I’ve been meaning to write up a good practice guide for Business Objects developers for some time now.
I have finally managed to find some time to put together a list of tips I’ve used over the years. I hope you find the following list of tips useful in some small way.
If, after reading this article, you come away with something new, then I’ll be very happy.
I’m going to keep this post going for a while, adding to it as and when I can. But more importantly, I’d love for you to share your developer tips. I know there is a wide community of Business Objects experts out there, so please help develop this list so we can all benefit from our shared experiences.
If you’d like to share a tip or two, please add a comment to the post. I’ll then add it to the main list and give you a name check!
Universe Designer Application
• Use business terminology for all universe objects – always. The universe is the semantic layer between the business and IT.
• Wherever possible, and certainly for measures and calculated objects, add a description to the object.
• Ensure your object is formatted correctly. Pay particular attention to dates and numeric fields.
• Create filtered prompts wherever possible. Anything which can help the user get their data more efficiently helps. Remember to allow the selection of all data.
• Ensure your List of Values (LOV) is sorted. Don’t assume it will be.
• Don’t create a LOV on a measure.
• If you have to create a LOV on a dimension with many data items, consider grouping those items.
• Group your universe objects into a logical order. Create something that the business will understand and be familiar with.
• Create different time periods – users like to compare current year to date to last year to date, quarter to date, month to date, etc.
• Contexts – these can be incredibly useful. But they can become terribly confusing to the end user. If you really have to put in a context, do so, but make sure it has a meaningful name and description. If your user is presented with a context prompt box, it should be absolutely clear to them which context to choose. If it isn’t, you need to revisit your context or approach. The purpose of the tool is to help the end user get their results. Remember that.
• For fields based on free form input text, consider trimming the field. Profiling your data beforehand should help you identify these fields.
The Query Panel
• Only use objects that you are going to use in your report.
• Put the objects in the order that you would like them sorted on the report if possible.
• Move your measures to the end.
• Use pre-defined (universe) filters if available.
• Avoid hard coding values wherever possible – use prompts instead.
• If you are not familiar with your data set, restrict the number of rows returned by your original query. If you are happy with the result set, you can remove this restriction. The last thing you need is to run the mother of all queries.
• Make prompts meaningful. If you are prompting for a date, include an example of the expected format. If a user can enter a ‘*’ to select all values, say so.
• Get the server to do the work rather than the report. Not all end user PC’s are high spec.
• If you are going to use multiple queries, give each query a meaningful name.
• Create a report header tab/page – this should list the report title, parameters used, descriptions, values entered at the prompts, last refreshed date.
• Ensure the report opens on this header page. This should result in a faster opening report.
• Keep the report look and feel the same throughout – table positions, headers, fonts and colours.
• For tables that span across several pages, repeat the header on every new page.
• Dates – who is your target audience? US and UK data formats differ. If you do not know who your audience is, or if it is a worldwide audience, specify the format of the dates in the header page of the report or the column heading.
• Decimal places – how many decimal places do you need to report on? How many decimal places are used in the calculation?
• When working with large numbers with several digits, consider dividing by a thousand or a million. Ensure the report states that you are reporting in that way.
• Include error handling in any calculations performed on the fly. Pay special attention to DIV/0 errors. Use the IsError function. This post may help.
• For all new formulas you create, consider putting them into a variable to allow reuse and easy maintenance. Prefix the variable name with v_ to make them easily identifiable.
• Include a cell at the bottom of your report to show the last refreshed date.
• Build in a very visible partial refreshed alert. All too often that yellow triangle is overlooked by users. This post may help.
• Number your report pages. ‘Page n of n’ works well.
• Give the report a meaningful name. Include a description. Keywords are helpful too for searching.
• Consider a numbering scheme for your reports. E.g. first digit represents dept/function, second digit represents area, etc. E.g. 104 – Sales Report London, 105 Sales Report New York, 204 – Finance Report London. This will allow rapid identification of a report series.
• Use autofit width or wrap text for longer fields. If you only want to display the first n characters of a field in a report, create a formula to do so (e.g. Left(,20))
• Be consistent with your fonts.
• Use a sensible font like Arial. Some of the more fancy fonts do not print too well, especially if a smaller size font is used.
• Consider how your users will use the report. If they tend to print reports, and they do not have a colour printer, avoid colour on your report. Colours on report don’t lend themselves well to output on a black and white printer. If you must use shading, consider greyscale. If you users only view reports on screen, colours can help identify sections of data.
• Use alerters. These are a great tool for bringing data to the users attention. Given them a meaningful name and description.
• Optimise your report for speed – reduce or eliminate unused formulas, alerters and formatting. Keep it simple.
• Consider the use of section breaks to break up large tables. These can be easily searched in the side panel.
• Before publishing or releasing a report to production, do a print preview. Does the report print out as you’d expect? Do you need to save the report in landscape. Also save to Excel. Are any fields appearing incorrectly formatted?
• Big reports – Do your users really need that 100 page report? Really? If all they are after is a data dump, a more efficient way to deliver this is to write a piece of SQL code and save the result as a CSV file. But always ask why they want a data dump in the first place. What they do with the data? BOBJ is a powerful tool. Show them how to do what they want to do. Use it.
• Save your work regularly. You never know when the power will go, or an application will crash. If you have spent the last couple of hours working on a report, it’s not going to hurt to hit that save button. I normally save every significant change as a new version. Once I have a report ready for production, I’ll delete those versions.
• Images such as a corporate logo can be inserted into reports. But keep the images small. Use a file format that compresses the image to a sensible size. Don’t even think about inserted a large 1mb image on every page. This post may help.
• Make use of multiple tabs in the report. However, avoid hard coding date values in these tabs. You will have to maintain these later.
• If you have a particularly complex report, add a tab that provides the detail behind the report. This will aid future developers.
• If you have one set of users that simply love to play with data, and others who want a formatted report, consider adding a Data Dump tab at the end of your report. This should be a simple table with zero formatting.
• Prior to release to UAT/Production, delete any unused variables and formulas.
• If running multiple queries, ensure the relevant dimensions are linked.
• Purge your report before release to production. The data may be obsolete by the time the user retrieves the report and it will be bigger than it needs to be.
• For corporate reports, consider developing a standard report template
• If your company reports in several currencies, ensure the measures on your report are clear as to what currency is being used.
• Avoid falling into the trap of creating several versions of essentially the same report. Remove hard coding and make the report as generic as possible. Use prompts. If a user wants a new field, can it be added to an existing report or do you need to create a duplicate with the new field?
• If you use graphs, keep them simple. Avoid 3D graphs. Read a good book on Data Visualisation to understand how to deliver data in a visual format. This post may help.
• If you have multiple tabs in a section, ensure that they are positioned relative to each other. Same applies to floating cells. Ensure they are relative to something so that they don’t overlap with other objects on your report.
• Percentages should be calculated within the report to ensure it is in context.
• If you create a grouped variable, you need to check that all possible values for that dimension are grouped. If not, the next time you run that report, you may find data that is not sitting in a group. This could lead to misleading results.