Checklist for Dossier Finalization

Objective : To ensure Panorama analyses present visualizations with a consistent look and feel. This checklist is a reminder of items to review as analysts polish their dossiers.

This is also an opportunity to document what is done and what is outstanding.

Dossier Checklist

Requirements Fidelity

Anatomy

Cover page with content in HTML format https://github.com/ICPI/DIV/tree/master/PAWS
Chapters given short descriptive names
Pages given short descriptive name

Selectors per Page

Made use of org level selector
- Janice may need to do this for all dossiers
Selectors In the right order (not all selectors apply to all indicators or visuals) Even if they are in the org level selector, it may still be necessary to include it separately.
- Org Level
- Period
- Prioritization
- SNU1
- SNU2
- Agency
- Mech/Partner
- Special Selectors should come after standard ones so the user doesn't get confuces – example – add (Modality, Age, Sex, etc)
- Named consistently (with other dossiers)
Set to "select multiple" (as appropriate)
Located on the top or right – NOT on the bottom or left
What is outstanding – on which page?

Metrics

All metrics working
What is outstanding – on which page?

Font: Size and Color

Font: Arial

Sizes:

21point 18 point16 point 12 point 10 point8 point

Visualization Title

Font: Arial

Size: 10 – 12

Font Color – Dark Blue

Background Color – Default - white

Table/Grid Title and Body:

Font: Arial

Size: 8-10

Title Background Color: Grey

Title Font Color: Dark grey/black

Body Background Color: white/light grey

Body Font Color: Dark grey/black

Selector/Filter Title and Values

Font: Arial

Title Size: 9

Font Color = Blue 00467f

Background = White

Body/Filter Size = 8

Font Color = Dark Grey

Axis Titles/Labels and Marks

Font: Arial

Size: 8 – 10

Title can be slightly larger or the same as the labels and marks

Graph Legend Title and Labels

Font: Arial

Size: 8 – 10

Err on the smaller side – 8

Tool Tips

Font: Arial

Size: 8 – 10

Err on the smaller side – 8

Notes or Text Boxes

Font: Arial

Size: 8

Colors

Color palette selected
Color palette respected for universally defined categories (e.g. sex)
Categorical colors sufficiently different
Title Colors/Size
Body Colors/ Size
Table Color/Size

Labels and Titles

Data labels
Chart titles
Data Notes

Legends

Visible
Orientation of legend is top or RIGHT
Labels do not take up too much space

Numbers

100K instead of 100,000
Consistent use of decimal points

Symmetry

Make sure that the filters, tiles, visuals, etc are all in line and not scattered or disjointed – do this in presentation mode!

Notes Box

Analytic Questions (ideally at the top )
Notes/Instructions/Caveats
- Are notes inserted on pages where there is a need to insert an analytic question, or a data caveat, or to explain what filters were applied?
- Ideally placed at the bottom

Selector that shows which OU was selected

investigate

Spelling

Check spelling
cover page, titles, notes, labels

All OU Version (if relevant)

All metrics working
What is outstanding – on which page?

Viz overflow

Using the "Fit to contents" / "Fit to container" options, make sure that the table/viz can overflow beyond the end of screen (i.e. requires scroll bar) or not, based on your requirements.

Space saving

Where appropriate, delete X axis titles, column headers and row headers.
Where appropriate, limit the number of characters (L String function) of attributes (e.g. IM) shows in the viz or legend.

Zoom

Most users will view pages with browser zoom set at 100% by default. Make sure your layout works at this level of zoom.

Vitara charts

Formatting/style on vitara charts are different / more involved from formatting/style on MSTR viz—please take extra time to ensure that formatting/style are appropriate and complete.
Vitara charts often don't automatically add axis titles (the way MSTR viz do)—this has to be done manually.
Vitara charts sometimes set the axis origin as the min value (across all tiles of a small multiple). However, you may want to set the axis origin as 0 (otherwise the tile that contains the minimum value will seem to have a missing mark)—this must be done manually.
By default, Vitara chart legends are on the bottom and need to be moved to the top.
Vitara charts often seem to render text at a different (smaller) scale than MSTR viz. So if the standard axis title font size in MSTR is 9, you may need to make it 10/11 in Vitara.

Viz page overflow

Using the "Fit to contents" / "Fit to container" options, make sure that the table/viz can overflow beyond the end of screen (i.e. requires scroll bar) or not, based on your requirements.

