Piaza.ai · Own your research
The study that could answer today's question usually already exists. Piaza turns the three files every study already produces, from any agency, into one interactive file you own. Ask the next question, cut it any way, and see the math behind the figures.
The quarterly review
A sharp question lands. The respondents already answered it. The answer is still days, even weeks away.
Your research comes from several agencies, each in its own format and its own tool. Comparing across them, or across years, is its own project.
Answers arrive weeks after fieldwork, and a follow-up question means a request, a quote and a wait, on data the respondents already gave.
The study lives behind a login you rent. When a licence or a relationship ends, the ability to ask it anything ends with it.
Three promises
Your agencies keep designing and fielding. Piaza is an extension of your trusted agencies, not a replacement: it changes what you get back.
The analysis file is ready within hours of fieldwork closing. Any new cut is significance-tested in seconds, and you can ask in plain English.
Piaza takes the data, codebook and questionnaire every study produces, new or archived, from any agency or platform, and turns each into the same format.
Each study becomes a file your company holds, not a login you rent. It keeps working when a licence or agency relationship ends, and you decide who opens it.
Unlike a black-box LLM answer, every number shows its working and round-trips to source. Trust you can defend to the board or a regulator.
Works with any agency, panel, or platform: Qualtrics, Forsta, Decipher, anything. Piaza layers on top; there's nothing to rip out.
Every study becomes a reusable benchmark. Your research history stops being dead PDFs and makes each new study smarter.
The real cost
For a CMO, re-cutting data is rounding error. The real cost is strategic, and it's the reason slow, static research quietly hurts the business: the call that had to be made before the read landed. When a new cut is instant, curiosity stops being a line item.
The read arrived after the go/no-go was already locked. You launched on a hunch and hoped.
Creative and media were committed before the audience data was in. The course-correction cost a quarter.
Positioning set on last wave's numbers, because this wave was still stuck in production.
You can't put a clean number on any of these, which is exactly why they get ignored. Piaza collapses time-to-insight from weeks to hours, so the answer arrives while the decision is still open.
Your research is an asset
Because Piaza works with any agency, every study, whoever ran it, can feed one owned benchmark library. The more you run, the more each new study is worth.
What we're changing
Important numbers leave the team that made them as a deck, a PDF or a spreadsheet nobody can question, or as a dashboard behind seats, logins and a server. A PIA file is both: it travels like a report, answers like an app, and proves every number.
A new category
One self-contained file with five layers inside: the data, a dictionary that declares what every column means and how every measure may be combined, a deterministic analysis engine, the approved analysis frozen with its evidence, and an access layer. Open it in the PIA Viewer app or your Piaza portal.
One file. Email it, drop it on a USB, open it on a plane. Once it is in the PIA Viewer, it works with no connection.
Filters, banners, drill-downs, verbatim exploration. A full analytical experience, not a static export.
AI insights frozen at approval. Every conclusion linked to source data. No drift. Same file, same answer.
Use cases
Proven in market research and in finance: two engines, one format, one viewer.
Insights and marketing leaders who commission studies from several agencies and need answers while the decision is still open.
Agencies and research teams who want to deliver faster and hand clients something better than a deck.
Deal teams sharing a company's numbers with bidders, investors and advisers during a process.
Finance leaders who send management numbers to boards, lenders and investors who do not have the reporting tool.
The data must leave the team that produced it.
Readers need to slice it, not just read it.
Every number must survive scrutiny.
Wherever all three are true, a PIA file fits. One table. Declared rules. An engine that travels with the data.
For agencies
Your clients want more than a deck. They want to ask the next question themselves.
Partner with Piaza · 1:27 · captions on
What's inside a PIA
The analysis runs inside the file, and the engine fits the data. Pick a world to see what it does.
For surveys and studies: respondent by question. Readers slice by region, segment or wave, and every figure recomputes on the device. Same question, same answer, every time.
Nested AND/OR segments, your supplied weights, low-base suppression and significance testing, fully filterable.
Quantified themes with prevalence + confidence intervals, demographic skew, and theme-to-metric bridges.
Divergences, consensus, gap analysis, and driver rankings surfaced automatically across the study.
Cross-study and cross-wave benchmarks, plus industry and custom norms, in one shareable file.
Transcripts are split into speaker turns under each topic, word for word and never rewritten, and sit in the same file as the survey.
Click a figure for its inputs, formula, and a plain-English verdict. Nothing is a black box.
Type the next question and get an answer computed from the respondents, or an honest refusal that says why the study cannot answer it.
A complete Top Line grid, and an AI report written at production from the survey, open ends and focus groups. Same report for every reader.
MaxDiff, Van Westendorp and Gabor-Granger pricing, and comparative tests, each with its own working on every point.
For financial data: period by entity by measure. Management accounts, diligence packs and portfolio reporting, sliced by the reader and reconciled in front of them.
Every measure declares how it combines: flows add up, balances take the latest value, ratios are re-derived from their parts, indices take a weighted mean.
Readers filter entities, regions and periods, and change the period range. Totals, bridges and ratios recompute on the device.
