LAZY EIGHT® /WRITING

Fouettés WITH Data.

How we are using AI to model data that makes our
creative agency more efficient and transparent.

Fouettés WITH Data.

How we are using AI to model data that makes our
creative agency more efficient and transparent.

This dance with data began in 2016 with a fairly simple question: "How should we price our services?" At that time, we noticed that creative agencies used various pricing models—hourly, project-based, success-based, etc. As a service provider, we understood that it ultimately boiled down to time, skill, and deliverables.

Around the same time, Slack introduced its APIs, allowing developers to build Slack bots. To experiment with Slack's new APIs, we decided to create a simple time-tracking bot, which we called Acht (8 in German). Acht allowed any team member to start and pause it while working on a task or cycle. Little did we know that this small experiment and the data it would collect over the years would become the foundation of Lazy Eight today.

DATA MODELS WITH AI

Although we had been playing around with data and manual models for a while, in 2023, we decided to experiment further as AI models gained momentum. Like our bot experiment, this began with a simple question: "If we input all our data into these models, what type of insights can we gain?" And holy moly, did we start seeing some interesting trends and insights. We quickly recognized that we had a rare opportunity to leverage and better understand all the data we had been collecting for nearly a decade across thousands of projects. This post covers some examples of how we are currently utilizing this data and implementing the insights to help us build an even more efficient and transparent agency.

Average Time Per Task/Cycle

we gained precise insights into the average time required for specific tasks. While there are always caveats and deltas, years of data have allowed us to quickly correlate the expected duration of various exercises. This enables us to predict potential costs accurately and measure new hires against existing benchmarks. Below is a sample plot of our last 250 immersion exercises.

DATA MODELS WITH AI

Average Time Per Task/Cycle

A scatter plot of the last 250 immersion exercises we have done. Median time: 14.6 Hours.

Project Costing

Although we don't bill our clients hourly, we use past data to determine project costs. This approach is agnostic of the client's size, meaning our pricing model remains consistent whether the task is for a startup or a large enterprise. Clients appreciate this transparency, knowing they receive a fair and data-driven estimate. Our estimate deltas are now down to an avg of 10%.

Principal Utility

By measuring the hours our Principals put in, we can easily gauge utility rates and provide signals to our HR team for hiring needs. For example, if a particular type of Principal is 90%+ utilized over several quarters, it indicates a need to hire more people. Conversely, underutilization signals the need to seek projects to engage these roles or adjust team strength. It's simple logic, made effective by acht's data collection over the years.

Dynamic Hourly COST OF TIME

As we delved deeper into data, we considered pricing our time based on this information as well. In 2017, we started using dynamic pricing based on supply and demand economics. When our average utility rate was high, we inferred high demand and adjusted prices accordingly. For calculating prices in currencies outside INR (e.g., USD), we use a modified version of Purchasing Power Parity (PPP) through a unique "reverse Netflix index." We won't bore you with the math or more jargon in this post, but it's pretty cool. This ensures that our prices are accurate and fair across different currencies. Below is a sample window of our historical price chart in INR.

Project Costing

Principal Utility

Dynamic Hourly COST OF TIME

NOTE THE price fluctuation DURING THE COVID WAVES

Activity Patterns

One non-obvious insight we discovered was the specific activity patterns of different roles. For example, Project Managers were busiest on Mondays, and THURSDAYS, while creatives and engineers had peak activity on Tuesdays, Wednesdays, and Thursdays. This insight helped shape our work-life balance culture, allowing team members to work 2-4 days a week based on their role.

Black DotS ARE CREATIVE Execution ACTIVITY. Grey Dots ARE Project Management RELATED ACTIVITY.

Creativity as a Utility

One of our most innovative insights is the utility model. We price our engagements on a project basis with milestone-based payments, but often, clients wish to continue working with us. Rather than charging retainers, which we find impractical, we offer a prepaid model. Existing Clients can purchase 40 hours upfront, using them for any task. When they run out, they can simply refill their account. The hours never expire, and if there’s no work in a month, clients aren't billed—unlike traditional retainers. This flexibility and fairness have made our utility model a client favoUrite.

At Lazy Eight, our love for data is an obsession. It drives our decisions, shapes our pricing, and enhances our efficiency. Our data-driven approach not only ensures fair and transparent pricing but also fosters a work culture that values balance and innovation. We are committed to continuing our experiments with AI to model our data and derive more insights. As we gain more findings, we look forward to sharing them.

Now with ai, We believe that more agencies should start using data to enhance their operations, even if they are not technically deep - ai has made it easier. The benefits of a data-driven approach are immense, and we hope our journey inspires others to explore this path as well.