Atlas GIS Version 3: Infoworld Test Center Results

When you want to know the real truth about a product or service, go to an impartial third party -- not to the vendor. Here is how Infoworld tested Atlas GIS, and how the product fared.

Desktop mapping programs of today are highly flexible, robust applications that can be used to solve a variety of business problems. Our test plan was designed to test their overall usability and performance.

Our first task focused on a product's capability to create new maps, as well as its capability to modify map elements such as polygons, lines, and points.

First, we created a simple map from scratch by digitizing an existing map, noting the ease of establishing coordinates and selecting a perspective. We also counted the number of digitizers supported.

Then, using the existing map, we attempted to make the following changes to map elements: merge boundaries to create a larger region out of two smaller ones, reshape a line or polygon, draw a new polygon element representing a lake or a park, change the colors of map elements, attach labels to several map objects, set the viewing level of map elements and labels, crop the map and its data and then save it as a new file, change the map's projection, attach a raster image to a point on the map, lay in a background raster image and line it up to data coordinates, and create an interstate highway sign and place it on the map.

To receive a score of satisfactory, a product had to let us create map objects, including regions, lines, and points, and change the styles of those objects. If a product let us merge two regions into one larger one, we raised its score. We also gave bonus points if a product allowed us to attach raster images to maps and geocode them. We awarded extra points for ease of use and for providing additional capabilities, such as polyline smoothing and snap-to features. We subtracted points if we found the process of creating map objects unintuitive or if there was no feature that allowed us to do so.

Most of the improvements in Version 3.0 increase the program's map editing capabilities. Atlas still can't match MapInfo's capabilities in this category, but the improvements keep Atlas competitive.

Atlas now offers built-in digitizer support for tablets that use either the Windows tablet driver, Version 1.x, or the Virtual Tablet Interface, Version 2.10, another common third-party interface. This means you can now create map elements -- lines and regions -- by drawing on a tablet, which allows you to trace hard-copy maps in order to generate computerized maps.

Even more welcome, Atlas now lets you edit map regions once you've created them. Previously, you could only delete a region and create a new one. Now you can move into edit mode, then add and move nodes to reshape map objects. However, the program still falls short of the editing capabilities offered by MapInfo and Maptitude.

Although Atlas offers a snap-to-node feature for precisely aligning objects, it does not have MapInfo's capability to smooth polylines. And it still has a maximum limit of 4,096 nodes per object, compared with the more than 32,000 nodes allowed by MapInfo and Maptitude. The 4,096-node limit can be a problem in creating complex objects and importing complex maps from other programs. Atlas makes up, in part, for this problem by providing a utility for reducing the number of nodes in objects according to user specifications. (Editor's Note: The upcoming Atlas 4.0 will have a dramatically higher maximum number of nodes).

We had a few concerns about changing the style of map objects. You can use Atlas' well-designed, centralized utility for controlling map layers and the styles of objects on them, but when you change a style, the new style is applied to every object on the layer.

Atlas handles other modification tasks smoothly, such as merging two regions or creating new regions. (Atlas includes rectangle, circle, and polygon tools but no oval, square, or curve.)

Atlas doesn't allow you to attach a raster image to map coordinates, as MapInfo does, without resorting to an optional programming package. However, you can attach raster images, vector images, video clips, or any other launchable file to fields in tables. You can also attach a raster image to map points as custom symbols. Unfortunately, custom symbols must be specified in the program's symbol table, and the file must be in .BMP format.

Because Atlas only allows you to keep a maximum of 30 custom symbols, it isn't designed for attaching photos to maps. (Editor's Note: A Raster Underlay feature was added as an Atlas App to version 3.03, but is very rudimentary).

The first part of this task examined a product's basic capabilities to query and display data on maps. First, we opened a state map, a ZIP code boundary file, and a street-level map file, noting the ease of doing so. Then we attached two separate data files to the map -- one with demographic data and another with data on retail sales for a chain of electronics stores. We then performed a simple calculation on one of the map attributes (we selected all ZIP code areas where the median age of the residents was more than 27), created a bivariant thematic map, panned it, moved it to a specific ZIP code, and created a live, editable inset in the same map (using the same open data).

To look at a product's advanced analysis tools, we generated a new database file of 100 clients in the Seattle area; each database record contained seven variables. We performed a multivariable query and created a derived field. Then we merged the data file with another data file. Finally, we examined each product's capability to query external database files.

To receive a score of satisfactory, a product had to be capable of performing all the tasks listed in the previous two paragraphs, with the exception of creating inset maps and derived fields and the external SQL query; these latter tasks earned a product a higher score.

Products got extra points for superior ease of use, support of inset maps, and additional query and display capabilities. If a product could not perform simple calculations of the map attributes; create a flexible, bivariant thematic map; or perform a multivariable query, we deducted points.

