VoiceInsights Africa
Operating Model & Implementation
VoiceInsights Africa
From research question to governed evidence, implementation and decision support
“Every Voice. Every Language. Every Insight.”
How institutions work with us
Not every organisation wants to operate a research platform. Some need the evidence delivered. Some already have a team and need the infrastructure behind it. Some need the platform itself as institutional capability. Everything below follows from which of those you are.
These are engagement models, not product tiers. The same evidence chain runs underneath all three, and so do the same boundaries — what changes is who holds the work.
Managed research delivery — we run the study
Who it suits
Programme teams with a deadline and no field operation of their own; donors and foundations commissioning an evaluation; ministries needing an independent read.
What we do
Instrument design, translation into the collection languages, enumerator arrangement and training, collection, quality review, analysis and the governed publication.
What your team does
Defines the question, approves the instrument, reviews findings and makes the decisions. No platform administration.
How it starts
A scoping conversation, then a written proposal with the design, the sample, the channels and the timeline.
Technology-enabled delivery — your team runs it, with us alongside
Who it suits
INGOs and research firms with field capacity; national programmes with M&E officers; consortia running one design across several countries.
What we do
Implementation and configuration, training for each role, design review, channel setup where a communications provider is used, and support through the collection window.
What your team does
Builds and launches the campaign, manages contacts and audiences, runs the enumerators, reviews quality and authors the findings.
How it starts
An implementation scope, then configuration and role training before the first campaign.
How institutions work with us continued
Enterprise platform — you operate it
Who it suits
Multi-country programmes, government agencies with several departments, and organisations whose reporting obligations recur.
What we do
Platform operation, governance configuration, identity and access setup, upgrades, and the assurance evidence a procurement review asks for.
What your team does
Everything operational — designs, audiences, campaigns, field teams, quality, analysis and publication — under your own permissions and scopes.
How it starts
A procurement conversation. The evidence pack states what is implemented, what is configurable and what is not yet assured.
What does not change between them
The evidence chain is the same in all three, and so are its boundaries: the same evaluation boundary, the same collection routes, the same evidence rules, the same language architecture, and the same invitation-controlled access. What changes is who holds the work, not what the work has to satisfy.
Ways we participate in programmes and procurement
An assignment rarely matches one supplier exactly. VoiceInsights can hold the whole engagement, a defined work package inside somebody else’s, or the infrastructure underneath both.
| Mode | Typical situation and what we own | Contracting structure |
|---|---|---|
| Prime / lead contractor | The assignment sits inside our scope and capacity. We contract directly and coordinate the approved delivery partners. Remaining with others: specialist domains named in the bid. | Direct contract with the buyer; subcontracts to approved partners. |
| Joint venture / consortium partner | The assignment needs combined expertise across research, sector knowledge, digital systems, field operations or evaluation. We take the components we can evidence. Remaining with the lead: overall delivery and client relationship. | Consortium or JV agreement formed for that opportunity, with a written scope split. |
| Technical implementation partner | Another organisation leads the research; we own the digital infrastructure, collection workflow, configuration, training, governance setup or reporting technology. Remaining with the lead: methodology and interpretation. | Subcontract or partner agreement for the technical work package. |
| Research / data collection subcontractor | A defined research, survey, data-collection or field-operations package inside a larger programme. Remaining with the lead: evaluation judgement and reporting to the client. | Subcontract with a named deliverable and acceptance criteria. |
| Platform provider | The institution procures VoiceInsights Workspace as infrastructure and operates it. Remaining with the client: all operational delivery. | Licence, with implementation and support scoped separately. |
| Managed research partner | We manage a full or bounded research assignment end to end within the approved scope. Remaining with the client: the question, approvals and the decision. | Deliverable-based assignment contract. |
What work packages we can own
Stated as a lead or partner can plan against — and bounded, because a capability listed without its condition will be read as unconditional.
Research operations
Study and instrument workflow design; digital survey deployment; multilingual collection; voice-enabled collection workflows; enumerator workflow and assignments; offline field data capture; field data collection.
Evidence and quality
Data quality review and double entry; controlled data processing; evidence traceability and lineage; governed analysis against the registered design.
Reporting infrastructure
Report generation infrastructure; publication workflow and approval control; institutional dashboards; governed exports.
Implementation
Platform configuration; role-based user training; institutional rollout; technical support; controlled integrations, configured per deployment.
