Technical Discovery & Needs Qualification Questions
The diagnostic front half of a technical sale: how a presales or solutions professional uncovers a prospect's business objectives, pain points and technical environment, and qualifies the opportunity. Covers preparing for and structuring discovery calls and workshops, building credibility and handling guarded or skeptical technical contacts, discovery questioning and frameworks (SPIN, MEDDIC), mapping stakeholders, decision criteria, champions and the approval path, capturing non-functional requirements (performance, security, compliance, availability, integration), mapping the customer's current-state architecture, integrations and network or operational constraints, verifying budget, timeline and measurable success criteria, running a quick feasibility check on custom asks, handling infeasible or conflicting requirements found in discovery, and packaging findings into a discovery deliverable or CRM record. Stops short of building the ROI case or negotiating terms; eliciting product requirements from end users and internal stakeholders outside a sale is covered elsewhere.
A prospect asks whether your product will work with their identity provider for single sign-on. How do you turn that into a discovery conversation: what do you need to learn about their identity setup and requirements, and how do you explain dependencies and likely timelines without overpromising?
Sample Answer
Direct answer
"Does it work with our identity provider?" almost never has a one-word answer, so I would reply "very likely, here is what it depends on" and then use the question to learn their identity setup, requirements and timeline drivers. I would ask which IdP (identity provider, the system that holds identities and authenticates users, for example a corporate directory or cloud identity service) and which protocol, who needs access and how users are created and removed, then explain dependencies and a timeline range in terms of what the customer must do, not just what we do.
Terms: SSO (single sign-on) lets users sign in once with their corporate identity. SAML 2.0 (Security Assertion Markup Language) and OIDC (OpenID Connect) are the two standard protocols for federated sign-in. SCIM (System for Cross-domain Identity Management) is a standard for automatically creating, updating and removing user accounts. MFA is multi-factor authentication.
What I need to learn
| Topic | Question | Why |
|---|---|---|
| IdP and protocol | Which IdP and version? Do they use SAML 2.0 or OIDC for other apps? Any identity federation (several directories or organisations trusting one another's sign-ins)? | Confirms support against our documented list; I check our product documentation rather than answer from memory |
| Scope | All employees, or some groups? Contractors, customers, partners signing in too? | Decides tenants (separate customer workspaces in our product), roles and licensing |
| Sign-in flow | Do users start from our login page or from the corporate portal? Is IdP-initiated sign-in required? | Not every flow is supported equally. In the usual flow the user opens our login page and we send them to the IdP (service-provider-initiated). In IdP-initiated sign-in the user clicks our app's tile in the corporate portal and the IdP sends the sign-in to us without us asking first. Some products support only one |
| Provisioning | Are accounts created manually, just in time (the account is created automatically the first time the person signs in), or via automated sync (SCIM)? How are leavers (people who left the company) removed, and how fast? | Offboarding is the security team's real concern |
| Authorisation | Are roles in our product mapped from IdP groups? Who owns the group-to-role mapping? | Often the hidden work |
| Security policy | MFA, conditional access (IdP rules such as "only from managed devices" or "only from the office network"), session length, device rules, audit logging requirements | May constrain what we support or force a change |
| Ownership and process | Who administers the IdP, how long do their change requests take, is there a non-production IdP to test with? | Typically the longest lead time |
| Existing constraints | A dated go-live, a security review, a certificate rotation cycle (the schedule for replacing the certificates that sign SSO messages, after which the connection must be updated) | Sets the timeline |
Explaining dependencies and timeline without overpromising
I split it into what is standard and what depends on them: "On our side the SSO configuration is normally a matter of days once we have your IdP metadata (a file or URL describing your IdP's endpoints and signing certificate), assuming your IdP uses a protocol we support. The part that varies is on yours: getting the application registered in your IdP, mapping groups to roles, and your security review. I would plan in a range, with the dependency named: for example, two to three weeks if your IdP admin can register the app in the first week, longer if there is a change-control queue." Where that range comes from, as an illustration: week 1, your admin registers the app in the IdP and sends us the metadata (customer, 1 to 5 days); early week 2, we configure SSO and test one sign-in (us, "a matter of days"); later week 2, group-to-role mapping and a pilot group; week 3, your security review sign-off and go-live. If the registration slips to week 2, everything after it shifts, which is why I name it as the assumption. I make clear it is an estimate I will firm up once I have seen their answers, and I record it as assumptions rather than a promise.
