Market research tool
Brand funnel analysis
How results are checked
Named statistical review pending. Selected automated reference cases are published on the validation page.Biggest leak
34.4%aware → would consider
Brand C: would consider among the aware
Other brands combined 60.1% · z = -11.1, p < .001 · weighted, Kish n_eff
A brand funnel (purchase funnel) shows what share of respondents is aware of each brand, would consider it, has bought it, bought it recently and buys it most often. The useful number is the conversion between steps — for example the share of aware respondents who would consider the brand — because it separates a reach problem from a persuasion or loyalty problem. Comparing each brand's conversion with its competitors shows its biggest leak.
Analyse a brand funnel from your tracker
Open a respondent-level file and pick the question behind each stage: a set of yes/no columns per brand (SPSS multiple dichotomies such as aware_1 … aware_5) or a column listing the brands chosen ("Brand A, Brand C", codes, or a single brand). Add a weight and a segment if you have them, or type stage percentages and base sizes instead. Everything runs in this browser tab and the file is never uploaded. "Use sample data" loads a synthetic five-brand snack tracker.
CSV, Excel, SPSS, Stata, SAS, R, Parquet, SQLite or JSON. The file is read in this browser tab and never uploaded. For Triple-S, select the .sss or .xml metadata file and its data file together.
Separate values with commas, semicolons or tabs (paste from a spreadsheet works). Conversions assume a nested funnel: everyone at a stage is also at the stage before, so the conversion base is the previous stage's share of the base n. The example rows are the synthetic sample's unweighted funnel.
Biggest leak
Brand C: Aware → Would consider
34.7% of respondents at “Aware” move on to “Would consider” for Brand C, against 61.1% for the other brands combined.
Two-proportion z = -11.88, p < .001 — significant at 95%.
From entered percentages; Wilson intervals at 95% confidence.
Funnel by brand
Bars show the % of all respondents at each stage; the thin line is the 95% confidence interval.
| Brand | Aware | Would consider | Ever bought | Bought recently |
|---|---|---|---|---|
| Brand A | 90.3%88.1–92.2 | 56.1%52.6–59.5 | 44.3%40.9–47.8 | 24.4%21.6–27.5 |
| Brand B | 74.0%70.9–76.9 | 42.1%38.7–45.6 | 27.5%24.5–30.7 | 15.4%13.1–18.1 |
| Brand C | 81.9%79.1–84.4 | 28.4%25.4–31.6 | 19.9%17.3–22.8 | 10.5%8.6–12.8 |
| Brand D | 56.4%52.9–59.8 | 31.6%28.5–34.9 | 25.0%22.1–28.1 | 8.4%6.7–10.5 |
| Brand E | 47.4%44.0–50.9 | 33.9%30.7–37.3 | 22.4%19.6–25.4 | 12.9%10.8–15.4 |
Conversion heatmap
Each cell is the share of the previous stage that reaches the next one. Colour shows the gap to the other brands combined: blue above, amber below; ▲/▼ mark significant gaps, the outlined cell is the brand's biggest leak.
| Brand | Aware → Would consider | Would consider → Ever bought | Ever bought → Bought recently |
|---|---|---|---|
| Brand A | 62.1%+9.8 ptssignificantly above the other brands | 79.0%+9.3 ptssignificantly above the other brands | 55.1%+5.3 pts |
| Brand B | 56.9%+2.5 pts | 65.3%−9.1 ptssignificantly below the other brands Biggest leak | 56.0%+5.6 pts |
| Brand C | 34.7%−26.4 ptssignificantly below the other brands Biggest leak | 70.1%−2.7 pts | 52.8%+1.5 pts |
| Brand D | 56.0%+1.4 pts | 79.1%+8.0 ptssignificantly above the other brands | 33.6%−21.8 ptssignificantly below the other brands Biggest leak |
| Brand E | 71.5%+19.2 ptssignificantly above the other brands | 66.1%−7.7 ptssignificantly below the other brands Biggest leak | 57.6%+7.3 pts |
Conversion table
| Brand | Step | Conversion | 95% CI | Base n | Other brands | p | Note |
|---|---|---|---|---|---|---|---|
| Brand A | Aware → Would consider | 62.1% | 58.5–65.6 | 722 | 52.4% | < .001 | |
| Would consider → Ever bought | 79.0% | 75.0–82.5 | 449 | 69.7% | < .001 | ||
| Ever bought → Bought recently | 55.1% | 49.9–60.2 | 354 | 49.8% | .100 | ||
| Brand B | Aware → Would consider | 56.9% | 52.9–60.8 | 592 | 54.3% | .269 | |
| Would consider → Ever bought | 65.3% | 60.1–70.2 | 337 | 74.4% | < .001 | Biggest leak | |
| Ever bought → Bought recently | 56.0% | 49.4–62.4 | 220 | 50.4% | .134 | ||
| Brand C | Aware → Would consider | 34.7% | 31.1–38.4 | 655 | 61.1% | < .001 | Biggest leak |
| Would consider → Ever bought | 70.1% | 63.8–75.7 | 227 | 72.8% | .393 | ||
| Ever bought → Bought recently | 52.8% | 45.0–60.4 | 159 | 51.3% | .725 | ||
| Brand D | Aware → Would consider | 56.0% | 51.4–60.5 | 451 | 54.7% | .594 | |
| Would consider → Ever bought | 79.1% | 73.7–83.7 | 253 | 71.1% | .009 | ||
| Ever bought → Bought recently | 33.6% | 27.4–40.4 | 200 | 55.4% | < .001 | Biggest leak | |
| Brand E | Aware → Would consider | 71.5% | 66.8–75.8 | 379 | 52.3% | < .001 | |
| Would consider → Ever bought | 66.1% | 60.3–71.5 | 271 | 73.8% | .010 | Biggest leak | |
| Ever bought → Bought recently | 57.6% | 50.3–64.6 | 179 | 50.3% | .074 |
Conversion = respondents at both stages ÷ respondents at the previous stage (the conditional base). p compares the brand with all other brands' respondents at the same step pooled (two-proportion z-test); the same people answer for several brands, so read it as a guide. Bases under 30 are marked “low base”.
