Responses go missing
An attachment is dropped when a response is re-keyed, or a letter's central argument is summarised away. Nobody can say what was lost.
NexVoice by LigeroNex
A record of consultation analysis that a committee, regulator or court can follow.
For bodies that consult. NexVoice is designed to account for every response, make campaigns visible and count them, look for rare positions and keep them, and put a named person behind every judgement. A model may propose themes; people decide, and every count and quote comes from the stored responses. Everything below is an interactive preview with fictional data.
- Interactive preview
- Fictional data
- No sign-up needed
The problem
Thousands of responses, one deadline, and a record someone may test later.
When a consultation closes, someone must read, code, count and summarise every response, then show months later how each one was handled. Courts have quashed decisions where the substance of a response was lost between receipt and decision.
Campaigns are handled quietly
Template letters and petitions are removed, merged or weighted without anyone saying so, and the count changes with them.
Rare positions disappear
Automatic theming condenses many themes into a few. A view held by a handful of people can vanish inside a bigger one.
Nobody can show who decided
A spreadsheet holds the codes, not who chose them, why, or what a model suggested first.
How it works
A model may propose. People decide, and the record shows it.
NexVoice reads the files your consultation platform, inbox and transcribers produce, and writes a pack you keep. Between the two, every response is accounted for and every judgement has a name on it.
- 1Every response, attachment and link is imported and counted
- 2Campaigns and petitions are grouped and counted
- 3A model proposes themes
- 4Named officers decide the codeframe
- 5Rare and unplaced positions are decided
- 6The export gate checks, then the pack is written
- 01
Account
Was every response looked at?Rows in equal accepted, blank and quarantined, with attachments and links counted too. Every accepted response ends coded, marked as having no content, or unplaced and decided by a named person.
- 02
Group
How were campaigns handled?Duplicates, templates, petitions and paraphrased campaigns are grouped and shown with respondents and distinct texts. Nothing is removed or weighted by the software.
- 03
Decide
Who chose the codes?Model themes load as proposals. Named officers accept, merge, split or discard them with reasons, and freeze versions. Mapping runs only against a frozen version.
- 04
Keep
Were rare positions lost?Small themes, themes merged away by the model and answers with no code go to one register. A rare position is merged only by a named person, with a reason.
- 05
Answer
Was each detailed point dealt with?Each point in each organisation's response is logged against its passage, with where the report deals with it, or why it does not.
- 06
Export
Can someone else check it?The pack is written only when every check passes. Its chain of decisions can be verified against digests you hold, without LigeroNex.
Counts and quotes never come from a model. Every count, total and campaign size is computed from the stored responses, and every quote is an exact passage of a stored answer, checked at export. With the model switched off, NexVoice still imports, groups, supports coding by hand and writes the pack.
Try it yourself
Confirm, decide, keep, then let the gate check it.
Northwold Combined Authority, a fictional body, has closed a consultation on its bus network: 1,284 responses. You are Sam Okoro, a consultation officer. Confirm a campaign the software proposed, decide the codeframe, keep a position the model merged away, answer the open points from organisations, and let the export gate check the record before the pack is written. It takes about four minutes. The authority, the people and every response are fictional.
This window is interactive. Click the menu, the rows and the buttons.
Limits
What NexVoice never does.
A record is only worth having if it is honest about what it shows. These are rules, not settings.
- No generated numbersNo count, total, share or campaign size comes from a model. Every figure is computed from the stored responses and traces to a table.
- No generated quotesEvery quote is an exact passage of a stored answer, held with its position and checked at export. A machine translation is never quoted.
- A proposal is not a decisionA model may propose a theme, a mapping or a campaign. It stays a proposal until a named person decides, and the export cannot read a proposal.
- No promised outcomeIt makes no promise about how any decision will fare if it is questioned, and warrants no accuracy or retention figure. It measures and reports instead.
- Campaigns counted, never weightedCampaigns and petitions are grouped and counted openly. How much weight they carry is the decision-maker's judgement, recorded in the pack.
- Never a bare percentageEvery table shows respondents and mentions with its denominator stated. Nothing is reduced to a percentage on its own.
In the design, not yet built
- Every response accounted forAt import, after grouping and at export. If one response is in none of the four end states, the pack is not written.
- Rare positions keptSmall themes, merged-away themes and unplaced answers go to a register that a named officer decides before export.
- Welsh stays WelshWelsh answers are coded and quoted in Welsh. A translation is a labelled aid for officers, never a replacement and never quoted.
- Your data stays with youThe first paid step is proposed to run in your own environment, with your approved model service. NexVoice's development environment holds synthetic data only.
- Officers are not ratedBulk actions and time on item are reported in aggregate, so the decision-maker knows how the work was done. They are never used to rate officers.
- A preview, not the productNexVoice is at design stage. Nothing in this preview is built yet, and no customer has used it.