What Deckr Can Do

A plain-English guide to Deckr's capabilities, written for clients and prospects.

In one sentence

Deckr turns your survey data into an editable PowerPoint deck of charts — every figure calculated from the file and verified before it ships, drawn in your house style, and rebuilt the same way wave after wave, so a tracker looks the same every time it reports.

How it works, from your side

1. Upload your survey data. Drop in the export from whatever platform you used.

2. Check the plan. Deckr reads the file, works out what kind of study it is, and lists the questions it found. Untick any you don't want in the deck.

3. Add a short brief. The study name and audience. Most of this is optional.

4. Click generate. Deckr builds the deck, runs its quality check, and you download a finished PowerPoint file.

Charts-only decks

A charts-only deck is the working deck a research team builds first: every question in your file, drawn as a native, editable PowerPoint chart, each slide titled with the question that was asked. The slides follow your file's own order, whether it is a tabulated export or a raw data file. Deckr writes no commentary on a charts-only deck, so there is nothing on a slide that is not in your data.

• Question selection — untick questions on the plan screen and they are left out. A selection only removes slides: every question you keep is drawn exactly as it would be in the full deck, and the selection is stated in the methodology record.
• Chart types chosen by what the question is — rating batteries and numbered rating scales as stacked bars in scale order, tick-box questions as sorted bars, NPS as its own distribution, numeric answers as a histogram with the mean.
• Answers that form a scale stay in order — age and spend bands, household size, shopping frequency, social grade and satisfaction are drawn in their scale order, never ranked by size; 'Prefer not to say' stays at the end.
• Many subgroups as a crosstab table — Total and up to two subgroups over up to ten answers stay a chart; more subgroups or more answers are drawn as a table, Total first, each subgroup with its base and ▲/▼ beside a figure that differs from Total.
• Batteries as one clean chart — a grid of statements rated on one scale becomes one chart, not a slide per row.

A new organisation's decks are charts-only by default. Switch on written analysis for one deck on the upload page, or change your house's deck shape in Settings (or per request through the API).

Verified figures and honest bases

• Every figure is calculated in Python from your file. Top-2-box scores, NETs, means and differences are worked out from the exact (and, where weighted, weighted) shares — never estimated.
• An impossible figure stops the build. Before a deck is drawn, every figure is checked for things that cannot be true — a share over 100%, a mean outside its scale, a subgroup larger than its total. A deck with one is refused, not shipped.
• A quality check before download. Every deck must pass an automatic quality check before you are allowed to download it.
• Base sizes appear where your house style puts them. By default a chart states its base. A house that keeps bases in its tables gets none on its full-sample charts, and a house that wants one on every chart can choose that. Either way, a question asked of only part of the sample always states a smaller base, and says who it counts.
• Small bases are flagged — a base under your thresholds (30 and 50 respondents unless you set others) is starred and noted wherever the base is shown.
• Weighting — Deckr finds the weight column in your file (or you choose it) and applies it to the charts, the trend lines and the tables. The base shown is still the number of people who answered. (Two less common charts — a 0–10 question that is not NPS, and a coded market question charted as an answer — are not weighted yet.) A column that is only called "weight" but is not one is refused.
• Significance you already ran is shown, not re-run. Where your banner tables carry significance letters, Deckr shows them as ▲/▼ markers and states the confidence level your tables were tested at. It never recalculates or changes your testing.
• Nothing is dropped silently. A question that could not be charted, a category too small to show, a field excluded from analysis — each is named on the deck or in the methodology record, an optional page or appendix that states what the build did.
• Personal details stay out. Names, phone numbers, email addresses and identifiers in a raw file are excluded before analysis — never charted, tabulated or sent to a model — and named as excluded in the methodology record.

Branding and house style

• Your own PowerPoint template — Deckr builds the deck inside it, using your layouts, placeholders, theme fonts and colours. Or just set brand colours, a font and a logo.
• Learned from a reference deck — hand Deckr one of your finished decks and it learns your house conventions from it: how you show significance, how you word base footnotes, how many decimal places you use, how you order scale answers, your chart colours, your fonts and your logo placement. Deckr only adopts a convention it saw consistently, and asks you about the rest.
• Stored and shown — what Deckr learned is kept as your organisation's house style, shown on the Settings page where you can confirm or correct each item, and every deck says what each build decided and why.
• Your fonts travel with the deck — an uploaded brand font is embedded, so the deck looks right on a machine that does not have it installed.