In this category also, Atlas emphasizes ease of use more than it does power. Accordingly, basic data query and display operations are easy to perform, but the program has more trouble with advanced operations.

Although easy to use, the thematic tools in Atlas do have limitations. First, Atlas doesn't include any charting feature. Second, although the program displays two variables per layer, it is surprisingly limited in the combinations of themes. You can't, for example, display a ranged fill for one variable on a layer while using proportional symbols for another variable on the same layer. You can do so by copying one variable's data to another layer, however.

Atlas also provides a solid set of tools for internal database operations, though there are limitations. Specifically, whereas MapInfo and Maptitude feature relational databases, Atlas employs a flat-file internal database, so we couldn't share columns across tables, and we spent time copying data from one table to another. Apart from these limitations, we found Atlas' database features to be strong.

Atlas has the muscle for advanced database queries, too. The program handled our multivariable queries with ease and let us construct complex arguments using the mouse to enter fields, functions, and operators. Only values were entered manually.

We were able to create a derived column quite easily, with the single proviso that the column that would ultimately hold the data had to be created before we performed the calculation.

The program has built-in SQL capabilities for querying external databases.

Our next task examined a product's capability to perform spatial queries. We found named objects such as ZIP codes, counties, and highways; did a radius search on multiple data points; performed a polygon search on multiple data points; created a buffer around a highway for queries; did a multilayer query that involved splitting a ZIP code area; performed a "build-to-value" operation, expanding an area around a point until the data within the area attained a specified level; balanced sales territories; selected all areas adjacent to the selected area; and performed a point aggregation and multilayer, area-weighted aggregation, noting the methods employed by the program.

To receive a score of satisfactory, a product had to be capable of performing each of the aforementioned tasks with the exception of the multilayer, region-splitting operation; the build-to-value operation; the territory-balancing operation; and the area-weighted aggregation. These latter tasks were worth additional points. Products without ease-of-use problems received a higher score. Products earned extra points for providing additional capabilities, such as line-of-sight analysis or global positioning satellite (GPS) capabilities. Products that could not do basic queries, such as radius or polygon searches, lost points.

Atlas features multilayer spatial-query tools, which enabled it to turn in a strong performance on this task. Like the other programs we tested, however, Atlas lacks advanced spatial-query capabilities such as multilayer area-weighted analysis and topographical tools.

The program is effective at finding a named feature, such as a ZIP code or an address, although its tools are not as flexible as those of MapInfo. When you use the Find utility, Atlas only allows you to search the key column in data tables and only three key columns in geocoded files; query by value uses data from any column in any table.

The expected radius and polygon search tools are present and nicely implemented.

Atlas' buffer tools are in the same league with those found in MapInfo and Maptitude, though the program can't match Maptitude's trick of creating multiple buffers around an object in a single operation. What the program does allow you to do is determine the width of a buffer according to data attached to the object being buffered. You might, for example, specify that buffers around airports vary according to the number of takeoffs and landings each accommodates. This capability to generate data-driven buffers is not the same thing as the capability to perform build-to-value operations. (A build-to-value operation is when you specify a value -- for example, a specified number of households -- and build a buffer around each object until it contains the specified number of households.) Data-driven operations are not as useful as build-to-value operations because you don't have as much control over your data.

You can perform a build-to-value operation, however, in a roundabout way by using Atlas' point aggregation tools. Getting to the right total, of course, is a trial-and-error process, so we couldn't give Atlas a bonus for build-to-value capabilities.

We geocoded our database of 100 clients using two methods, address matching and ZIP+4. We noted any additional supported methods and checked whether the program allowed progressive passes employing different geocoding methods. To receive a score of satisfactory, a product had to have geocoding capabilities. We docked products for any limitations to their geocoding functions. Extra points went to products with exceptional ease of use or extra capabilities, such as multipass, multiple-criteria geocoding.

We examined each product's control over printing maps and other elements, such as graphs, charts, and legends. We created a map report, selectively printing map data and features, as well as formatting (including adjusting fonts and colors) and annotating the map features. Next, we tried to include in the printed report a table showing data from a radius search.

Finally, we created a chart to accompany the map, noting the variety of chart types provided and the ease of modifying those charts.

We gave a score of satisfactory if a product provided basic formatting control over printing maps, including adjusting margins and fonts. To earn a higher score, products needed to give us the choice to include tabular data in the report. Programs that provided batch reporting, report catalogs, and other special capabilities received bonus points, too. Because the capability to produce hard-copy results selectively is invaluable, we subtracted points if a product was missing this feature.

Atlas, like the others, lets you geocode not only addresses but map points for which you can provide latitude and longitude. Atlas, however, is the only program that lets you work with either traditional (degrees, minutes, seconds) or decimal notation.

