Skip to main content
Shameek ConyersMarketing Data ScienceResumeContact

Services

Start with the decision, look at the data you already have, and build the smallest thing that answers the question.

Work built around a decision

I help teams answer questions that sit between marketing, customer behavior, and data systems. I start by pinning down the decision, then check whether the data you have can support it. A model or dashboard comes later, if it helps.

A first conversation is worth having when you have campaign, lead, customer, sales, review, document, or spreadsheet data and can't connect it into an answer you trust. Each section below lists the questions I work on and what you get at the end.

Campaign and revenue measurement

Questions I work on

  • Which campaigns bring in paying customers, not just clicks?
  • Where do prospects drop off between the ad, the lead, the visit, and the purchase?
  • Does a high acquisition cost pay off in later customer value?
  • What should we start tracking now so next quarter's comparison works?

What you get

I join campaign, lead, visit, offer, and sales data wherever the identifiers allow it, and I agree the definitions with you first: what counts as a lead, a verified visit, a conversion, and attributable revenue. When the identifiers don't exist yet, I help set them up, from dated flyers to personal offer codes.

Typical deliverables are a measurement map, a clean analysis table, a campaign comparison, a funnel report, and a plain-language summary of what the numbers say and what to do next. When a question needs a true incrementality answer, I design the holdout or comparison that can give one.

For one hospitality client, this work cut cost per attributable revenue dollar on lead campaigns by about 70%.

Prospect and demand analysis

Questions I work on

  • Which prospects should the sales team call first?
  • Which audience, offer, event, or product deserves a small test?
  • Which basket combinations lift the average ticket?
  • What should we compare before spending more?

What you get

I score or rank prospects from your existing records, and I can train a model when the data supports one. For a wholesale supplier's sales team, I built a lead CRM with a random-forest ranker that raised closing rate per call by about 50%. For demand questions, I combine sales, audience, platform, and local-market data, including scraped text from forums and group chats, and turn the finding into a test you can run.

Typical deliverables are a ranked contact list, a demand brief, a test design with a defined outcome, and a follow-up analysis.

Read how local demand analysis pointed a cafe to language-exchange nights. Foot traffic doubled within one quarter of the launch.

Player and market research

Questions I work on

  • What do players of comparable games praise, and what drives negative reviews?
  • How do players rate a design choice across similar games, once blockbusters are set aside?
  • Which community posts gain traction, and how often should we post?

What you get

I pull public reviews and market data, set the comparison and the rules for a finding before looking at results, and report rates with intervals. You get a written study, the data behind it, and a dashboard when one helps.

Read the strategy game market study or the player research case study.

Document search and data workflows

Questions I work on

  • Can analysts find the right passage in old reports without opening every file?
  • Can a language model write the explanation without doing the math?
  • Can messy spreadsheets become a reliable reporting input?
  • Which rows failed validation, and where did they come from?

What you get

For document search, I build ingestion, retrieval, drafting, and evaluation as separate parts, and the system shows the source passages behind every answer. For reporting, I import, join, standardize, and validate the source files. Rejected rows stay visible with their source and a reason, so nothing disappears during cleaning.

Typical deliverables are a retrieval prototype, an evaluation set, a data dictionary, validation rules, database tables, a recurring report, and documentation.

Read about the retrieval assistant behind about 40% less analyst research time or inspect the spreadsheet consolidator.

How an engagement usually runs

Most projects start with a short diagnostic:

  1. Agree on the decision and who will use the answer.
  2. Look at the data you have and how records connect.
  3. Write down the measures, the comparisons, and what the data can't tell us yet.
  4. Build the smallest report, model, or workflow that answers the question.
  5. Walk through the result with the people who will act on it.

If the first project turns up a recurring need, the work can continue as implementation or ongoing reporting. I don't publish fixed prices because the effort depends on data quality, access, and the comparison you need.

Start a conversation

Email shameekconyers@gmail.com with the decision you're trying to make, the data you have, and who will use the result.