University of Liverpool, Department of Architecture & Building Engineering

Citation

Walker, T. (2001), "University of Liverpool, Department of Architecture & Building Engineering", Property Management, Vol. 19 No. 5. https://doi.org/10.1108/pm.2001.11319eae.001

University of Liverpool, Department of Architecture & Building Engineering

University of Liverpool, Department of Architecture & Building Engineering

Karl Anthony BlythRegistration date: 1st October 1997Expected submission date: Summer 2001

A computer model that forecasts the cash flow of construction projects using stage payments at the tender stage

Cash flow forecasting is an indispensable tool for construction companies and is essential for the survival of any contractor at all stages of the work. Cash flow forecasting at the tendering stage needs to be simple and fast, considering the short time available and the associated cost. A multiple linear regression computer model that forecasts the cash flow of building projects for both the traditional monthly valuations approach, and via a stage payment approach was produced. This involved standardising the programmes of work for 50 building projects in order to compare them more accurately. A total of between 20 (one storey building) and 39 (seven storeys) standardised activities were produced, along with their set sequence and established activity dependencies. Multiple linear regression analysis of the sample data shows that the costs, durations, start and end dates of the pre-determined standardised activity groups, can be predicted from a combination of the most effective interval level and non-interval level data variables. An 85 per cent minimum accuracy when comparing actual with predicted values was achieved in each case. The maximum absolute percentage errors for any project, for overall cost, duration, and time lags, were 5 per cent, 11 per cent and 9 per cent respectively. A mean SDY of 2.4 was produced for the cumulative cost curves. The model was tested successfully on a further six projects that were not used in the initial regression modelling. The maximum absolute percentage error for the six test projects, for overall cost, duration, and time lags, were 8 per cent, 14 per cent and 15 per cent respectively. A mean average SDY of 3.885 was produced. The results are very reliable despite, as expected, being slightly less accurate than the data used in the development of the model. The 85 per cent minimum accuracy was still maintained. The results were compared with those from previous models and proved more reliable.

Keywords: Cash flow, financial planning, programme of works, s-curves, stages/milestones, standardisation

Description of research

The report "Constructing the Team" by Sir Michael Latham (1994) contained some important proposals concerning tendering, contractual issues and the current, traditional monthly valuation approach. It suggested that it would be a more favourable system of payment and valuation if payments were based upon stages or milestones. The majority of contractors' concerns relate to cash flow and how interest on cash could be obtained. Stage payments will affect cash flow and if they become an acceptable form of payment, contractors will have to reassess their margins. A questionnaire was constructed and sent to companies in four areas of the building industry i.e. contractors, clients, quantity surveyors and consultants. The purpose of this was to determine whether stage payments were becoming more frequently used and whether or not their use was supported. Results were analysed and response patterns are discussed below.

A more recent report entitled "Rethinking Construction" by Sir John Egan (1998) built upon these initial proposals by Latham, and contained suggestions on delivering improvements in construction quality and efficiency, including integrating the project processes via standardisation. As a response to the findings of this report an innovative framework that proposes a standardised programme of works for construction projects, with characteristics generic to most building construction programmes, and with similarities to the concept of concurrent construction, was developed. An initial sample of 50 buildings, collected via interviews, was analysed to assess the existence of similarities and repeated operations in each construction project. A minimum of 20 standardised elemental options for a single storey building, increasing to 39 for a seven-storey building, were developed. A logical pre-determined sequence of activities, including the respective dependencies based on both analysing each programme of works and expert knowledge, was produced.

A computer model that predicts both overall project and activity cost and duration, and start and end dates of the activities (based on time lags) was produced using multiple regression techniques. The model is based upon a number of pre-determined project characteristics, identified via a questionnaire completed for every project obtained. The results show that the model can produce reliable estimates of these factors, and hence can predict the whole programme of works for the building projects involved, and can be compared accurately to the original.

The model can therefore also predict the contractor's cash flow of these building projects. Cash flow forecasting is an indispensable tool for construction companies and is essential for the survival of any contractor at all stages of the work. Cash flow forecasting at the tendering stage needs to be simple and fast however, considering the short time available and the associated cost. Contractors rarely prepare a detailed construction plan at this stage, and usually wait until winning the contract. Therefore a simple and fast technique of forecasting cash flow quite accurately is required. The majority of existing cash flow forecasting models have been based on standard value S-curves, developed using historical data. The model proved to be more accurate than existing models.

Two cash flow models were produced. The first cost and valuation model was created from these results, based on the traditional monthly payment approach. The second stage payment model was created in conjunction with the first and uses an interface in which any combination of stages can be used to suit individual project needs. These were combined and linked with MS Project to complete the model.