What I would say
"Yes, we support SAML and OIDC single sign-on, and I would like to check the specifics for your setup. Which IdP do you use, and is automatic account removal for leavers a requirement? That second one changes the design more than the first. On timing, I would plan around two to three weeks, assuming your admin can register the app in the first week. I will firm that up once I have your answers, and I will tell you early if anything moves it."
Worked example
Customer: "We use a cloud IdP; users in the engineering group only; leavers must lose access within 24 hours; MFA enforced by the IdP." I record: protocol SAML 2.0 (assumed, not yet stated by the customer, to confirm in the next call), provisioning needed (SCIM or equivalent), role mapping by group, MFA handled at the IdP. I flag one open item: "does our SCIM support cover group-to-role mapping? I will confirm by Wednesday" and give the range above with that assumption stated.
Trade-offs and pitfalls
- Saying "yes, we support SSO" and discovering later that provisioning and group mapping were not covered.
- Quoting a go-live date before checking the customer's change-control delay.
- Naming support for a specific IdP from memory instead of checking our documentation and, where needed, a test.
- What would change my answer: if their requirement is an unusual flow (IdP-initiated only, or custom claims, meaning extra attributes beyond the standard ones that the IdP must put in the sign-in message), I escalate to our product team before giving any date.
A prospect opens with: we need 10,000 concurrent users, 99.95% availability and a strict budget. What do you ask before accepting those numbers, what technical information do you collect, and how do you help them prioritize between cost, reliability and time to market?
Sample Answer
Direct answer
I would not accept any of the three numbers yet. Each one is a conclusion someone reached for a reason, and my first job is to find the reason. I would turn "10,000 concurrent users" into a request rate and a peak shape, turn "99.95% availability" into a measured definition and a monthly downtime budget, and turn "strict budget" into a number with a time horizon. Then I would make the customer rank cost, reliability and time to market, because they cannot all be maximized, and recommend the cheapest architecture that meets the one or two they rank highest.
(Terms. Region: a geographic area where a cloud provider runs data centres. Availability zone: one physically separate data centre within a region. Failover: the automatic switch to a standby copy when the main one fails. Warm standby: a smaller copy kept running, ready to take over. Active-active: two or more sites all serving traffic at once. Graceful degradation: continuing with reduced function instead of failing outright. Load-test: deliberately sending simulated traffic to see where the system breaks. Concurrent users: people active at the same moment. Availability: the share of time the service works, here 99.95%. SLA: service-level agreement, the contractual availability promise. SLO: service-level objective, the internal target. Time to market: how soon the first usable version must be live.)
1. What I ask before accepting the numbers
Concurrency (is it 10,000 browsers open, or 10,000 requests in flight?). These differ by orders of magnitude. By Little's law (concurrent users = arrival rate x time between a user's actions, so rearranged, arrival rate = concurrent users / time between actions), if each active user sends one request every 30 seconds:
requests per second=30 s per request10,000 users≈333 requests/s
If each user acts every 60 seconds it is about 167 requests/s. That is a 2x spread (333 against 167) from one unstated assumption. So I ask: where did 10,000 come from (a measurement, a marketing forecast, a competitor's number)? Is it average or peak? When is the peak (a Monday 9am login, a product launch)? How fast must it be reached (steady growth or a launch day spike)? What share of users are only reading, and what share write?
Availability. What is the 99.95% measured on (the whole service or one API), over what window, and does planned maintenance count? What happens to the business when it is down: lost revenue per hour, a regulatory duty, or an annoyance? Is it a contractual SLA with penalties, or an internal goal? The budget it implies, with 30-day month = 43,200 minutes:
| Target | Allowed downtime per 30 days | Per year |
|---|---|---|
| 99.9% | 43.2 min | 8.76 h |
| 99.95% | 21.6 min | 4.38 h |
| 99.99% | 4.3 min | 0.88 h |
Budget. Strict compared with what: a monthly run cost, a total project budget, or a number tied to revenue? Over what horizon (12 months, 3 years)? Does it include people, licences and support, or only infrastructure? Is there approved money or only an expectation?
