Shepherd · Solutions
Case Studies

A few projects we're proud of.

Real engagements across Power BI platforms and AI solutions. Names are generalized to respect client confidentiality.

Our productAI Solutions

DealDocket.ai, our proprietary AI real estate investment platform

Shepherd Solutions (proprietary product) · Real estate / private investment · Built and maintained in-house

Visit live site

DealDocket.ai is our own AI-driven real estate investment platform. It automatically sources deals, runs the underwriting, and stress-tests scenarios, replacing weeks of manual work per deal.

Challenge

Real estate investors hunt deals manually, build one-off Excel underwriting models, and re-run scenarios from scratch every time financing or assumptions shift. Good deals slip through and capital sits idle. We built DealDocket to fix it, and to prove what's possible when AI is at the core of a product.

Approach

  • Deal sourcing pipeline that ingests MLS, off-market, wholesaler, and public-record feeds, then de-dupes and scores against the investor's buy box
  • AI-assisted underwriting engine: cap rate, cash-on-cash, IRR, BRRRR, and value-add models with sensitivity analysis baked in
  • Scenario engine for financing, rate, vacancy, capex, exit, and market shocks, fully reproducible with stored assumptions
  • Natural-language assistant on top of the deal database, ask 'compare these three deals at 7% rates with 5% vacancy' in plain English
  • React + TypeScript UI with map view, deal pipeline, and one-click LP-ready memos generated from the model output
ReactTypeScriptNode.jsPythonPostgresOpenAI / ClaudepgvectorMapbox

Outcomes

Days → minutes
to underwrite a new deal end-to-end
10x+
deals evaluated per analyst per week
1 click
LP-ready memos from any scenario
AI Solutions

AI-driven SaaS website with self-serve billing baked in

B2B SaaS startup · SaaS · 8 weeks

Designed and shipped an AI product website where prospects can chat with an AI assistant, sign up, try the product, and start paying, all without ever leaving the site or talking to sales.

Challenge

The team needed to launch an AI product fast, but every off-the-shelf option required stitching together AI, billing, auth, and onboarding from scratch. Months of plumbing before a single customer could pay.

Approach

  • Marketing → product → billing flow with AI woven through every step
  • Embedded AI assistant that answers product questions and qualifies leads on the marketing site
  • Stripe Checkout and customer portal for self-serve subscriptions, plan upgrades, and invoices, no manual quoting
  • AI-powered onboarding that personalizes the first-run experience based on the customer's stated use case
  • Auth (email magic links + SSO option), tier-based access control, and usage metering tied to billing
  • Production deploy on Vercel with cost and quality monitoring from day one
ReactNext.jsTypeScriptStripeOpenAI / ClaudePostgresVercel

Outcomes

Weeks
from kickoff to first paying customer
0
human touchpoints required for a customer to subscribe
100%
of code, AI prompts, and billing logic owned by the client
Power BI

Unified Power BI platform for a field-services operator

Regional field-services company · Field services / HVAC · 8 weeks

Replaced 30+ spreadsheets and three siloed reports with a governed Power BI semantic model and executive dashboards.

Challenge

Operations leaders were piecing together weekly numbers from disparate spreadsheets. Definitions of 'revenue' and 'completed job' differed by team, making leadership meetings about reconciling, not deciding.

Approach

  • Workshops with ops, finance, and field leadership to agree on metric definitions
  • Star-schema model in Power BI with dataflows from the dispatch system, ERP, and CRM
  • Row-level security so franchise managers only see their territory
  • Executive, dispatcher, and technician dashboards plus a paginated report for invoicing
Power BIPower QueryDAXSQL ServerAzure Data Factory

Outcomes

75%
less time spent on weekly reporting
1
single source of truth across the business
3 days → 1 hour
to close the weekly ops review
Power BI

CFO reporting pack: month-end close cut from 6 days to 1

Multi-entity services group · Professional services · 10 weeks

Designed a governed Power BI finance model and CFO dashboard pack covering P&L, cash, AR aging, and entity-level drill-throughs.

Challenge

Finance was rebuilding the same Excel-based pack every month from three GL exports. Errors slipped in, the CFO couldn't trust the variance commentary, and the team was burning out at month-end.

Approach

  • Standardized chart of accounts mapping across entities in dataflows Gen2
  • Star-schema model with calculation groups for time intelligence and currency
  • Executive P&L, cash, and AR dashboards with drill-through to entity and account
  • Deployment pipeline (dev → test → prod) with a documented release checklist
  • Recorded handoff and a measure-authoring guide for the FP&A team
Power BIMicrosoft FabricDataflows Gen2DAXCalculation Groups

Outcomes

6 days → 1
month-end close cycle
100%
of variance commentary tied to source numbers
0
spreadsheet-only reports left in the close pack
Power BI

Sales performance model rebuilt for a 200-store retailer

Specialty retail chain · Retail · 6 weeks

Rebuilt a 9 GB Power BI model into a tuned star schema with sub-second visuals, and freed the analytics team from refresh anxiety.

Challenge

The existing Power BI model was a single flat table fed by 14 queries. Refreshes failed weekly, dashboards took 30+ seconds to load, and the team was scared to add a single measure.

Approach

  • Vertipaq analyzer audit and a performance baseline against business-critical visuals
  • Refactored to a star-schema with dimension tables and incremental refresh
  • Rewrote 40+ DAX measures using variables and calculation groups
  • Added RLS for store managers and district leaders
  • Power BI Projects (.pbip) in Git with a PR review checklist
Power BIDAX StudioTabular EditorGit / .pbipSQL Server

Outcomes

30s → <1s
median visual load time
9 GB → 1.4 GB
model size after Vertipaq tuning
0
refresh failures in the first 90 days
Want something like this? Talk to us
Free 30-min discovery call

Have data, ideas, or a backlog you can't get through?

Tell us what's slowing you down. We'll come back with a focused plan, a realistic timeline, and a fixed scope for the first milestone.

Book a discovery call See what we build