Atlas also offers more controls than the others over the address-matching process; you can geocode either by street address or by ZIP+4 code. If you geocode by address, you can instruct Atlas to employ any or all of four address-matching methods in sequence: address, ZIP+4 centroid, ZIP+2 centroid, and ZIP code. If the match can't be made using the most precise method -- specific address matching -- the program will attempt to match the record using the next method.

The program also allows you to selectively relax any of five criteria -- direction designations on streets, street type, street name, house number, and ZIP code -- to increase the match rate. In short, you should be able to geocode any data by manipulating these variables.

Unlike MapInfo, Atlas can geocode addresses in multiple counties -- or even states -- in a single pass without running out of system resources, because Strategic Mapping has separated the geocoding information from the actual street files. Accordingly, you don't need to load map files to perform geocoding, and all geocode information can fit on a single CD-ROM disc.

In light of its flexible controls, it might seem ironic that Atlas turned in the fewest matches -- 68 -- on our address matching test. But our sample of 100 addresses is far too small for this to be statistically significant, and we tested using only the most rigorous address-matching method.

In this category we looked at various types of expandability, including cross-platform support, the capability to develop custom interfaces or interface elements with the application's programming language, compatibility with lower and higher end versions of the program, real-time hooks for GPS, foreign-language versions, and add-on data sets.

A product had to come bundled with a programming language (or offer one at an added cost) that could be used for building applications to earn a score of satisfactory. Higher scores went to products that provided versions on multiple platforms or interoperability with lower and higher end applications. The more add-on data sets available for a product, the more bonus points we awarded.

Atlas falls a bit behind MapInfo in platform support -- it's only available on Windows and DOS -- and it doesn't have as many foreign-language versions. But it makes up ground with its extra-cost programming languages, Atlas ScriptVB and Atlas ScriptC, which allow you to customize Atlas or generate new applications in Visual Basic and C environments.

Atlas also fits into Strategic Mapping's tiered product strategy; it is on the low end and a full-fledged, networked market-analysis database is on the high end, so you can leverage an investment in Atlas if your mapping needs extend to end-user or professional levels.

Although Atlas doesn't come bundled with as much data as Maptitude, it offers a lot. Strategic Mapping bundles U.S. state and county maps, five-digit ZIP code centroids, U.S. interstate highways, Canadian provinces and territories, and countries with their capitals and major cities marked.

Atlas also offers the broadest set of extra-cost demographic, economic, and financial data of any of the programs in this comparison, much of which is updated at least yearly. Finally, Atlas is backed by a broad array of third-party applications, maps, and data sets. (Editor's Note: ESRI has dramatically improved the bundled data set that will ship with Atlas 4.0).

This category examined the file formats supported by each program for import and export. Products that let us open ASCII, dBase, Lotus 1-2-3, and Microsoft Excel files were given a score of satisfactory. If a product did not support one of these four file formats, we deducted points. We gave bonus points for products that could import files in public-domain geographic formats, such as Tiger files.

Atlas' importing and exporting capabilities are unchanged from those of the previous version. The program uses dBase as its native file format, which means you can not only read but also directly edit any files in that format. In addition, you can import Microsoft Excel, Lotus 1-2-3, and tab- and comma-delimited ASCII files. When it comes to exporting, however, Atlas is limited to ASCII formats.

An extra-cost module is available that allows you to import and export Tiger, AutoCAD .DXF, and PC ArcInfo files. The module, Atlas Import/Export, can also import .DLG and Etak files; it lists for $795. We were impressed when we found that Atlas even supports MapInfo's file formats. (Editor's Note: Atlas 4.0 will open the ArcView SHP and MapInfo MIF formats, as well as BNA files and the native AGF format).

We loaded each product's street map for Washington state, then cropped it to display the Puget Sound region, with the state ZIP code boundaries and street maps present. We timed how long it took the product to perform six subtests: load a workset consisting of a map and two separate data files; redraw a screen after changing one map variable (the color of Interstate 5); zoom to a set level; geocode a test file; change the map's projection and save the change; and print the map. We measured the time from issuing the print command until control returned to the user.

When we had an option, we used uncompressed -- and therefore editable -- instead of compressed files. We scored the individual subtests using the following criteria, then calculated the final speed score by averaging the scores for the subtests.

We scored the open workset subtest as follows: A time of less than 20 seconds earned a score of excellent; a time between 20 and 39 seconds earned a score of very good; a time between 40 seconds and 1 minute, 19 seconds earned a score of good; a time between 1 minute, 20 seconds and 1 minute, 59 seconds earned a score of satisfactory; a time between 2 minutes and 2 minutes, 39 seconds earned a score of poor; a time equal to or greater than 2 minutes, 40 seconds earned a score of unacceptable.