Time to market. Is there a fixed external date (a contract, a conference, a regulatory deadline), and what is the cost of missing it?
2. Technical information I collect
- Workload shape: read/write ratio, payload sizes, session state, burst pattern, growth over 12 months.
- Data: volume, consistency needs (can two users briefly see different data?), retention, residency.
- Dependencies: third-party APIs and identity providers, because a chain of dependencies sets the availability ceiling.
- Current state: what exists today, its measured peak, its current availability and incident history.
- Latency target (for example "95th percentile under 500 ms", meaning 95 out of 100 requests are faster than that).
- Team: who operates it, whether there is an on-call rota (people on duty to respond to alarms), and cloud experience.
- Failure tolerance: what degrades acceptably (read-only mode, a queue, a cached page) and what must never fail (payments).
3. Testing the availability number against the architecture
Availability multiplies down a chain of dependencies that all must work. An illustrative serial chain: load balancer and app tier at 99.99%, a database at 99.95%, a third-party identity provider at 99.95%:
0.9999×0.9995×0.9995≈0.9989
That is about 99.89%, which is about 47.5 minutes of downtime per 30 days against a 21.6 minute budget. This chain is the no-redundancy worst case, which is why it falls below even 99.9%. In option A below, two zones and automatic failover lift the app and database figures above their single-component numbers, which is how A can plausibly reach 99.9% while this bare chain does not. So 99.95% overall needs either redundancy (a standby copy across a second failure zone) or a graceful-degradation design, not just "a reliable database". These figures are illustrative, not any provider's published numbers; I would take the real component SLAs from the provider documents and the customer's own incident history.
4. Helping them prioritize cost, reliability and time to market
I ask them to rank the three and give each a "must" or "nice" tag, then show three options priced roughly (illustrative, to be replaced with real quotes):
| Option | Design | Cost | Availability expectation | Time to market |
|---|---|---|---|---|
| A | Single region, spread across two availability zones, managed database with automatic failover | Baseline | Plausibly meets 99.9%; 99.95% needs the dependency check above | Fastest |
| B | A plus a warm standby in a second region | Roughly 1.5 to 2x A (ESTIMATE, depends on data transfer and standby sizing) | Meets 99.95% more defensibly | Slower by the cross-region work |
| C | Active-active across regions | Highest, plus higher engineering cost | Beyond what 99.95% needs | Slowest |
My recommendation for a customer who says all three matter: start with A, load-test at 2x the stated peak, and ask what an hour of downtime costs. Then compare like with like: the extra downtime A allows each month (21.6 minutes between 99.9% and 99.95%, i.e. 0.36 hours) times the hourly loss, against B's extra monthly price. If the loss is smaller than the price, A is enough and the 99.95% becomes a target to measure, not a promise to sign. If it is larger, or if the contract carries penalties, B is justified. I would say plainly that C is not required by the stated number.
Worked example
Customer: an online ticketing start-up. Answers: the 10,000 comes from the best launch day they expect; users act about every 30 seconds, so about 333 requests/s average, and they agree to test at 2x (about 667 requests/s). 99.95% was copied from a competitor's page, outage cost is "maybe 2,000 dollars an hour in lost sales" (the customer's estimate), the budget is 8,000 dollars a month, and the launch date is fixed by a festival.
Check on the decision (the 5,000 dollar baseline for A is an assumed figure for illustration): B at 1.5 to 2x A costs 7,500 to 10,000 dollars a month, so 2,500 to 5,000 more, and sits from just under (7,500) to well over (10,000) the 8,000 budget, so only the low end fits and only with no headroom. The extra 21.6 minutes of downtime B avoids is 0.36 hours x 2,000 dollars = about 720 dollars a month of lost sales. Paying 2,500 or more to avoid 720 does not pay back, so A wins.
Outcome: launch on option A with a 99.9% SLO and a clear path to B once real traffic exists. Their ranking becomes time to market first, cost second, and reliability third, which is a decision they made with the facts, not one I made for them.
Pitfalls
- Sizing infrastructure to 10,000 before learning whether it is a hope or a measurement.