Description of methodology

Data collection and stage payment questionnaire

A structured letter and questionnaire was sent out to over 400 building companies asking for the programmes of work for any recent projects. The accompanying questionnaire consisted of 14 multiple-choice questions regarding the characteristics of the project e.g. cladding type, service intensity, site access. A total of 56 programmes of work for recent projects from a combination of interviews and gathering of historical data published in Building Magazine, were obtained. The data encompassed 11 building types including commercial, retail, and industrial projects, ranged from one to seven storeys, and took between seven and 33 months to complete.

A multiple-choice stage payment questionnaire survey was constructed and sent to 180 companies including: clients, contractors, consultants and quantity surveyors. Overall there were a total of 27 stage payment-related questions and statements to answer e.g. frequency of payments, effect on cash flow. For each of the questions there was space left for additional comments if necessary and these were taken into account in the analysis. The purpose of the questionnaire was:

1. a.

   to learn more about stage payments and whether they are becoming more frequently used;

2. b.

   to see if the industry agrees with the advantages of stage payments as proposed by Latham; and

3. c.

   to see what the industry thinks about the rest of the relevant Latham proposals.

Out of the 41 per cent response rate, the results showed that out of the nine proposed advantages of stage payments, the respondents only agreed with three of them e.g. they are cost effective, they spare the contractor costly measurement of incomplete and unfixed materials. With regards to the other Latham proposal, the respondents disagreed strongly with four out of the seven proposals e.g. incentives for exceptional performance, reforming the JCT 80 format to keep up with industry change. An overall total of 42.9 per cent of the respondents said that stage payments had no effect on cash flow, compared with 35.7 per cent who said they had a moderate adverse effect and 21.4 per cent who thought they had a favourable effect.

Developing the models

Standardising activities, sequences, dependencies. Fifty buildings encompassing a total of 11 different project types were surveyed and analysed. The other six were reserved for testing. The sample was then investigated further to assess the existence of similarities and repeated operations in each individual construction project. Each building was compared with every other to determine the existence of similarities, and hence look for stereotypes. The philosophy behind this survey was to elicit knowledge from the sample, and subsequently to summarise and attempt to standardise the same construction activities for any building. A logical pre-determined sequence of activities, including the respective dependencies based on both analysing each programme of works and expert knowledge, were produced. A total of between 20 (single storey building) and 39 (seven storey building) new standardised activity groups were produced i.e. site set-up, foundations, to external works, clean and hand-over.

All of the costs, durations etc. were amalgamated and classified into these standardised groups. The start dates were determined by calculating their respective percentage time lag from their relevant dependent activity predecessor. Each activity was dependent on a previous activity in the programme sequence and hence could not start until the earlier one had started. There were to be no negative lags. This produced 50 new programmes of work, based on actual data. Each of the activity costs for all of the projects were grouped i.e. table one would consist of all activity one (site set-up) costs, table two for all activity two (foundation) costs. This was repeated for all the duration and the time lags.

The regression data. The 14 project characteristics obtained via the questionnaire would be used as data variables, in the development of a multiple linear regression model. A further eight characteristics were produced based on the initial 14 to increase the amount of variables, and perhaps increase the accuracy of the regression. The non-interval level data e.g. project type and location were assigned numbers, based upon a ranking system, determined by the popularity of a given category. A total of 16 computer and manual regression methods, using both Excel and SPSS, were developed to see which were the most accurate and produced the best results when compared to the originals. The chosen regression method involved eliminating outliers and modelling unique project types separately. The analysis of the data was completed in Excel.

For every regression method, the accuracy of the equations produced were tested in percentage error terms when compared to the original values. Anything above an absolute error of 15 per cent was neglected and deemed too inaccurate. Multiple linear regression analysis of the sample data shows that the costs, durations, start and end dates of the pre-determined standardised activity groups, can be predicted from a combination of the most influential data variables. An 85 per cent minimum accuracy was achieved in each case. The maximum absolute percentage error for any project, for overall cost, duration, and time lags, were 5 per cent, 11 per cent and 9 per cent respectively.

To test further the accuracy, the predicted S-curves were calculated manually using these equations and compared with the actual, for each project. The initial results of each method were also visually tested based upon practitioner's knowledge of S-curves. A huge difference in shape and closeness suggested that the results were not accurate enough, and further methods were produced. The equations were deemed effective and utilised when there was less than an absolute error of 15 per cent and the S-curves were very close to each other.

