Factors Influencing The Precision of Estimates for Tester-Developer Ratios

There are many authors that tend to analyze the factors influencing the accuracy of tester-developer ratio estimates. As a result, we can get familiar with the key points for using these ratios, represented in the form of rules:

1. The combination of inconsistent measurement systems and the existing differences in the ratios lead to a margin of error in the resultant ratios values on the order of 10x. Consequently, it is useless to compare the ratio of your project with any other ratio. The impact of the measurement system cannot be analyzed separately from the existing differences in the ratios.
2. If the proposed value of the ratio differs markedly from the actual industry values by order of magnitude, the comparison may prove useful. For example, the suggestion that the entire department should work at a ratio of 1 to 30 in all projects is easily countered by referring to the reported industry ratio values. Do you know that qa consulting companies look to help people manufacture high quality software? If you are unsure of successful release of your project it makes sense to address distinguished quality assurance consultants and get their authoritative opinion on this issue.
3. The ratios that were used in past projects can be applied under the following conditions:

• A consistent measurement system is being used.
• There is almost no information regarding the intended functionality of the system (otherwise the cost estimates, for example, based on a work breakdown structure will be much more accurate.
• The obtained ratios are adjusted subject to known differences between projects.
• A margin of error of 25-50% is considered acceptable.

If you want to make an initial cost estimate based on the ratio of testers to developers, the papers such as “Estimating Tester to Developer Ratios (or Not)” by Kathleen A. Iberle and Susan Bartlett’s and “It Depends: Deciding on the Correct Ratio of Developers to Testers” by Johanna Rothman provide enough information to help you perform the task intelligently.

However, it is senseless to use only one approach. Combining the empirical rules of J. Rashka, J.Paul, Dustin with Rothman’s risk-based methods and ratio-affecting-factor model engineered by Iberly and Bartlett, rough estimates can be refined. By making various estimates using a variety of empirical rules, including obtaining average estimates, you can improve the accuracy of the final estimates. The difference between the models gives the best-case and worst-case estimates. Nevertheless, remember that if you need more accurate estimates, in all likelihood, the best option is to use work-breakdown structures.