- Agreeing to an availability SLA you cannot control, such as one that depends on a third party's uptime.
- Presenting the options as a menu without a recommendation.
- Treating the budget number as fixed before asking what the customer would pay to avoid downtime or a missed date.
Which operational realities of a customer's environment can gate a deployment, and how do you find out about them in discovery? For the ones you uncover, how do you adjust rollout, upgrade cadence and support so the customer still reaches value quickly?
Sample Answer
Direct answer
Operational realities that can gate a deployment (hold it up or block it) are the things around the software rather than inside it: change freezes (periods when production may not be changed, such as year-end) and release windows, who runs and supports the product, existing monitoring and ticketing tools, environment availability, regulatory validation, and the customer's team skills. I find them by asking about how they ran their last similar rollout, and I adjust rollout, upgrade cadence and support so the customer reaches a first slice of value quickly (time to value, the elapsed time from signing to the first measurable benefit), even if full rollout takes longer.
Realities that gate a deployment, and how I ask
| Reality | Discovery question | Adjustment |
|---|---|---|
| Change freezes and release windows | "When can production change, and who approves it?" | Schedule go-live inside a window; prepare cut-over (switching from the old system to the new one) steps in advance |
| Change advisory board (CAB, the group that approves production changes) | "How far ahead must a change be submitted?" | Submit early with a rollback plan |
| Environments | "Is there a test environment that mirrors production?" | Offer a hosted test tenant (a separate test copy of our service that we run for them) if theirs is missing |
| Team skills and capacity | "Who will run this day to day and what have they run before?" | Enablement plan (below) |
| Monitoring and ticketing integration | "Which tools must alerts and incidents flow into?" | Provide connectors (prebuilt integrations) or webhooks (automatic messages our system sends to theirs when an event happens) early |
| Support model | "Do you need 24x7 cover, or business hours, and who calls us?" | Agree support tiers and escalation names |
| Regulatory validation (proving to a regulator that a changed system still works as documented) | "Does a change require formal validation?" | Fewer, larger, validated upgrades, because each upgrade carries a fixed validation cost |
| Upgrade tolerance | "How often can you absorb change?" | Match cadence to their capacity |
Adjusting rollout, upgrade cadence and support
- Rollout: phase it. A pilot with one team gets value in weeks and reveals operational surprises cheaply.
- Upgrade cadence: if the customer can only absorb change twice a year, agree a longer-supported release stream (a version that keeps receiving security fixes for years without feature changes) rather than push monthly updates they cannot take. Example for a regulated customer: if every release needs about three weeks of documented re-testing, twelve monthly upgrades would consume 36 weeks of their year, so two validated upgrades a year (6 weeks of testing) is the cadence that works.
- Support: align hours and escalation with their on-call model, and name the engineers who will answer.
When the operations team lacks the skills
A concrete enablement plan tied to the agreed timeline, for example:
| When | Activity |
|---|---|
| Weeks 1 to 2 before go-live | Role-based training for administrators and operators |
| Weeks 3 to 4 before go-live | Hands-on exercises in a sandbox using their own scenarios |
| Weeks 5 to 6 before go-live | Runbooks (step-by-step operating procedures) for the top incidents, reviewed with their team |
| First 30 days after go-live | Extended support with named contacts, then a review of what incidents occurred |
The timings are illustrative; they are tied to whatever go-live date was agreed.
Why this protects time to value
A customer who is ready to run the product reaches value faster than one who is technically live but cannot operate it. If no operations capacity exists, I propose a managed service (we or a partner run the product for them) or partner for the first phase, rather than promise self-service.
Pitfalls
- Treating operations as an afterthought handled by customer success after the sale.
- Training the sponsor and not the operators.
A customer's requirements document says only that 'the system must be highly available'. Turn that into several measurable requirements you would propose to the customer, say what each would mean in contractual language, and explain how each ties to business impact.
Sample Answer
Direct answer
"Highly available" is not testable, so I would convert it into a small set of measurable targets, each with a number, a measurement method, a contractual term and a business-impact reason. I would propose them as a package: availability percentage, recovery time objective, recovery point objective, performance under load, and maintenance and support commitments. I would also ask the customer to rank them, because every extra nine costs money. A "nine" is one more 9 in the percentage: going from 99.9% to 99.99% is one extra nine and allows ten times less downtime.
