Technical Discovery Before Building Custom B2B Software

Why skipping discovery creates expensive rework

Requirements documents written without operators describe an ideal process, not Tuesday morning. Edge cases live in exceptions approvers grant verbally, in report exports finance trusts more than the UI, and in integration fields someone renamed in ERP last year. Engineering estimates without discovery assume happy-path APIs, perfect test data, and instant security approvals. When reality arrives, velocity collapses and trust erodes between business and delivery. Discovery reduces ambiguity on four axes: who does what in the workflow, which system owns each data element, what must be true at go-live for operations to continue, and what can wait until phase two. Without those answers, every sprint becomes a negotiation. B2B contexts amplify risk: role segregation, audit trails, multi-site configuration, and ERP and enterprise integrations that were 'out of scope' until week six.

• Undocumented exceptions and workarounds operators rely on daily
• Integration and security lead times underestimated in quotes
• Conflicting priorities between departments surfaced only at UAT
• MVP scope that includes reporting parity with legacy exports

Deliverables that make discovery actionable

Discovery should produce artifacts engineering and procurement can reference, not slide decks alone. Minimum set: current-state workflow summary (as-is), target workflow for MVP (to-be), entity and ownership table, integration inventory with direction and frequency, non-functional requirements (latency, availability, residency, audit), risk register with owners, and a phased roadmap with non-goals. Add a vertical slice definition: one end-to-end path you will build first, with acceptance criteria operators can validate. Example: 'field technician submits inspection, supervisor approves, status posts to ERP as draft record' with explicit failure behavior. Include open questions log with decision deadlines. Unresolved items become change orders if they slip past sign-off. For SaaS-bound products, align discovery output with multi-tenant architecture decisions early: tenant hierarchy, SSO expectations, and entitlement rules belong in discovery, not sprint three.

• As-is and to-be workflows with named roles and systems
• Entity ownership and integration contract notes
• NFRs: performance, security, residency, retention
• MVP slice with acceptance criteria and explicit non-goals

Activities that surface real constraints

Interviews alone are insufficient. Combine: structured interviews with operators and approvers; shadowing a half-day of real work (or screen recordings with consent); review of exports, emails, and tickets that represent exceptions; read-only access to staging ERP or CRM to validate fields; and a technical spike on the riskiest integration or permission model. Ask operators to show the last time something went wrong: what did they do, which system did they trust, who did they call? Incidents reveal authoritative behavior better than happy-path demos. Map permissions: which roles create, approve, view PII, export bulk data, impersonate users. Security questionnaires arrive faster when you already documented role boundaries. Timebox spikes to two to five days. Goal is learning, not production code. Spike outcomes: 'ERP create order works with these payloads' or 'SSO blocks test users without SCIM'. For internal tools replacing spreadsheets, cross-check build vs buy for internal tools so discovery does not assume a greenfield process that nobody follows.

Integration, security, and operations questions for B2B

Discovery must answer operational questions buyers forget until launch week: backup and restore expectations, on-call ownership, deployment windows, rollback rules, and monitoring on business journeys (not only server CPU). Integration checklist: systems involved, API vs file vs manual bridge, sandbox availability, credential approval path, rate limits, idempotency needs, reconciliation owner, and month-end freeze rules. Security checklist: authentication method, MFA policy, data classification, retention and deletion on churn, penetration test timing, and whether customers require dedicated infrastructure. If modernization is in play, link discovery seams to incremental legacy modernization so phase one does not secretly become a full rewrite.

Turning discovery into MVP scope and milestones

MVP means smallest scope that delivers measurable operational value, not 'fewer buttons'. From discovery, rank workflows by pain, frequency, and build risk. Cut reporting parity, nice dashboards, and secondary integrations unless contractually required. Translate roadmap into milestones with demo audiences: operators for workflow milestones, IT for integration milestones, leadership for risk reduction. Each milestone should have a rollback story if production trial fails. Attach estimates only after discovery sign-off. Ranges should name assumptions: sandbox ready week one, single ERP instance, no air-gapped environment. When assumptions break, re-estimate transparently. Budget conversations should use realistic development cost drivers with discovery listed as its own line, not absorbed into 'Sprint 0'.

Running discovery with a contractor vs in-house

Contractor-led discovery works when you need senior facilitation plus technical spikes, and internal product ownership is thin. In-house-led works when domain experts are available daily and you only need external help for spikes. Contract structure: fixed fee or timebox (one to three weeks typical for a focused B2B workflow), clear deliverable list, joint sign-off with business owner, and IP ownership of artifacts in your repository from day one. Red flags: discovery billed hourly without cap; no access to operators; deliverables only wireframes; refusal to document integration risks; or pressure to skip discovery to start 'real coding'. Handoff criteria to build phase: signed scope, open questions below an agreed threshold, sandbox access confirmed, and named internal owner for product decisions during build.

Timeline and budget for discovery

Focused B2B discovery for one primary workflow often runs one to three weeks with a senior engineer plus product owner time from your side. Multi-site, multi-system, or regulated environments extend calendar time because scheduling stakeholders and obtaining sandbox access dominate. Cost is usually a fraction of build phase (often ten to twenty percent of MVP build) but pays back in avoided change orders. Underfunded discovery shows up as 'surprise' integration sprints priced later. Do not conflate discovery with free pre-sales consulting. A short feasibility call is fine; a proper discovery is paid work with defined outputs. After discovery, re-baseline timeline for build. If build was quoted before discovery, treat the quote as hypothesis until sign-off.

Next steps

List your riskiest unknown: integration, permissions, compliance, or operator adoption. Schedule three operator interviews and one export review this week. Decide who signs off scope on your side. See other resources, book a short call to outline a discovery timebox, or send a message with industry, systems involved, target go-live, and whether you have staging access to ERP or legacy apps.

FAQ