Stage percentages versus conversion
Stage percentages are shares of everyone interviewed, so a small brand looks weak at every stage. A conversion is a share of the previous stage only: of the people who know the brand, how many would consider it; of those who have bought it, how many bought it recently. That puts big and small brands on the same footing, and it is the number to compare.
A brand that converts well but has low awareness needs reach; one with high awareness and weak consideration has an image or relevance problem; weak repeat after trial points to the product, price or availability. To cut the funnel by more variables at once, build the stage flags as variables and tabulate them in the workspace.
The method
Formula
Conversion(k) = respondents at stage k and stage k−1 ÷ respondents at stage k−1
where
- stage %
- weighted share of all respondents at that stage, with a Wilson score interval
- conversion
- a conditional proportion: its base is the previous stage, and its Wilson interval uses that conditional base
- vs other brands
- two-proportion z-test (pooled) of the brand's conversion against all other brands' conversions at the same step combined
- biggest leak
- the step where the brand converts furthest below the other brands combined
- n_eff
- with weights, the Kish effective base (Σw)² ⁄ Σw² replaces n in intervals and tests
Worked example
The synthetic tracker has 800 respondents and five brands, weighted to correct a sample with too many 18–34s (effective base 736.5). Brand A leads every stage: 90.1% aware, 55.4% would consider it and 19.7% buy it most often.
Brand C is almost as well known (81.5% aware) but only 34.4% of those aware would consider it (95% CI 30.7%–38.2%, conditional base 655), against 60.1% for the other brands combined — its biggest leak, z = -11.1, p < .001. Brand D's leak comes later: 33.2% of its past buyers bought it in the past three months, against 55.4% for the others. By age, Brand E's awareness is 64.5% among 18–34s against 33.7% in the rest of the sample.
The actions differ: Brand C needs a reason to be considered, Brand D needs to win back lapsed buyers, and Brand E's growth depends on reaching older buyers. The brands and every number here are synthetic.
Using it well
When to use it
Use it when
- Use: Brand tracking and usage-and-attitude studies that ask awareness, consideration and purchase for several brands.
- Use: Finding the step where a brand loses the most potential buyers compared with its competitors.
- Use: Comparing funnels between segments, regions or waves (one wave per file).
Not the right tool when
- Avoid: Funnels from different questionnaires or samples — stage definitions must be identical across brands.
- Avoid: Single-brand studies with no competitors to compare against — only the absolute conversions remain.
- Avoid: Reading the brand-versus-others test as exact: the same respondents answer for several brands, so the samples are not independent.
How to interpret it
Rule of thumb
Compare conversions, not stage percentages. A leak that is both significant and large (several percentage points on a base of at least about 30) is worth acting on. Answers that skip a stage (considering a brand without being aware of it) are reported, counted in stage percentages and left out of conversions. Benchmarks for what a "good" conversion is depend on the category; the tool compares brands with each other rather than with outside norms.
Questions
Frequently asked questions
What is a brand funnel?
A brand (or purchase) funnel is the share of a market at each step from awareness to purchase and loyalty for each brand — for example aware, would consider, ever bought, bought recently and buy most often — plus the conversion from each step to the next.
How is funnel conversion calculated?
Conversion is the number of respondents at a stage and the previous one divided by the number at the previous stage. Its base is the previous stage, so its confidence interval uses that smaller conditional base, not the total sample.
How does the tool decide a brand's biggest leak?
For each step it compares the brand's conversion with the conversion of all other brands combined at the same step. The biggest leak is the step with the largest shortfall; the z-test shows whether that shortfall is statistically significant. With only one brand, the step with the lowest conversion is shown.
Can I use weights and multi-select data?
Yes. Weighted percentages use the Kish effective sample size for intervals and tests. Stages can be SPSS-style multiple dichotomies (one 1/0 or yes/no column per brand), comma- or semicolon-separated multi-select columns, or a single-choice column of brand names or codes.
What if I only have the funnel percentages?
Switch to 'Enter stage percentages' and type each brand's base n and stage percentages. The tool then assumes a nested funnel, so the conversion base is the previous stage's share of the base n.
Sources and next steps
References & further reading
- [1]Lavidge, R. J., & Steiner, G. A. (1961). A model for predictive measurements of advertising effectiveness. Journal of Marketing, 25(6), 59–62.
- [2]Wilson, E. B. (1927). Probable inference, the law of succession, and statistical inference. Journal of the American Statistical Association, 22(158), 209–212.
- [3]Agresti, A., & Coull, B. A. (1998). Approximate is better than 'exact' for interval estimation of binomial proportions. The American Statistician, 52(2), 119–126.
- [4]Kish, L. (1965). Survey Sampling. New York: Wiley.
Try it on your own data — free, no signup
Upload a CSV, Excel, SPSS, Stata, SAS, R, Parquet or JSON file. Everything runs in your browser; your file never leaves your device.
Open the workspace →