The redraw screen subtest had ranges of less than 10 seconds for a score of excellent; between 10 and 19 seconds for a score of very good; between 20 and 39 seconds for a score of good; between 40 and 59 seconds for a score of satisfactory; between 1 minute and 1 minute, 19 seconds for a score of poor; and 1 minute, 20 seconds or more for a score of unacceptable.

For the zoom subtest we gave an excellent score for times of less than 5 seconds; a very good for times between 5 and 14 seconds; a good for times between 15 and 34 seconds; a satisfactory for times between 35 and 54 seconds; a poor for times between 55 and 1 minute, 14 seconds; and an unacceptable for times of 1 minute, 15 seconds or more.

The change projection subtest was scored by assigning an excellent for times of less than 2 minutes; a very good for times between 2 minutes and 2 minutes, 59 seconds; a good for times between 3 minutes and 4 minutes, 59 seconds; a satisfactory for times between 5 minutes and 6 minutes, 59 seconds; a poor for times between 7 minutes and 8 minutes, 59 seconds; and an unacceptable for times of 9 minutes or more.

We used these criteria for the geocode subtest: Times of less than 10 seconds earned an excellent; times of 10 to 29 seconds earned a very good; times of 30 seconds to 1 minute, 9 seconds earned a good; times of 1 minute, 10 seconds to 1 minute, 49 seconds earned a satisfactory; times of 1 minute, 50 seconds to 2 minutes, 29 seconds earned a poor; and times of 2 minutes, 30 seconds or more earned an unacceptable.

The print subtest used the ranges of less than 1 minute for a score of excellent; between 1 minute and 1 minute, 59 seconds for a score of very good; between 2 minutes and 3 minutes, 59 seconds for a score of good; between 4 minutes and 5 minutes, 59 seconds for a score of satisfactory; between 6 minutes and 7 minutes, 59 seconds for a score of poor; and greater than or equal to 8 minutes for a score of unacceptable.

Atlas didn't have a problem keeping pace with the fastest geographical information system applications. Atlas led the pack when it came to screen redraws and geocoding.

Atlas only stumbled during the change-projection test. It was twice as slow as MapInfo and lagged behind Maptitude.

At a minimum, the hard-copy documentation had to have thorough, accurate information on how to use the product. We gave a higher rating if it had additional in-depth information on data development and data analysis strategies, clear writing, helpful examples, quick-reference material, and useful on-line help.

The support score was based on the quality of service we received during anonymous calls to the vendor and on a vendor's support policies. We expected to receive prompt, courteous, and correct answers. We gave bonus points for technicians who offered extra information or made follow-up calls to ensure that they had fixed our problem; we deducted points for incorrect answers, unreturned calls, and long waits on hold. At a minimum, a vendor's support policies had to include some free support. We gave extra points if a company offered a toll-free support line; a money-back guarantee; information via CompuServe, fax, the Internet, or a private BBS; and access to technicians for more than 10 hours each weekday or on weekends; as well as for negotiating custom support plans.

We based our pricing score on the list price of a single-seat license, not including support contracts. The less expensive a product, the better the score. Prices of less than $400 earned an excellent; between $400 and $599 earned a very good; between $600 and $999 earned a good; between $1,000 and $1,399 earned a satisfactory; between $1,400 and $1,799 earned a poor; and $1,800 or more earned an unacceptable.

In terms of hard-copy documentation, Strategic Mapping provides an adequate Getting Started guide, an extensive tutorial manual, and a thorough reference manual. The tutorial does a good job explaining the program's features task by task. The reference manual delves into each command available in the program. The manual is clear, although the stiff writing style made reading it a bit tedious. Atlas' context-sensitive, on-line help is also extensive and well written, and its hypertext links made it easy to move from one topic to another. The program could, however, provide much better pop-up cues for toolbox icons and other display elements (Editor's Note: these indeed were improved in Atlas 3.03). We found Strategic Mapping's technical support staff to be knowledgeable and available almost immediately. When we had trouble getting through, we were kicked out to a voice mailbox and left a message; Strategic Mapping returned calls within a reasonable amount of time.

Strategic Mapping backs up Atlas with a 60-day money-back guarantee. The company also provides 90 days of free telephone support, available weekdays from 7 a.m. to 4 p.m. Pacific time. Various levels of extended support are available at extra cost, and users can also turn to the company's fax-on-demand service, as well as a forum on CompuServe and an Internet list server, for assistance or information. In addition, Strategic Mapping sends a free quarterly newsletter to all registered owners and offers training classes and application-building services for a fee. (Editor's Note: with SMI now sold to ESRI, ESRI support policies and contacts are in effect; ESRI Tech Support can be reached at 909-793-3774. Support for Atlas Select, now owned by Claritas, is available through Claritas at (800) 328-6667. Better yet, just ask us at RPM at 818-831-7607. The October 1997 price of Atlas GIS 3.03 is $795, with educational copies $250).