so I can pass it over to a digital designer.’
How lean teams can get started with AI reporting
For lean teams that want to try AI marketing analytics, Dubay’s advice is to start with the tools already in their tech stack. Specifically, she advises teams to:
- Consolidate and integrate your tech stack before you begin. ‘Get your tech stack as integrated as you can’, she suggests.
- Learn from others. ‘I’ll look up TikToks on how to optimise this specific thing within Claude’, she says. ‘There's a wealth of information out there.’
- Audit MCP servers against data sources. ‘Look at the connectors, then see out of your tech stack what you can connect to, and what you can't’, she recommends.
With easier access to Dagne Dover’s data, Dubay and her team can spend less time with spreadsheets and more time with strategy.
Related content