Terms: availability is the share of time the service works as agreed. SLA (service-level agreement) is the contractual promise, usually with remedies. SLO (service-level objective) is the internal target behind it. RTO (recovery time objective) is how long an outage may last. RPO (recovery point objective) is how much recent data may be lost. P95 (95th percentile) latency is the response time that 95% of requests beat.
The measurable requirements
| Requirement | Example target | Contractual wording (sketch) | Business impact tie |
|---|---|---|---|
| Availability | 99.9% monthly for the production service | "Service Availability measured per calendar month as the percentage of minutes the service is up, excluding agreed maintenance windows; service credits (a refund or bill discount the vendor owes when the target is missed) apply below 99.9%." Maintenance windows are pre-announced periods when the service may be down without counting against the target. | Cost of one hour of downtime during trading |
| Recovery time (RTO) | 4 hours for a regional failure (an outage of a whole cloud region, meaning one geographic cluster of data centres) | "Service will be restored within 4 hours of a declared disaster." | Time the business can run on manual process |
| Recovery point (RPO) | 15 minutes of data | "No more than 15 minutes of committed data lost in a disaster." | Number of orders or records that would have to be re-entered |
| Performance under load | P95 response under 500 ms at 200 requests per second | "The 95th-percentile response time of requests to the production service shall not exceed 500 ms in any 5-minute window in which the request rate does not exceed 200 per second, measured at the service boundary." | Abandonment, SLA to the customer's own customers |
| Maintenance and notice | Planned maintenance at most 4 hours per month, with 5 days notice, outside 06:00-22:00 local | Scheduled-maintenance clause | Operational planning, avoiding peak periods |
| Support response | Severity 1 acknowledged in 30 minutes, 24x7 | Support schedule | How fast a human starts working the outage |
What the availability numbers mean (computed in a 30-day month)
Downtime allowed per 30-day month is (100 minus availability) percent of 43,200 minutes:
- 99.9% allows 43.2 minutes.
- 99.95% allows 21.6 minutes.
- 99.99% allows 4.32 minutes.
Per year (365 days, 525,600 minutes), 99.9% allows 525.6 minutes, about 8.8 hours. Series dependencies (components that must all work for the service to work) multiply: two components each at 99.9% in a chain give 0.999 x 0.999 = 99.8001%, about 99.8%, which allows roughly 86 minutes a month. I use that to show the customer that a headline number for the service is lower than for each part.
Business impact conversation
I ask: "What does one hour of downtime cost you in revenue or manual work, and what is the worst time of day or year?" If an hour costs 50,000 dollars at peak, then the 43.2 minutes allowed at 99.9% is about 50,000 x 43.2 / 60 = 36,000 dollars of exposure per month. At 99.95% the allowance is 21.6 minutes (18,000 dollars) and at 99.99% it is 4.32 minutes (3,600 dollars). Whether the extra nines are worth their price is then a comparison the customer can make, and they may prefer a higher availability tier (a stricter target at a higher price) for the peak period only. That turns a vague word into a priced choice.
Worked example of the final sentence
"The system shall be available 99.9% per calendar month, measured at the public endpoint, excluding up to 4 hours of notified maintenance; in a regional failure it shall be restored within 4 hours with no more than 15 minutes of data loss."
Trade-offs and pitfalls
- Promising 99.99% without the architecture (multi-zone, meaning copies in separate data centres within a region, or multi-region, meaning copies in separate geographic regions) and cost to support it.
- Leaving out how it is measured and what is excluded: those clauses decide disputes.
- Treating the SLA as the engineering goal: set the internal SLO tighter than the contractual SLA.
- Mixing up RTO and RPO: time to recover versus amount of data lost.
- What would change my numbers: the customer's actual cost of downtime and the price they will pay for each extra nine.
A hospital network wants to use your ride-booking platform to move patients between facilities. During discovery, what non-functional requirements do you probe for, and which of them would most change the architecture you propose?
