Launch for Quant Survey Analytics Firm Crowdmines

In Toronto, Canada, a team comprising MR and statistics professionals, and AI developers, has launched Crowdmines, an agentic AI system for professional quant MR. The offer includes conjoint, MaxDiff with TURF, segmentation, driver analysis and pricing, plus a respondent-quality audit tool.

Eighteen months in development, the software features hard-coded statistical procedures, which means that the same inputs produce the same output on every run. This is in contrast to the likely performance of AI if put in charge of 'the maths'. In this case, the firm says, language models are used for three (other) things: understanding the research question, selecting the analytical approach that best answers it, and drafting interpretation from computed results.

Crowdmines' process starts with a researcher stating a business question: the platform then tests several methods against that question across a range of solutions, returns the comparative fit evidence, and allows the researcher to choose. The firm stresses that researchers can accept, adjust or override at every step. In the words of VP of Research Allen Turner (pictured): 'Researchers tell us they save hours on the modelling and then spend those hours checking the output, because they can't be sure what the model did.'

Crowdmines.ai offers four separate modules, starting with the assumption that researchers have raw SPSS files ready to work with. Survey Guard is an independent audit of a completed survey file, running 61 built-in checks, which researchers can edit and add to if required - every respondent is marked clean, review or exclude, and every flag returns the question, the respondent's answer and the reason it was raised, rather than a score. Research Helper lets a researcher ask plain English questions of a completed SPSS file, from variable checks and cross-tabs to subgroup analysis and correspondence maps. Specialist Agent Hub adds features including segmentation, a typing tool for classifying new respondents into existing segments, driver and pricing analysis (Shapley, Van Westendorp, Gabor-Granger...), and data exploration. Trade-off Platform offers choice-based conjoint with hierarchical Bayes estimation - benchmarked across ten synthetic populations - MaxDiff, and TURF.

Crowdmines.ai with API access is available now, with MCP access to follow shortly. The company, co-founded by Shams Tabrez and Khoa Doan, can be found online at https://crowdmines.ai .