Let’s make a performance prediction

Ideally, you want to know project outcomes and risks before they happen with as much clarity as possible.

You’re already analyzing past projects and revisions to keep your baselines in check, but just so we’re on the same page let’s run through how to get those stats. We’ll assume you’ve been exporting the data consistently over the years into a format for analysis. If not, there’s no better time to start than right now and remember once you’re out of the backlog, maintain it! So grab all of those MS Project and Primavera files and export them into a format like CSV.

Project type, phase, and task names are important but not actually critical. If you have a consistent naming format that’s excellent you’re a big step ahead. Critical attributes are Expected Duration, Resource, Actual Duration. The next part is optional only if you have task name consistency: group your Task Names into categories, go through each row and assign a high-level category that makes sense, use Master Format Divisions as a guide. Do the same for Resource categories.

Find the average (mean) of all durations. The aggregate Actual Duration average should be almost identical to your Expected Durations as this is the basis for your baseline. Now we’re all on the same page. You have definitive data points for tasks, task categories (baseline), resources, and resource categories (baseline).

So far we’ve been measuring Actual outcomes, to make a true prediction we need to know inferred outcomes. This is where your real-time percentage of completion and revisions come in. Whichever construction app your org is using, one of its features is likely daily reporting with a percentage of completion.

Take the Expected Duration and multiply it by the percentage being reported, that’s your inferred duration. Now take your Expected Start date and Current Date to create an inferred percentage of completion. Do this for each task. The result of your percentages to their respective durations when subtracted from each other (actualDuration – inferredDuration) will return a positive or negative number. If your durations are in hours you’ll have a Gain or Loss in hours.

At this point, you’ve got everything you need to make a prediction. Filter into tasks, categories, resources, resource categories, phases, and projects. You’ll sum all of the Gains and Losses into a total value for each filter. I.e. filter by task and then resource, total up Gains and Losses and now you have a value of estimated time as a Gain or Loss for a task group as predicted from the performance of the resource group. Either add or subtract that number from your baseline to get a prediction for that task based on the resource or resource group.

Do this on categories to get a benchmark to measure against and you’ll know if your expected outcomes are above or below average.

Depending on what you can extract from your construction app you can further augment these results with contextual attributes like issues, engagements, notes, mark-ups, and other types of deviations.

A little about us. Harbr is a Construction Performance platform, we offer a project management app and a risk and performance prediction service for legacy data for orgs that already have project management software.


Leave a Reply