Recast’s Post

Many Recast customers have tried open-source media mix modeling (MMM) tools before deciding to use Recast. While these packages are interesting, they only solve the easiest part of MMM: running the model and producing some output The real challenge lies in operationalizing those results. How do you validate the model? Communicate the insights to marketers? Get them to trust and actually use the findings? That's where the open-source tools often fall short without substantial engineering investment. We've heard from organizations whose data science teams run MMMs, but the marketing team is completely unaware of or disengaged from the results. If marketers aren't leveraging the insights, what's the point? Customers turn to Recast because they recognize the potential of MMM but need help making it actionable. They need a solution that ensures consistency, stability, and trust in the results. Pushing data through an algorithm is easy. Operationalizing the insights is the real challenge.

Lesson #3 - Clean Equals Speed Last year at Green Line Digital, we onboarded a shiny new data platform (Adverity). Prior to using Adverity, the company's data analytics stack was primarily combo of Funnel and NinjaCat. Adverity was a fantastic tool to switch to, as it provides the bulk of the capabilities the data team needs for ETL. But of course, as with any transition, change was tough. Our team had some pretty hard deadlines to meet, as we wanted to move off of NinjaCat for dashboards and reporting as quickly and efficiently as possible. While we were actively transitioning to the new platform, we were also LEARNING Adverity. And while Adverity is a low-code solution, the learning curve was still steep! We spent roughly six weeks as a team with a rep learning the ins and outs of the platform -- how we wanted to structure our workspaces, how to set up data collection, practices for transforming/enriching data, and of course the entire loading and visualization side of things. Sometimes you just gotta go with what you've got. We transitioned to the new platform in a couple of months (with not a few bumps along the way 😉), but we DID IT. Then in the later part of last year and earliest part of 2024, our team began evaluating our entire setup. Because while we were able to build useable dashboards, there were many consistent challenges: data not loading quickly enough, dashboards that weren't as user-friendly as possible, and we often lacked flexibility and standardized data. To combat these issues, our team built out a new approach with the idea of data governance and standardization in mind. We wanted clean data, clean transformations, and cleanly built dashboards. To our delight, we found that as we embraced this approach, we not only gained on performance time (an average 30% decrease in processing time). Our ability to QA and troubleshoot data anomalies quickly GREATLY increased. Also, our dashboards and reports began to load snappier, and we're able to mockup new ad-hoc reports quickly as well. Data governance and standardization can often come across like a "negative" -- a way for a data team to just say "no." Ultimately, our end goal and result has actually been the opposite. Standardization allows us to say "yes!" more often. Because clean equals speed -- for everyone.

When the co-founder and CEO of a digital analytics company says digital analytics has been like hell to use... you know it's bad.  It has been hard to get started because companies need a team of engineers to set up the data pipeline and instrument events. It has been hard to use because teams need to know what to query, how to set up cohorts, and what to measure. And it has been hard to get to value because most people don't know how to analyze the data or translate it into actions.  We're changing that with #AmplitudeMadeEasy. Spenser Skates shares more:

When the co-founder and CEO of a digital analytics company says digital analytics has been like hell to use... you know it's bad.  It has been hard to get started because companies need a team of engineers to set up the data pipeline and instrument events. It has been hard to use because teams need to know what to query, how to set up cohorts, and what to measure. And it has been hard to get to value because most people don't know how to analyze the data or translate it into actions.  We're changing that with #AmplitudeMadeEasy. Spenser Skates shares more:

Data Strategy on ID vs. Cohort measurement. In this analogy the leaves are individual IDs, the river is movement and groups of IDs are cohorts. There are two ways to measure the flow of a river. 1) Individual Tracking: We track the path of each leaf as it floats downstream, recording its position and velocity over time. 2) Cohort Tracking: We set up a fixed grid in the river and count the time in-between and direction of groups of leaves as they cross boundaries in the grid. Individual Tracking requires cross party clean rooms, log-ins, identity resolution and a device graph. Log files are very helpful. It is the world of using a sample of deterministic measurement to infer probabilistic measurement. Cohort Tracking requires the deployment of a marketing data model, naming conventions and aggregated data standardization. It is the world of micro-mix modeling and geo experimentation. Is Cohort Tracking a paradigm shift for marketing measurement or is it another example of the old becoming new again and it is already embraced by practitioners? What are the pros and cons of each approach?

Thoughtful perspective from Nathan Woodman on grouping IDs in cohorts as a paradigm shift in marketing. For years, we have focused on the individual for "1:1 targeting," but with the industry shifting to be more privacy-focused and ID controlled, identifying audiences, performance vectors, and optimization levers on a cohort basis in a tactical way to drive performance at scale while respecting privacy requirements. If you can get the same results or better from targeting and optimizing at a cohort level using geo and mix-modeling, while also staying well ahead of privacy standards, why wouldn't you embrace this strategy?

May be it is "consent" vs "probabilistic" measurement mapping ID vs Cohort. The ID measurement is all about managing/having consent (& tools like clean room). As there is not going to be a complete coverage with consented ID's, a hybrid measurement model gives the best coverage? The time, geo travel dimensions seems like fundamental to cohorting.

Great blog post on the value your company can get from embedding analytics to your customers. #EmbeddedAnalytics #DataDrivenDecisions

This Is for You If Your Problem Is Unstructured Data. In a world of unstructured data, finding the right information is like searching for a needle in a haystack. That’s where embeddings step in transforming search and insights with semantic intelligence. With embeddings, we’re helping enterprises: • Perform semantic search that understands meaning, not just keywords. • Cluster and categorize massive datasets automatically. • Personalize recommendations with unparalleled precision. For example, our solution for a giant used embeddings to power product recommendations, driving a 40% increase in customer engagement. Embedding intelligence isn’t just about search it’s about unlocking actionable insights buried in your data.

Great blog post on the value your company can get from embedding analytics to your customers. #EmbeddedAnalytics #DataDrivenDecisions