No roster. One record per unit on every collection surface. Member-level rosters — a household enumerated person by person — are held relationally by the platform, and there is no roster control on the
No causal claim. Change is reported descriptively — a value, a later value, and the difference between them. No causal or attribution claim is produced, because the platform implements no identification stra
No media capture. Photographs, GPS points and file attachments are not question types a respondent or enumerator answers. Location is recorded about a submission, not asked as a question.
Bounded logic. Skip logic is one condition on one earlier question — “if they said no, do not ask the follow-ups” — evaluated by the server as well as the device. Multi-condition expressions, arithmetic ac
Working with partners, joint ventures and consortia
We do not claim internal expertise in a domain simply because an assignment requires it. Where the work needs a sector or thematic specialist, an independent evaluator, a local field partner, an academic institution, a statistical, geospatial, clinical or costing specialist, an implementation partner or a specialist technology integrator, that expertise comes from a partner — named in the bid, not implied by ours.
What we do instead is define, integrate and govern our work package inside the agreed structure, so the lead can see exactly where our responsibility starts and stops.
The commercial and implementation journey
A controlled route from need to operating service. Durations are set per assignment; none is fixed here.
What shapes a proposal
Scope follows operating reality rather than a price list. These are the factors we ask about, and each one moves the shape of the work.
Structure and people
Organisation and institutional structure · programme or assignment grouping · users and roles · collectors and enumerators · specialist partners where required.
Collection reality
Channels · languages · response volume · audio processing · geography · field logistics.
Data and systems
Storage · data products · integrations · hosting · security · reporting requirements.
Service
Training · support and service levels · implementation duration and phasing.
Inside a research engagement
Eight stages, each with a named owner and a reviewable output. Whether your team runs these or we run them for you is what the engagement models answer — the stages themselves do not change.
| Stage | What happens |
|---|---|
| 1 · Design | The research question becomes an instrument: the questions, their order and logic, the languages it will be administered in, and the consent statement a respondent hears first. The design is registered, so analysis later runs against what was planned rather than what was remembered. |
| 2 · Field preparation | Contacts are normalised into one unambiguous form. An audience is a rule over those people. A contact is only dialled where consent was recorded — a blank consent field is treated as unknown, never as permission. Enumerators receive assignments with an area, a supervisor and due dates; offline packages are prepared on the device before the team leaves. |
| 3 · Collection | A person replies however suits them, across the registered routes. Recording a voice does not depend on which language is spoken, and the recording is kept as the evidence of what was said. |
| 4 · Getting the work back | Often the longest stage. Work collected offline sits on the device until receipt is confirmed. An interrupted interview resumes where it stopped; a device running out of room refuses a new interview rather than silently failing to save one; a lost confirmation is re-sent and recognised rather than stored twice. |
| 5 · Evidence management | Every response carries its consent status, origin and lineage. Original audio is retained — including a recording with no discernible speech, because a refusal to answer is a finding and the recording is the proof of what happened. |
| 6 · Quality control | Responses are checked for completeness, duration, consent validity and duplication, and flagged where something does not hold. Double entry compares two independent records of one interview. A supervisor decides; the platform records who decided and when. |
| 7 · Analysis | Descriptive results with the cautions that belong to them. Qualitative coding keeps themes, contradictions and minority views rather than averaging them away. Findings reference registered evidence, so a claim can be traced to the responses behind it. |
| 8 · Reporting and decision support | A governed publication in the formats an institution circulates. A publication whose evidence cannot be resolved is refused rather than produced. Export is a permission, not a button everyone has. |
Who does what
Responsibility moves with the engagement model. This is the allocation we propose from; it is confirmed in writing per assignment rather than assumed.
| Activity | Managed research | Technology-enabled | Enterprise platform |
|---|---|---|---|
| Discovery | Joint | Joint | Joint |
| Methodology | VoiceInsights leads | Joint | Client leads |
| Instrument design | VoiceInsights leads | Joint | Client leads |
| Instrument approval | Client leads | Client leads | Client leads |
| Sampling and audiences | VoiceInsights leads | Client leads | Client leads |
| Translation of the instrument | VoiceInsights leads | Configured by agreement | Client leads |
| Enumerator sourcing | VoiceInsights leads | Client leads | Client leads |
| Enumerator training | VoiceInsights leads | Joint | Client leads |
| Field supervision | VoiceInsights leads | Client leads | Client leads |
| Collection | VoiceInsights leads | Client leads | Client leads |
| Quality review | VoiceInsights leads | Joint | Client leads |
| Data management | VoiceInsights leads | Joint | Client leads |
| Analysis | VoiceInsights leads | Client leads | Client leads |
| Interpretation of findings | Joint | Client leads | Client leads |
| Report drafting | VoiceInsights leads | Client leads | Client leads |
| Report approval | Client leads | Client leads | Client leads |
| Publication approval | Client leads | Client leads | Client leads |
| Platform administration | VoiceInsights leads | Joint | Client leads |
| Security and access control | VoiceInsights leads | Joint | Client leads |
| Export | Configured by agreement | Client leads | Client leads |
| Support | VoiceInsights leads | VoiceInsights leads | Configured by agreement |
| Closeout | Joint | Joint | Joint |
| Specialist domain expertise | Partner or specialist where required — named in the proposal | ||
Multi-channel, offline and multilingual delivery
The respondent decides how they can be reached, not the research design.