A reconciliation panel shows whether totals and bridges tie out, so a reader can see the numbers close rather than take it on trust.
Each company's measures sit against its peers, inside the same file.
Every figure carries the tag of the rule that produced it, and the reveal prints the rule and its inputs.
A measure with no declared rule is not combined, and a ratio missing a part says why. A wrong total never ships quietly.
The file is encrypted, and access can be set to expire, so it stops opening when the process ends.
Ledger data is a complete record, not a sample, so the engine reports the figures as they are rather than testing them for significance.
Boards, lenders, bidders and investors open the file with no licence or login to your systems.
A snapshot of the numbers, not a forecasting model. No AI writes or answers in the finance file.
Report + judgment
The report is written once, at production, to your research brief, from the survey numbers, the open ends and the focus group transcripts. Then it is baked into the file: every reader sees the same report, and no AI runs when they open it. Your team can edit the words; the figures stay locked, and anything that cannot be verified is withheld before delivery. Stories included. Receipts attached.
Under-35s rate value lowest: 38% give it a top-two-box score, against 51% of everyone else. The gap is significant and holds for men and women alike. Price is the theme they raise most, named by 29% of them, and the groups said the same thing in their own words.
Demonstrated, not asserted
Statistics in a PIA open their own working. The reveal shows inputs → formula → intermediate values → result → a verdict in plain English.
Data accuracy
Paste your data into a chatbot and it will confidently hallucinate figures and quotes. Piaza's AI is grounded: every number it emits is re-derived from the source data before it's baked, and rejected if it doesn't match.
And unlike a SaaS dashboard, the insight is frozen at approval. It won't drift when a model updates, and every stakeholder opens the identical analysis.
Side by side
Static files and SaaS make you do the work by hand. DIY AI chat is fast, but it hallucinates the numbers, sends your data to a third party, and still hands you a static Office file. Piaza automates the analysis and locks it to the evidence.
| Dimension | Piaza.ai PIA files | Office files | SaaS analytics | DIY AI chat |
|---|
Inside the file
Banners, verbatim, and comparison reports, all interactive, all in one file, all built from the same evidence.
| Measure | W1 | W2 | W3 |
|---|---|---|---|
| Awareness | 41% | 44% | 49% ▲ |
| Consideration | 33% | 35% | 34% |
| vs norm | −2 | +1 | +4 |
Simple by design
Upload the survey data, codebook and questionnaire from any platform, plus focus group transcripts if you ran groups. Map questions, themes, and demographics.
More than thirty specialised AI agents process verbatims, quant, banners, and reports. Every insight is evidence-audited and round-tripped to source.
Generate a standalone interactive .pia file. Email it to your client or stakeholders. Done.
From raw data to interactive deliverable in one session.
How the AI thinks
The order matters. Ground truth wins over LLM convenience, always.
Demographics link to quotes by code-level matching. Stats come from the real respondents, never invented.
Researchers add context and commentary that travels with the file.
The LLM organises and narrates. It never sources the numbers. Every figure round-trips back to the data.
Research, done properly
A general AI chat tool can give different answers to the same question, with no evidence, on a model that changes. A PIA file cannot.
Your agency's weights, the banners you specify, and false-discovery control across every test.
Comments are grouped by code into a theme map. The AI writes the theme names; it never decides who belongs.
Industry norms, question targets and your own private norm sets, each comparison stated with its base.
The report quotes the groups as their own source, beside the survey. A survey answer is never passed off as something said in the room.
Rewrite any paragraph of the report. Figures are locked to the data, so an edit cannot change one.
Every agency, wave and year in one client portal, with roles and an archive you own.
Under the hood
Specialised agents process your data in stages, with deterministic checks between them. Where an exact rule is enough, no model is called at all.
Every AI conclusion links to source data. A six-layer validation stack round-trips every number and blocks anything it can't verify. Full audit trail embedded in the file.
Built from the data, codebook and questionnaire a study already produces, whoever fielded it, including archived studies.
Researchers add context and interpretation directly into the file. Notes and comments travel with the data permanently.
Trust & sovereignty
Each PIA file carries the study data it needs, so recipients never need access to your raw files. They open it in the PIA Viewer app or your client portal, and access can be set to expire.
Role-based access, embedded. A PIN or password gate plus per-viewer roles, all inside the file:
Delivered as a hardened, compressed .pia container that passes email scanners and opens offline in the PIA Viewer.
Works with no connection in the PIA Viewer
A file can lock when access ends
Read, comment or edit, per viewer
You own the file, not a login
Commercial model
No subscription. No per-seat fees. No recipient costs. You commissioned the research, so you should own the deliverable.
One flat fee to generate each PIA file. No usage meter, no renewal conversations. Predictable, project-based pricing.
Share with anyone: your CMO, your board, your global team. Your cost is the same whether 5 people read it or 500.
The file lives in your filesystem. No rent, and no lost access when you switch agencies.
Any agency. Yours to own. Answers on demand. Own your research. Built for the curious.
Or email hello@piaza.ai
Caution on the 55+ read: base is small (n=24). Treat as directional, and recommend a follow-up boost before greenlight.
Agree. Let's flag this in the client readout.