Trackers, waves, markets and segments

• Repeatable by design. The same file with the same settings makes the same deck, and the upload page and a scripted build make identical decks (checked part by part). Your house style and question selection are stored, not retyped each wave.
• Re-run with new data — from your deck history, re-run a past deck on the next wave's file: the same structure, the new figures, with the new wave's own baseline.
• Wave-on-wave change — Deckr detects waves from a wave column in your file, or from a previous-wave file you upload, and reports the change.
• Trend lines — with three or more comparable waves, Deckr draws trend lines and marks only the wave-on-wave movements that test as significant.
• Projects — keep each tracker's waves together in a project, with its own settings where they differ from the organisation's.
• Multi-market — a raw respondent file with a country or market column can produce one overall deck plus one deck per market, delivered together as a ZIP. On a tabulated export with a market banner, a deck per market is offered as an option you tick — never made automatically.
• Custom cuts — name any set of cuts (for example "Female 25–34", or combined conditions) and get one deck per cut.
• Audience segmentation — compare results across the audiences in your banner, or get one deck per group from a segment column in your data.

What kinds of data Deckr accepts

Deckr reads the content of the file, not just its name:

• Excel (.xlsx) or CSV — one row per respondent
• SPSS data files (.sav) — including value labels and weights
• Qualtrics exports — the standard multi-row-header format
• Pre-tabulated banner tables — Decipher / Forsta tables, and the common banner layouts from other tab software (for example Quantum, WinCross and SPSS Custom Tables), with their significance letters kept
• Experience diaries (Decipher loops) — a raw Excel or CSV export where each respondent records several experiences is recognised as a loop and analysed one row per experience. Questions asked once per person are reported on respondents, loop questions on experiences, and every loop chart states both. On request, Deckr also draws the diary views: brand share of experience, touchpoint heat-maps and a touchpoint engagement map. (Loops in an SPSS .sav file are not read yet.)

It also handles real-world messiness: unusual text encodings, question wording in French, German or Spanish kept as written, a decimal comma read correctly, and survey-platform column codes.

Optional extra files: a banner/cross-tab file for segment comparisons, a previous-wave file, your PowerPoint template, and your logo.

File formats Deckr does not read today: IBM Dimensions (.mdd/.ddf), SPSS portable (.por), raw PDF or XML. If you upload one of these, Deckr tells you clearly rather than producing a broken deck.

What kinds of studies Deckr understands

Deckr recognises the type of study from your data and you can change it:

• Post-event / experience surveys — satisfaction, Net Promoter Score, return intent, importance ratings and verbatims.
• Brand tracking — the brand funnel (awareness → consideration → preference → usage → advocacy) for your brand, brand image, and competitor comparison.
• Concept & copy testing — concept scorecards (appeal, purchase intent, uniqueness, relevance, believability). A 0–10 purchase-intent question is never mistaken for NPS — NPS needs a genuine "would you recommend" question.
• Usage & attitudes — usage frequency, occasions, repertoire and attitude statements.
• Customer satisfaction / CX — correctly-scaled satisfaction, touchpoint and attribute Top-2-Box, NPS where it was asked.
• Other / bespoke quantitative — segmentation, media and one-off designs, reported as they are, with anything Deckr could not confidently classify flagged rather than guessed.

Analysis Deckr calculates when the data supports it (no AI involved):

• Pricing — a Van Westendorp price-sensitivity design becomes the four curves with the optimal and acceptable price points read off them; a Gabor-Granger ladder becomes a demand curve. If the design does not fully reconstruct, Deckr shows the raw answers and says so.
• Trade-off (MaxDiff / choice) — a counting analysis: a best–worst ranking for MaxDiff, choice shares for a simple choice design, named as a counting analysis on the slide.
• Key-driver analysis (on request) — which attributes move most with an overall measure such as satisfaction or NPS, with a priority quadrant of importance against performance. The method is stated on the slide.
• Segment pen portraits (on request) — each segment against the total: side-by-side bars, an index score, and markers on the significant differences.

Written analysis, if you want it

Deckr can add written analysis to a deck. It is optional, and the figures are the same either way. When it is on, an AI model writes:

• a headline answer and an executive summary
• a storyline connecting themes across questions
• a "so what" line on finding slides, if you want one
• recommendations, if you ask for them
• themes and sentiment coded from open-ended answers, with representative quotes

Every figure the written analysis cites is checked against the data in Python; one that does not match is removed before the deck is built, and the removal is logged. A headline never rests on a flagged small base. An organisation can switch off every step that sends respondents' free-text answers to the model.

Output formats

• A PowerPoint (.pptx) file with native, fully-editable charts
• A data-tables Excel (.xlsx) book — every question crossed by your banner, significance-tested against the total, with NET / Top-2-Box and base rows. Ask for the deck, the tables, or both.
• A ZIP of decks when you've asked for one per market, cut or segment

For teams and platforms

• Team accounts — users per organisation with roles (Owner, Admin, Editor, Viewer), projects to organise work, and access so people only see the projects they're given.
• Developer API — send a data file (or tables as JSON) and get back a finished deck through a secure download link, with a notification when it is ready.

The short version

• Charts first — a complete, editable deck of every question, in minutes, with nothing written that isn't in the data.
• Verified — every figure calculated in Python and checked before you can download.
• Yours — your template, your house conventions, your fonts.
• Built for trackers — the same deck, wave after wave, re-run on new data.
• Analysis when you want it — written commentary is an option, with every figure it cites checked.