| Collection route | Status |
|---|---|
| Phone Call | Proven |
| Proven | |
| SMS / Feature-Phone Fallback | Proven |
| Web Platform Link | Proven |
| Offline Mobile Application | Proven |
| Proven |
Three language decisions, not one
Respondent language
What the person answers in. Recording a voice does not depend on which language is spoken.
Working language
What your team analyses in. A separate decision from the one above.
Report language
What the deliverable ships in. A third decision, made at publication.
Quality, governance and evidence control
Consent and lineage
Consent is recorded with the response, never added afterwards. Every response carries its origin and lineage, and a finding references the evidence behind it.
Review and publication
A person who did not collect the work reviews it. A publication that cannot resolve its evidence is refused rather than produced, and approval is a human act with a named approver.
Access and separation
Access is invitation-controlled — there is no public sign-up. Roles decide what each person sees, organisations are separated, and the audit trail records what was decided and by whom.
AI-assisted, not AI-authorised
Analysis assistance is separated from observation and from human interpretation, and presented as suggestion rather than conclusion. AI does not approve evidence, decide, or authorise publication.
Applicable law
Implementations are designed to support obligations under Tanzania’s Personal Data Protection Act, No. 11 of 2022, alongside the applicable agreement. Where a programme falls under the EU GDPR or another regime, those obligations are assessed and documented during implementation. No independent assessment of GDPR alignment has been carried out, and none is claimed.
Deliverables, commercial models and procurement
What you receive
| Engagement model | Typical deliverables |
|---|---|
| Managed research | An executive report, an editable working document, a presentation, a statistical workbook, the governed dataset export and the evidence pack. |
| Technology-enabled | A configured Workspace, your own governed publications and exports produced by your team, role training, the implementation record and operational support. |
| Enterprise platform | Institutional platform configuration, governance setup, controlled access, recurring reporting infrastructure, implementation and support, and the audit and evidence record. |
Not every format belongs in every assignment. The deliverable set is fixed in the proposal.
Commercial structures
The structure follows the scope. We do not publish a price list, because the factors on the previous page move the work more than any list could represent.
Assignment-based
Fixed assignment or deliverable-based engagement · managed research engagement · bounded pilot.
Platform-based
Implementation fee plus platform licence · annual institutional licence · usage or response-linked components where applicable.
Programme-based
Framework agreement · multi-year implementation and support agreement.
Partnership-based
Subcontract or defined work package · JV or consortium scope allocation.
Procurement pathway
We can engage through an RFI, an EOI, an RFQ, an RFP or tender, a framework agreement, a pilot, direct institutional procurement, a JV or consortium bid, or a subcontracting arrangement. Each is an engagement route subject to your own procurement rules and the requirements that apply in your jurisdiction — we make no claim about procurement legality anywhere.
Why the model is useful
One evidence chain across channels
A voice call, a WhatsApp reply and an offline interview land in the same governed chain, so a finding does not depend on which route the respondent used.
Responsibility is allocated, not assumed
The matrix in this document is written down before the work starts, which is when disagreements are cheap.
Less fragmentation
Collection, quality and reporting sit in one chain rather than three tools and a spreadsheet between them.
Repeatable workflows
A design that is registered can be run again — next round, next country, next year — against the same structure.
Controlled review and publication
What is published has been reviewed by a named person, and what cannot resolve its evidence is not published at all.
Institutional continuity
The audit trail and the governed record survive staff turnover, which is when most evidence chains break.
Movement between models
An organisation can begin with managed delivery and move to operating the platform itself without changing the evidence chain underneath.
We can be part of a team
The assignment does not have to be forced into one provider. We can hold a work package inside a consortium and govern it.
Contact us
“Every Voice. Every Language. Every Insight.”
Tanzania — Head Office
Ali Hassan Mwinyi Road, Kisutu3rd Floor, Haidary Plaza
P.O. Box 77139
Dar es Salaam, Tanzania