Survey results in Power BI, already weighted

Each project gets its own feed with a fixed structure. The fact table Results holds one row per question, answer, breakdown value and wave. Questions, Breaks and Waves supply the labels and sort order, and Meta records when the feed was last built.

Every result row carries the weighted percentage of base (PctOfBase), the weighted numerator and base, the unweighted base, the mean for numeric questions, and the wave test result as a number and as an arrow (▲ or ▼). Breakdowns and waves each have a total member with the ID T, so you reach the full sample without building your own sums.

Individual respondents’ answers are not in the feed. Power BI reads only aggregates that DataLion has calculated and stored in advance.

• Results: percentages, bases, means, significance
• Questions: question label, kind and order
• Breaks: breakdown variables and their values
• Waves: waves with start and end dates
• Meta: build time and cell count

Survey data follows rules a BI model doesn’t know about, weights and correct bases above all. With the BI feed, DataLion applies those rules before Power BI sees the numbers. In the feed settings you decide whether the project’s weighting applies, whether means are included and which alpha the wave test uses. The default is 5%.

One rule then applies in Power BI: PctOfBase is a finished percentage and must not be summed. For a percentage across several cells, divide the sum of NumeratorWeighted by the sum of BaseWeighted. Show the significance mark straight from SigMark.

• Weighting: as set in the project, always or never
• Wave test: two-sided z-test against the previous wave, adjustable alpha
• Means: optional, for numeric questions

You enable the feed per project. It stays off until an administrator switches it on and builds it for the first time.

Power BI signs in with an API token. In the sign-in dialog, choose Basic, enter any user name and use the token as the password. Treat the token like a password and revoke it under API access once you no longer need it.

• Under Admin → Projects → Settings → BI feed, select questions, breakdowns and waves. An estimate shows how many cells the selection produces.
• Enable the feed, save and choose Rebuild now. Once the status reads “ready”, the first version is available.
• Under Admin → Integrations → API access, create a token with the “Power BI” preset and copy it right away. DataLion shows it only once.
• Copy the OData service URL from the BI feed page.
• In Power BI, choose Get data → OData feed, paste the URL and load all five tables. Then relate Questions, Breaks and Waves one-to-many to Results.

DataLion rebuilds the feed whenever its basis changes: after a successful data import, after changes to the codebook or the waves, and when you save the feed settings. On top of that, every enabled feed is rebuilt each night. When a new wave arrives in a tracking study, DataLion calculates only that wave and keeps the rest.

Each build writes a new version that goes live only once it is complete, so Power BI never reads a half-built state. If a run fails, the last complete version stays available and the BI feed page shows the error.

How often Power BI fetches the data is set in its scheduled refresh. The Meta table holds the time of the last build. Show it in the report and every reader can see how current the figures are.

Only tokens belonging to an active administrator account can read the feed. Access profiles, which limit what someone sees in a dashboard, do not apply to it. The feed contains every result you selected for it. If different readers should see different data in Power BI, set that up in Power BI.

Of the OData query options, the feed evaluates only $count and ignores $filter and the rest. Power BI therefore loads every row on each refresh, in pages of 50,000. Power BI’s incremental refresh needs server-side filtering and does not work with the feed.

Open-text answers, MaxDiff, conjoint and advanced-statistics outputs are not in the feed. Small bases are not suppressed; use BaseUnweighted to hide them in Power BI. The wave test compares consecutive waves, not groups within one wave. Each project also has a cell budget, two million cells by default.