Shameek Conyers | Marketing Data Scientist and Analytics Consultant

Selected results

70%

lower cost per attributable revenue dollar

On lead campaigns, after I rebuilt ad audiences from verified visitors.

30%

higher lead conversion

Once prospect scoring chose who the team contacted first.

foot traffic within one quarter

After launching language-exchange nights picked from local demand analysis.

40%

less analyst research time

With a retrieval assistant the client kept as a recurring paid product.

Services

What I can help with

Every project starts with a decision someone needs to make and the data already on hand. These are the areas where I've done the most work.

Campaign and revenue measurement

Tie campaigns to leads, verified visits, and sales so you can see which spend brings in paying customers and where prospects drop off.

Prospect and demand analysis

Rank the prospects you already have, find underserved demand, and test an offer or event before spending more on it.

Player and market research

Learn what players of comparable games praise and complain about, and how they rate a design choice once blockbusters are set aside.

Document search and data workflows

Find answers in your own documents, keep language models away from the math, and turn messy files into data you can report on.

See all services →

Client and studio work

Case studies

What the business needed to decide, what I built, and what changed.

Connecting paid campaigns to verified visits

Dated flyers, phone numbers, and personal offer codes tied ad spend to people who walked in and what they spent.

Cost per attributable revenue dollar on lead campaigns fell about 70%.

Finding demand for a new event format

Reddit, Meetup, and group-chat analysis showed language exchange was underserved locally.

Foot traffic doubled within one quarter of the launch.

Searching prior research in plain English

A retrieval assistant answered analyst questions from a B2B media company's own reports and showed its sources.

Analysts spent about 40% less time on research.

Player research for a strategy game

A matched comparison of three battle turn systems across 31 games, plus the agent workflow behind the engineering.

A fix I directed made the pathfinding benchmark about 81% faster.

Read all case studies →

Public work

Technical work

Public projects with source code and seed data, so you can rerun the pipeline and check the numbers yourself.

Strategy Game Market Study

209,223 Steam reviews across 29 strategy games, with a transformer sentiment model adopted on a test written before the data was read.

AI-Verified Economic Analytics Pipeline

Calculates every economic claim in Python, lets a language model write the explanation, and checks each displayed value against SQLite.

Spreadsheet-to-Database Consolidator

Turns inconsistent spreadsheets into SQLite, validates every row, and keeps rejected rows with their source and the reason they failed.

Browse technical work →

Applied data science, from question to use

I have nearly five years of experience in marketing measurement, forecasting, experiments, NLP, retrieval, and data workflows, first at Ringer Sciences and then as an independent consultant. Math and econometrics taught me to set the comparison before reading a result. Consulting taught me the rest: an analysis only counts once someone uses it.

About my work View resume →

Start with the decision you need to make.

Send the question, the data you have, and who needs the answer. You don't need a technical spec to start.

Discuss a measurement problem Email directly