Outstanding Items Comments/Notes Page by Page

Insert comments
Insert comments

Engineering Checks for Dossier Review

Selectors (page filters)
- If a page contains multiple viz (or tables), make sure that all selectors target all of the viz on the page (or all the viz they are supposed to target)
- Check to see if added Selectors (selectors beyond the standard set) contain irrelevant choices that could confuse / distract the user. For example, the age band 18+ is only relevant for OVC, and is not pertinent for a page on HTS. Such choices can be filtered out with a chapter filter—but be sure it is not needed in any page in that chapter.
- Verify that all selectors are set to "Show option for All" and "Allow Multiple", unless requirements demand otherwise.
- Leave all filters in the state that you wish to be the "default" view for the user
Level selectors
- Org Level:
- Add the level variable as chapter filter to filter out Org Level = NULL
- Add the level selector as a viz filter with Does Not Contain=Dedup
- Age or Sex level filters:
- Add the level variable as chapter filter to filter out Org Level = NULL
- Add the level selectors checkbox "Include" viz filters to include only valid options (i.e. don't check NULL, N/A, anything else you don't want to show up)
- Try each option of every single level selector in the dossier to ensure that it produces the desired options (and no undesired options)
- Leave all level selectors in the state that you wish to be the "default" view for the user
Chapter filters
- Carefully examine chapter filters to make sure that they are valid / appropriate for all the pages in the chapter and don't inadvertently impact any single page.
Influence of the build OU
- Be aware of the limitations of the data of the OU that you used to build your dossier. That OU might not report a disaggregate that other OU do report. Not accounting for this disaggregate in your design could cause problems.
- Filters that are based on values that could change based on the OU should be carefully verified. If the build OU didn't report a disaggregate (e.g. Post ANC1), during the build that value won't show up to be included/excluded in a Modality filter. However, that value will appear (as unchecked—whether unchecked means include or exclude) in the filter if you switch to an OU that did report the disag.
Copying / importing
- If, during the development process, you created new pages/chapter by copying and then modifying existing pages/chapters, double check all chapter and viz-level filters to make sure there are no "vestigial" elements that are no longer appropriate to the new chapter.
- If you imported pages from other dossiers, you may have lost the chapter filters that supported the page in the original dossier. Verify that all appropriate chapter filters are present.
- If you moved pages from one chapter to another, of course any chapter filters that previously supported the page are lost.
Dataset cleanup
- Delete any calculated metrics that you created but are not using.
- Inform the systems team of any system metrics/attributes that you are not using and can be removed from the dataset.
Targets/Results: Be sure that you are only showing the data you intend
- If you are using a system metric, confirm whether it is created with Results or Value (targets & results)
- As appropriate, add chapter and viz-level filters to ensure that only the intended data are shown
Disaggregates
- Most system metrics are not filtered to a specific disaggregate, and will return data for all disaggregated. For example, HTS_TST would return data for the age/sex/modality disag, the KeyPop disag, and Total Numerator. The same goes for Numerator/Denominator. This is a frequent source of errors.
- To avoid double counting, apply a chapter filter (ensure no adverse effect on any other page in the chapter), viz-level filter, or have one of the disags in the viz itself (e.g. have HTS stratified by sex in the Viz.
Axes
- Consider if any of your viz do / could have outliers that could skew the scale and make the viz hard to read. If so, consider countermeasures, such as a quantitative filter or a axis max (documented in the page note).
- Consider if any of your viz have a narrow data range (e.g. VL suppression is rarely \<30%) that justifies narrowing the range of the axis scale to more easily differentiate the marks.
- For stratified viz, consider if any of your axis scales should be allowed to be independent by row. Depending on the analysis, this could be useful if large differences of scale are anticipated among strata (e.g. TX_NEW by \<15/15+) and the analysis question does not entail absolute comparison between strata (in the TX_NEW example, I'm probably not trying to compare the number of \<15 v 15+ TX_NEW—I'm likely more concerned with comparing TX_NEW among sex or IM within each age strata
- This also holds true for Small Multiples, which can be considered a sort of strata. If your small multiples show something that you want to compare across tiles (e.g. Linkage, or Volume) you probably want the axis scale to be standardized across all tiles so the eye can quickly scan across them to identify lower results / lower volume.
Changing age/sex bands
- Verify that viz employing age/sex stratifications are not inadvertently affected by changing age/sex bands.
- This problem is most common when showing absolute numbers (including retention proxies) by age or sex. For example, a trend graph of TX_NEW by sex will see an artifact increase with the introduction of sex disaggregation for \<10.
- Similarly, age/sex groups could be included/excluded from calculated indicators. For example, a trend graph of Linkage by sex won't include \<10 before 19Q1 but will include \<10 in 19Q1 and after.
Periods
- If possible, always use relative time attributes (e.g. Years Ago, Quarters Ago) for filters and calculated indicators, instead of fixed time attributes (e.g. Period, Fiscal Year).
- To introduce specific calendar time limits, use "Exclude", not "Include." For example, an indicator was modified in 19Q2, and you want to build a trend graph for this modified indicator. However, you don't want the graph to stretch back before 19Q2 (before the definition changed). You can't use relative time attributes here (if you set Quarters ago to 2, when Q4 comes the graph will only show Q3 and Q4), so you have to filter on specific periods. This should be done using Exclude (exclude all periods before 19Q2). If you use Include (include Q2 + Q3), when Q4 arrives it will not be checked as Include in your filter.
Calculated metrics
- In the Formula Editor, verify that all calculated metrics (e.g. created by you the developer at the dossier level) end with a "+" to allow the metric to conform to any applied filters.
Site
- Verify that non-site data (e.g. Military (Site="N/A")) are excluded
- Verify that community sites are / are not included as intended.

ICPI / DIV

Pano Style Guide and Checklist for Dossier Finalization #143

Style Guide and Checklist

Checklist for Dossier Finalization

Dossier Checklist

Engineering Checks for Dossier Review