The questions you ask after the report —
answerable now, not next week.
Subgroup cuts, predictor swaps, sensitivity tests — your team, in real time, on the data you already have.
One application of the Electric Insights engine, applied to concept-test data. The same engine also produces structurally identical deliverables across other commercial-research methods, public-opinion polling, and operational data — making cross-method and cross-domain reading possible for the first time.
We don’t replace your supplier.
We replace the slow, expensive part of working with one.
Most concept-test reports answer the questions the supplier thought to ask. The questions you actually have come up in the meeting after the report.
With your current setup, those questions become custom cuts that take a week and cost real money. Each one. Every time the brand team meets.
With this platform, those questions get answered in the meeting they come up in. The data your supplier delivered stops being a static report and starts being something the brand team can interrogate directly.
Three things this does that a report doesn’t.
These aren’t features. They’re capabilities the brand team has on their own, in seconds, after the report has landed.
Subgroup in seconds
“What does this look like for the 25–34 income tier? In the West?”
Click the subgroup, click run. Three answers in two minutes — not three custom cuts in three weeks.
The platform refits the model on each subgroup, with confidence intervals, calibration metrics, and a full simulator on every result.
Drop a predictor, see what changes
“If we remove brand fit from the model, does package appeal still drive purchase intent? What about quality perception?”
Uncheck a variable, click run. The model refits, the new effects display, the comparison to the previous run sits next to it.
Not a what-if scenario layered onto a fixed model — an actual refit, with the same diagnostics as the original.
It tells you when it doesn’t know
When the data won’t support an analysis — subgroup too small, predictors that perfectly separate the outcome — the platform refuses, names the variable causing the problem, and suggests fixes.
That sounds like a constraint. It’s actually a guarantee.
You’ll never present a finding that gets challenged in the C-suite because the underlying fit was unstable.
Where this fits with conjoint, BASES, and your tracker
Insights teams already work with conjoint, BASES, and brand trackers. Each does a job Electric Insights doesn’t do. The platform does a job none of them does — turning each method’s output into the same headline-explained-and-simulatable structure, so the brand team can interrogate every research finding in the same vocabulary.
Conjoint & MaxDiff
Their job: design a choice exercise inside the survey and produce share predictions for the attribute combinations that exercise tested.
Our job: produce the explanatory model and simulator on the binary outcomes around the choice exercise — chose-our-offer, picked-this-claim, top-2 purchase intent, top-box brand fit. Same headline-unit deliverable structure as every other method.
BASES & norms databases
Their job: tell you whether 47% top-2-box is good for your category, and convert the score into a year-1 volume forecast.
Our job: explain why the score landed where it did, and produce the simulator that lets the team test what would change it. Norms answer "is this good?"; we answer "what would move it?"
Brand trackers
Their job: report the score over time and across markets, with the data infrastructure to keep the series consistent.
Our job: separate signal from noise inside the tracker — which perceptions are doing the work on top-box brand fit, which are noise, and how that mix differs across markets. Reported in the same headline-unit structure as the concept test, the conjoint, and everything else.
Each of these is a distinct discipline. The strategic gain isn’t any one of them — it’s the four working together. See the full Approach page for the three buyer worlds — commercial research, public-opinion polling, and operational data — where Electric Insights produces structurally identical deliverables across every method or data source inside each.
Results in the units you already use.
The capabilities above only matter if you can act on the results. So effects come back in percentage points — the same units as the headline.
If the topline is “54.7% top-2 purchase intent,” a predictor’s effect reads as “+8.3 points” or “−5.1 points.” Most modeling tools stop at coefficients or log-odds and leave the translation to whoever is presenting. The translation is the platform’s job, not yours.
Modeling depends on Design.
The capabilities above — subgroup refits, predictor swaps, sensitivity tests — only work as well as the data permits. And the data only permits modeling when the study was designed with explanatory modeling in mind: outcome chosen up front, candidate drivers measured on usable scales, response coverage adequate across subgroups.
Most concept tests aren’t designed this way. They’re designed for descriptive reporting — toplines, attribute scores, demographic breaks. The good news: the gap is fixable, and it’s fixable cheaply if you catch it before fielding.
Two ways to engage:
- Already fielded? We work with what’s there. The diagnostic tier tells you what your data does and doesn’t support before you commit to a full engagement. See diagnostic.
- Still in design? A short pre-fieldwork review costs little and pays back the most. See methodology partnership.
What this is, and what it isn’t
The platform does the analytical work that turns concept-test data into a decision tool. Suppliers do the data collection; norms and forecasting providers do their respective jobs; we do explanation and simulation. Below is exactly what that does and doesn’t mean.
What this platform does
- Refits explanatory models on the subgroups you choose, in seconds.
- Shows the predictors that drive your outcome — with effect sizes, confidence intervals, and calibration.
- Runs simulators where the brand team can stress-test “what if uniqueness perception drops 10 points?”
- Surfaces the data’s honest limits — small samples, perfect-prediction artifacts, weak signal.
- Lives alongside your supplier’s report, not instead of it.
What it isn’t
- Not a norms or benchmarks database. We don’t tell you whether 47% top-2-box is good for your category. BASES, Nielsen, Kantar, and your tracker history do that.
- Not a volume forecast. Concept score to year-1 case volume is a different problem with different inputs. Keep your forecasting partner.
- Not a sample provider. Fielding, recruiting, weighting, and quotas stay with your supplier. We work on the data they collect — though the modeling layer depends on Design choices made before fielding (which factors get measured, on what scale, with what coverage), so there’s a small upstream conversation to have.
- Not the slide deck. The supplier’s analyst still writes the narrative. The platform gives the brand team the tool that sits beneath it.
Two paths if your data isn’t ready: Codebook Construction for studies already fielded, or a methodology partnership for studies still in design.
Beer concept test
A real concept-test dataset, n = 803 craft-beer buyers. Five binary outcomes — purchase intent, package appeal, brand fit, quality perception, premium perception. Eight subgroup variables. Twenty package-reaction predictors.
Switch outcomes, slice by income or region, refit the model, watch the predictors shift. The model behind the platform was built and validated against a published Stata 18 reference.
Open the demoWho this is for
Concept testing
Pre-launch package, formulation, or claim tests. Top-2-box outcomes with attribute batteries and demographic splits.
Brand tracking
Wave-over-wave brand tracking where the question is “why did awareness move” not just “did it move.”
U&A and segmentation
Usage-and-attitudes studies where the strategic question is which attitudes actually predict behavior, not which are most prevalent.
The platform handles any binary outcome with categorical or continuous predictors. If your study fits that shape, this works on it.
“How is this different from a Tableau dashboard?”
Tableau dashboards show you the data your supplier already cut. They don’t fit new models on the fly.
If you want to ask “what predicts purchase intent in the West region for 36–45-year-olds, controlling for brand awareness,” a dashboard can’t answer that — it would need to be pre-built.
This platform fits the model live.
Closer to interactive statistics than to interactive visualization.
How to put it on your data
Engagements are scoped per study, with a fixed-fee diagnostic as the usual first step. The Services page lays out tier options and pricing. Most CPG buyers start with a diagnostic on one existing concept-test dataset.
One sentence, three demos, twenty minutes.
The fastest way to understand this is to see it on a real concept test. We’ll walk you through the platform answering the questions your team would actually ask — subgroup cuts, predictor swaps, sensitivity tests — on the beer dataset or, if you’d rather, on yours.