Sample Answer
Direct answer
For a hospital moving patients between facilities, I would probe the non-functional requirements (NFRs, the qualities a system must have beyond its features: speed, uptime, security, compliance) in six groups: availability, latency and dispatch timing, privacy and compliance, auditability, integration with hospital systems, and safety and accessibility rules. The ones that would most change the architecture are privacy and compliance (it forces how data is stored, isolated and logged) and availability (it forces redundancy and failover), followed by integration with clinical systems. A seventh point sits outside the six probing groups because it comes from the product's own behaviour rather than from a hospital requirement: pricing and fairness rules (below).
What I probe, and why each matters
| NFR area | Questions I ask | Why it matters |
|---|---|---|
| Availability | "What happens if a transfer cannot be booked for 30 minutes? Is there a manual fallback?" | Sets uptime target and failover design |
| Latency and timing | "How fast must a dispatcher (the person or system that picks and sends the vehicle) see a vehicle assigned? How fresh must the ETA be?" | Drives caching, real-time update design |
| Privacy and compliance | "Will patient health information (PHI) be in the booking? Which regulations apply (for example HIPAA, the US Health Insurance Portability and Accountability Act)? Will we need a signed business associate agreement (BAA, the contract in which a vendor handling patient data promises to protect it as HIPAA requires)?" | Changes data model, encryption, tenant isolation (keeping one customer's data and workloads apart from every other customer's), which cloud services are allowed |
| Auditability | "Who must be able to see who accessed or changed a booking, and for how long are logs kept?" | Immutable audit logs (records that cannot be altered once written), retention |
| Integration | "Which systems must exchange data: patient scheduling, electronic health records (EHR)? Which standards, such as HL7 (Health Level Seven) or FHIR (Fast Healthcare Interoperability Resources), the common message and API formats hospital systems use to exchange patient data?" | Adds an integration layer and mapping work |
| Safety and accessibility | "Wheelchair or stretcher vehicles? Trained or vetted drivers? Escort rules?" | Changes the matching logic, not just the infrastructure |
That is six rows because the question deserves breadth; in a real call I would prioritise by asking which two the hospital would fire a vendor over.
Which ones change the architecture most
- Privacy and compliance. A consumer ride-booking app (riders request a vehicle, dispatch assigns one, and an ETA, the estimated time of arrival, is shown) mixes data from many riders. A hospital deployment may need PHI kept apart from general data, encrypted with customer-controlled keys (encryption keys the hospital holds, so we cannot read the data without them), restricted by region, and logged on every access. That can mean a separate tenant or deployment, not a configuration flag. Concretely: the consumer app keeps all riders in one shared database, whereas the hospital would get its own database and encryption key in an approved region, so its patient records never sit beside other customers' data.
- Availability. A missed patient transfer has clinical consequences, so single-region, best-effort uptime is not enough. That pushes toward multi-zone deployment (running copies in several separate data centres in one region so losing one does not stop service), failover (automatic switch to a standby copy), a tested fallback process and a stricter service-level agreement (SLA, the contractual uptime and response promise).
- Pricing and fairness rules (not one of the six groups; I raise it because our own product behaves differently from what the hospital needs). Surge pricing (raising fares when demand is high) is acceptable for consumer rides and plainly unacceptable for patient transport, so the pricing service must be switchable per customer.
Linking each business objective to one technical metric
The hospital's goals are business sentences. I convert each into one number I can test, using the surge-pricing example for contrast:
| Business objective | Single technical metric |
|---|---|
| Staff trust the price they see before booking | Price displayed within a latency target (for example, the 95th percentile, P95, under 1 second) |
| Prices never punish patients | Fixed-fare rule enforced: 0 bookings priced above the contracted rate (an invariant: a rule that must always be true, and a test can check it) |
| Vehicle status is trustworthy | Location and status updates delivered within an agreed update SLA (for example, within 10 seconds) |
| Regulators can audit us | 100% of PHI access events present in the audit log, retained for the period the contract states |
The figures above are illustrative: I ask the customer for their real thresholds.
Pitfalls
- Accepting "high availability" as a requirement. It is a wish until it has a number and a consequence.
- Treating compliance as paperwork at the end. It decides the architecture at the start.
Unlock Full Question Bank
Get access to all 14 Technical Discovery & Needs Qualification interview questions and detailed answers.
Sign in to ContinueJoin thousands of developers preparing for their dream job.