Constructing the formula templates for the traditional and stage payment models. All of the final 90 regression equations obtained were placed systematically into a linked structured formula template in Excel. The two models consist of eight formula templates. When the characteristics of a building are entered into the first template, the predicted costs, duration, and the start dates of each activity are produced immediately. The second template summarises this data, and where applicable, calculates the cost and duration for each floor, along with the unique project types, eliminates unnecessary data (for buildings with less than seven storeys), and specifies activities with no lags. The third template summarises the previous data, identifies relevant costs, calculates the week number an activity starts and ends, produces actual dates of these, and also an additional check to ensure all the figures are correct. The essential data i.e. cost, duration, start and end date of an activity, is copied and linked to use in the next template.

The fourth and fifth templates constitute the traditional monthly valuation model. The fourth template calculates the cumulative monthly costs of all the activities using determined formulae. The model would produce a zero (no cost) where applicable, until the activity began. The fifth template would then, based on these figures, automatically calculate cumulative cost, value and ultimately cash flow, including cumulative labour, materials, plant, and sub-contractors' costs and payments, based on assumed percentages. Hence producing the traditional monthly valuation approach model.

Templates six, seven and eight constitute the stage payment model. The stage payment approach has been linked with MS Project, where stages can be chosen manually by creating special filters. Any combination of stages can be experimented with to see how they will affect cash flow in practice. The sixth template contains the activity stages chosen in MS Project, including overall cost and duration for each stage, which consists of the activity sub-totals contained in each specified stage. Template seven calculates when the lump sum costs occur in the project calendar, using different formulas as above than the previous model. Template eight calculates the cumulative cash flow etc. in a similar way to before.

Graphs of cumulative costs, value, and cash flow are automatically produced for every project, for both models. Figures of 5 per cent retention, 30 per cent labour and plant costs, and 20 per cent material and sub-contractors costs respectively, were assumed, but can be altered to suit individual project needs.

Forecasting S-curves and calculation of SDY

Actual S-curves were produced for every project by calculating the actual cumulative monthly costs, and then expressing each month's total as a percentage of the overall cost. The predicted S-curves were then produced automatically via the cash flow model, and again the cumulative monthly totals were produced, along with percentage totals. Graphs were produced automatically to compare visually the actual versus predicted in each case. Every predicted S-curve compared well with actual curves in each case. The Standard Deviation about the estimate of Y (SDY) values were calculated via an established equation, based on these two sets of actual and predicted cumulative monthly totals, for every project. The lower the SDY, the more accurate the model. The S-curves were first standardised over ten months and then as a monthly percentage of overall total (actual duration). This would also determine how accurate the model really was. A mean average SDY of 2.400 was produced for all of the 50 projects. The results were compared with previous models and found to be more accurate than four out of the five studies. The studies produced mean average SDY of 1.750, 2.996, 5.229, 5.581 and 7.580 respectively.

Testing the model

To determine the true accuracy of the model, it had to be tested on a new set of projects that were not included in the development of the model. The programmes of work for all of the six test projects were all standardised successfully including amalgamating costs, durations, and producing time lags, the same way as before. The multiple linear regression models again produced reliable estimates and conclusively show that the model can produce reliable estimates of construction cost, duration and time lags, and hence can predict the whole programme of works for any building project in the sample.

The maximum absolute percentage error for the six test projects, for overall cost, duration, and time lags, were 8 per cent, 14 per cent and 15 per cent respectively. Again, the SDY was calculated for the six new projects, to emphasise the model accuracy. A mean average SDY of 3.885 was produced. As the SDY values suggest, actual and predicted S-curves were very close to each other. The results are very reliable despite, as expected, being slightly less accurate than the data used in the development of the model. Overall, this again confirms that the accuracy of all aspects of the model.

Contribution to the construction industry

The desired outcomes of standardising the programme of works, by establishing activities, sequencing and dependencies, would be to increase certainty in project delivery timescale, reduce administration and management costs and achieve a reduction and/or elimination in any unnecessary activities and project waste.

Construction time performance, cost and quality, have been identified as the three crucial success factors for a construction project. The fact that all of these factors of the various phases of a building can be accurately and reliably predicted ahead of the events, will reduce uncertainty and hence greatly benefit the client and the rest of the management team involved in the construction process.

The traditional and stage payment valuation models can be used for two important applications: cash flow forecasting and project cost control. Cash flow forecasting at the tendering stage needs to be simple and fast, considering the short time available and the associated cost. Contractors rarely prepare a construction plan at this stage, and usually wait until winning the contract. Therefore a simple and fast technique of forecasting cash flow quite accurately is required. Once the project characteristics are entered, the results are immediate, and hence can dramatically reduce the time taken for the majority of the planning tasks for a project. The models are also useful as they are developed using actual data from previous projects.

The fact that the models are simple to use and easy to understand make their use desirable. The minimal SDY values also prove that the model is reliable thereby making a significant contribution to the construction industry.