Lotus Improv

Lotus Improv
Company type	Software application
Industry	Software, Spreadsheet software
Founded	1991
Founder	Lotus Development
Defunct	April 1996
Fate	Discontinued after IBM acquisition of Lotus
Products	Spreadsheet application
Parent	Lotus Development (subsidiary of IBM from 1995)

Lotus Improv was a spreadsheet application developed by Lotus Development, initially launched in 1991 for the NeXTSTEP operating system and subsequently released for Windows 3.1 in 1993. Development ceased in 1994 due to underwhelming Windows sales, and the project was officially terminated in April 1996 following Lotus's acquisition by IBM.

The primary objective of Improv was to transform traditional spreadsheet usage by streamlining the processes of creation and modification. Unlike typical spreadsheets that combined data, formulas, and annotations within individual cells, Improv distinctly separated these components—cells contained only input or output data. Formulas and macros were maintained externally to cells, which minimized errors and facilitated easier adjustments. Additionally, it utilized named ranges for formula references instead of conventional cell coordinates.

Although Improv did not compete effectively in the mass market against dominant products like Lotus 1-2-3 or Microsoft Excel, it found a dedicated user base in specialized fields such as financial modeling. Its pioneering design influenced niche industries and led to similar software offerings, including Lighthouse Design's Quantrix.

Apple Inc.’s Numbers application integrates a comparable approach to formula management and naming conventions within a standard spreadsheet environment.

Background

The earliest spreadsheets were paper-based grids designed primarily for accounting tasks, featuring cells for entering data and displaying formula results. Berkeley academic Richard Mattessich championed the use of spreadsheets for financial modeling but observed that recalculations could take several days. In 1964, he suggested employing computers to instantly update calculations whenever inputs changed.

In business education, spreadsheet modeling was taught using chalkboards that simulated paper worksheets, enabling straightforward error correction and sharing. This concept was digitized in 1979 by Dan Bricklin, who created the first electronic spreadsheet, VisiCalc, launched on the Apple II. Its widespread popularity sparked numerous imitators.

Among these was a significant competitor developed by former VisiCalc programmer Mitch Kapor, who created Lotus 1-2-3. Tailored for the IBM PC platform, Lotus 1-2-3 sold 60,000 copies within its debut month, rapidly establishing Lotus as a leading software company worldwide.

ATG

In 1986, Lotus formed an advanced technology group aimed at simplifying spreadsheet setup processes. While finalized spreadsheets were generally user-friendly, many users found it challenging to design initial layouts—deciding whether to organize data by rows or columns, whether intermediate values should be stored alongside input or on separate sheets, and estimating the necessary workspace.

Developer Pito Salas addressed these challenges by examining practical examples and identifying that data, views, and formulas represented distinct elements. Traditional spreadsheets combined these aspects within single sheets, making it difficult to discern whether cells contained inputs, intermediate calculations, or outputs. Salas introduced a novel spreadsheet structure that separated data entry, formula logic, and output views into different components. This allowed users to categorize data based on purpose using names rather than relying on fixed cell locations. Consequently, moving data did not disrupt associated calculations.

Salas also highlighted that existing spreadsheets had limitations regarding output views. The intermingling of inputs and computations made layout modifications risky since they could affect results. By segregating these elements, users could create flexible views without triggering recalculations. For instance, summarizing monthly sales figures by year could be achieved simply by adjusting the view instead of inserting new formulas or additional sheets.

Back Bay

By mid-1986, Salas had created a demo resembling a slideshow called Modeler for the IBM PC. In February 1987, Glenn Edelson joined the effort to develop a C++ version. This concept demonstrated considerable promise for financial modeling applications. Bonnie Sullivan was responsible for drafting project requirements, while Jeff Anderholm conducted an analysis of the financial software landscape. During demonstrations in the summer, financial firms provided highly favorable responses.

