1-3
Purpose, benefits, and limitations
During 2006, the House Cleaning Alliance sponsored a statistical study to derive a formula for pricing recurring house cleaning assignments. The model predicts the man-minutes required for a team of professional cleaners to clean an occupied single-family home.
The model goes beyond house size. It considers physical attributes, household usage, and execution factors that can materially change cleaning time. Its purpose is to improve estimates, help operators examine their own operating profiles, and provide a rational foundation for pricing.
The study also recognized important limits. Several usage factors require subjective scoring; correlated variables make it difficult to assign every minute cleanly to one cause; and the model produces estimates rather than guarantees.
4-5
Data and house attributes
The dataset contains 179 observations collected from July through September 2006 by a large Denver house cleaning company. Each observation represents one recurring cleaning assignment.
House attributes included finished square footage, basement area to be cleaned, number of toilets, and showers in use. These objective characteristics form the physical foundation of the estimate.
6
Usage factors
Usage factors measure the effect of the residents on the work. They include the number of inhabitants and scored observations for lifestyle, floors, and pets. Higher scores mean additional expected man-minutes; lower scores mean less work than the model's average household.
7-9
Execution factors and correlation
The study considered team size, team-leader profile, late-day work, daily workload, whether the client was home, and other conditions affecting execution. It also examined correlation among explanatory variables.
Square footage and bathrooms naturally move together. Lifestyle and clutter were also strongly correlated. Because clutter and dusting added little independent predictive power, Model II removed them from the statistical model.
10-13
The model and principal findings
Two multivariable linear-regression models were tested. HCA selected Model II because it was more parsimonious, easier to apply, and slightly more accurate after statistically weak variables were removed.
Predictive value. Approximately 87% of Model II observations had an absolute predictive error below 25%. The report identified roughly 20% error as a practical tradeoff between precision and predictive coverage.
Lifestyle and floor. A one-point increase in either factor added approximately 0.6 times the logarithm of square footage in man-minutes.
Pets. A one-point increase in the pet factor added approximately one-half of the logarithm of square footage.
Team size. A three-person team carried an estimated penalty of approximately 33 man-minutes compared with a two-person team.
Bathrooms. The original model estimated roughly 32 man-minutes for toilets with and without showers in use, although the report cautioned that the coefficients were probably biased by the available observations.
14-18
Practical application
The recurring-cleaning model supplies the foundation for weekly and biweekly estimates. The initial-cleaning calculation was intentionally described as a "better than nothing" aid because the study did not statistically analyze initial cleans.
Operators then convert predicted man-minutes to prices using their chosen hourly rates and minimum charges. The original subscription Pricer also allowed each company to adjust coefficients and operating assumptions to reflect its own scope, teams, market, and experience.
From research to working tool
The restored HCA Pricer puts the operator in control.
The working replacement retains HCA's baseline model while keeping operator-adjusted coefficients, rates, minimums, and assumptions private behind the WordPress login.