Search Consultancy
Search consultancy as a practice — who does it, what they offer, and how to run one effectively.
Known Consultancies & Firms
| Company | Focus | People |
|---|---|---|
| OpenSource Connections | Search relevance consulting + evaluation tooling (Quepid, RRE) | Doug Turnbull, Giovanni Fernandez-Kincade |
| Sease | Apache Solr, Elasticsearch, neural IR (LTR, ColBERT, reranking) | Nicolò Rinaldi |
| The Search Juggler | Vendor-neutral strategy & audit, market positioning, events | Charlie Hull |
| BigData Boutique | Elasticsearch consulting | — |
Individual Consultants
- Charlie Hull — 25+ years, vendor-neutral solo practitioner, Haystack organizer
- Doug Turnbull — OSC founder, Relevant Search author, relevance methodology
- Giovanni Fernandez-Kincade — OSC principal, search relevance
- Nicolò Rinaldi — Sease, neural IR and LTR specialist
What Search Consultants Typically Offer
Strategy & Audit
Diagnose why search is underperforming — relevance issues, missing measurement infrastructure, wrong technology fit, poor team structure. Deliver a prioritized improvement roadmap.
Hands-On Technical Work
- Technology selection (Solr vs. Elasticsearch vs. Vespa vs. OpenSearch)
- Query design, relevance tuning, ranking model development
- Learning to Rank and neural reranking implementation
- Search quality evaluation setup (Relevance Program Setup)
Organizational Advisory
- Building a search team — roles, hiring criteria, structure
- Stakeholder communication: translating relevance metrics to business outcomes
- Relevance program governance: keeping eval infrastructure alive across quarters
Training & Knowledge Transfer
- Relevance fundamentals for engineering teams
- LTR and embedding-based search workshops
- “Search Optimization 101” style walkthroughs
Market Positioning (for vendors)
- Helping search product companies articulate differentiation
- Reviewing product offerings against competitive landscape
- Raising community visibility through speaking, writing, events
How to Run a Search Consultancy
Positioning
- Pick a lane: pure relevance/evaluation (OSC model), engineering-heavy neural IR (Sease model), or broad strategic advisory (The Search Juggler model). Generalists lose deals to specialists; specialists miss broad engagements.
- Vendor neutrality is a feature: clients trust advisors who have nothing to sell. Explicitly not being a platform SI partner commands a premium for unbiased guidance.
- Build in public: blogging, conference talks, open source tools (Quepid, RRE) create inbound without sales effort. The best search consultants are known before they pitch.
Engagement Models
- Hourly / day-rate — low commitment for clients, good for audits and quick advice; easiest to sell
- Fixed-scope project — audit + roadmap deliverable; sets clear expectations, easier to price
- Retained advisory — monthly hours for ongoing guidance; highest LTV but requires established trust first
- Offer Calendly or similar for frictionless first contact — remove barriers to “just a quick call”
What Clients Actually Need (vs. What They Ask For)
- They ask: “Is our search engine the right one?” — They need: a measurement baseline before switching anything
- They ask: “Can you fix our ranking?” — They need: a judgment set and an eval loop first
- They ask: “Should we add AI?” — They need: to quantify current relevance gaps before adding complexity
Key principle (Charlie Hull): “Even the most sophisticated search system will fail in some cases” — the goal is graceful failure handling and categorized understanding of problems, not perfection.
Specialist vs. Generalist
- Specialist (neural IR, LTR, e-commerce) earns more per project and wins RFPs in their domain
- Generalist with a network (The Search Juggler model) can cover broad engagements by subcontracting or referring specialists
Network as Infrastructure
- Actively maintain a network of specialists to refer work to and receive overflow from
- Contribute to the community (conferences, Slack groups, open source) — this is pipeline, not altruism
- Haystack, Berlin Buzzwords, and London Search & AI Meetup are core gathering points
What Makes a Good Search Consultant
- Understand that search is a product problem as much as a technical problem
- Measure before recommending — resist tuning before establishing a baseline
- Translate between engineering (NDCG, precision@K) and business (revenue per search, abandonment rate)
- Know your limits: refer out rather than bluff in unfamiliar territory
- Write and speak publicly — thought leadership compounds over time
Further Reading
- Relevance Program Setup — foundational methodology clients need before any tuning
- OpenSource Connections — dominant consultancy; Quepid tool; Doug Turnbull’s writing
- The Search Juggler — Charlie Hull’s solo practice; vendor-neutral model
- Sease — Italian firm; neural IR and Solr specialist
- Managing a Search Team — organizational context consultants work within