Lotus gave the green light to implement Modeler in September 1988. After reviewing various platforms such as DOS and Macintosh OS, they ultimately selected OS/2, which at the time was regarded as a commercially viable system. The initiative was internally named "Back Bay," inspired by Boston’s Back Bay neighborhood, with a mascot called Fluffy Bunny.

In October 1988, Steve Jobs visited Lotus to present the newly introduced NeXT computer. Impressed by the Back Bay project, he encouraged development on NeXT hardware. Despite enthusiasm for NeXT, Lotus persisted with OS/2 development, contending with its unstable Presentation Manager interface and immature platform issues that impeded progress.

NeXT release

After encountering difficulties with OS/2, the team resolved in February 1989 to transition Improv to NeXT. Steve Jobs endorsed this decision by sending flowers and assigning Bruce Blumberg, a NeXT software specialist, to support Lotus engineers with NeXTSTEP. Dividing the back-end into C++ and the front-end into Objective-C simplified debugging efforts. The use of NeXT's Interface Builder sped up user interface experimentation, allowing swift product iteration.

During an April 1989 visit, Jobs advocated for a more user-friendly categorization method, resulting in the creation of category "tiles"—icons that enabled quick rearrangement of output sheets. Jobs remained a proactive supporter, consistently motivating enhancements. Blumberg continued providing technical assistance during the release of NeXTSTEP 2.0.

Improv for NeXT debuted in February 1991 to widespread acclaim from leading computer and mainstream business media. It became recognized as a killer application for NeXT systems, significantly boosting sales among financial institutions. This achievement secured a lasting foothold for NeXT in finance sectors throughout the late 1990s, even after NeXT was acquired by Apple Inc..

Windows release

Following the success on NeXT (codenamed "Black Marlin"), Lotus pursued versions for Windows ("Blue Marlin") and Macintosh ("Red Marlin"). Due to substantial differences in APIs and programming languages, porting proved difficult. Lotus Improv for Windows v2.0 was released in May 1993 for Windows 3.1 and received praise for its user-friendly design.

Despite favorable critiques, sales on Windows were sluggish. In March 1994, Lotus repositioned Improv as an add-in for 1-2-3—capitalizing on its ability to import 1-2-3 files—yet this failed to boost sales. After a minor update to version 2.1, development halted in August 1994. The product was officially discontinued in April 1996 following IBM’s acquisition of Lotus.

After Improv

The disappointing sales and eventual cancellation of Improv on the PC platform have become emblematic examples of software market missteps. While limited adoption on NeXT was attributed to its small market share, failure on PCs largely resulted from internal resistance within Lotus. Sales and marketing teams, accustomed to promoting 1-2-3, found it difficult to position Improv effectively and instead concentrated on their established offering. Furthermore, Microsoft Excel’s inclusion in Office at an affordable price point combined with early shortcomings of Improv—such as absence of macro programming and undo features—impeded wider acceptance. Joel Spolsky attributed Improv’s downfall to its design being too narrowly focused compared to Excel’s broader appeal.

Although Improv disappeared during the 1990s, it continues to influence discussions and literature concerning spreadsheet innovation. Its withdrawal inspired imitators like Lighthouse Design’s Quantrix—a near replica aimed at financial users—that ultimately failed following Sun Microsystems’ acquisition of Lighthouse Design.

Concepts

Improv introduced a novel approach by distinctly separating data, views, and formulas into individual elements. The spreadsheet itself contained solely the raw input data, organized and labeled according to specific categories. Instead of relying on cell references, formulas pointed to these named data sets, allowing the flexible generation of numerous views—whether spreadsheets or graphical charts—without any fixed limit.

To illustrate, calculating overall sales involved labeling columns as "Unit Price" and "Unit Sales," then establishing a formula such as "Total Sales = Unit Price multiplied by Unit Sales." When a "Total Sales" view was added, the system automatically presented the computed results, interpreting the formula’s purpose seamlessly.

The categorization framework enabled data aggregation by dimensions like product categories or years. Users could interactively rearrange these groups to swiftly modify the displayed views—a concept that later became widely recognized through pivot table functionality in other applications.

External links

Review from 1993