Selected work

Case studies

Three systems, three different problems. Written the way I'd walk a peer through the decisions — including the parts that are still in progress.


01

Real engagementCRM / Data model

Rebuilding the CRM data model for Cleanfix

Problem. Cleanfix — my own small business — was running customer and job data across three disconnected spreadsheets and a CRM nobody trusted. Mileage wasn’t tracked, job status lived in someone’s memory, and no report reflected reality.

Approach.

  • Rebuilt the object model around jobs, not contacts — the unit of work the business actually runs on
  • Automated mileage capture from the field app instead of end-of-week reconstruction
  • Rebuilt pipeline stages to match what work actually happens, not a generic sales template
Field data
CRM model
Automated stages
Reporting

Where it stands. The rebuild is live. Revenue-impact numbers aren’t in yet — the metric I’m watching is time-to-invoice, and I’ll publish it once there’s a full quarter of data behind it.

02

AnonymizedEnrichment / Scoring

An enrichment and lead-scoring pipeline for a B2B SaaS company

Problem. Inbound leads were scored on form-fill fields alone — title and company size — which routed a lot of noise to sales and buried the leads worth a same-day call.

Approach.

  • Enrichment layer pulling firmographic and intent signals at time of form-fill, not on a nightly batch
  • A scoring model weighted against closed-won history instead of a generic ICP checklist
  • Routing rules that separated "score now" from "needs a human read" instead of one threshold
Form fill
Enrichment
Scoring model
Routing

Where it stands. Live and routing leads; I’m still tuning score thresholds against real outcome data rather than calling it finished.

03

Technical buildArchitecture

A cost-gated AI enrichment and prospecting system

Problem. Prospecting lists are cheap to generate and expensive to enrich well — most teams either enrich everything (and pay for it) or enrich nothing (and lose the value).

Approach.

  • Strategy pattern for enrichment sources, swappable per account tier without touching the pipeline
  • Decorator layer that gates expensive AI calls behind cheaper pre-filters, so spend tracks lead quality
  • Built to read the same for a consulting client’s stack and for an engineer reviewing the architecture directly
Source list
Pre-filter gate
AI enrichment
Ranked output

Where it stands. Functional end-to-end; cost/quality gating thresholds are still being tuned per source.

Have a similar problem?

Tell me what's broken in your pipeline and I'll tell you honestly whether it's a two-week audit or a longer build.

Book a discovery call