Use it or Bin it… Can we still use data from the last 3 years?

Data… We have heard it all before. It is the word on everyone’s lips, in every industry, uttered by most of our colleagues as the most critical asset in existence.
Great strides have been taken within companies large and small to harness their data to extract the value hidden within. Why? It’s because the data generated from a typical construction project can show us how to mitigate upcoming risks, peer into projects with granular levels of detail and even be used to predict the likelihood and quality of projects being delivered… And that’s just the start!

But what do we really know about our data, given recent events over the last few years? What is it composed of, is it still usable? I believe the answer comes down to 3 factors that should shed some light on the state of construction data as it exists today.

There has been a massive upswing in the industry’s adoption of data platforms, machine learning tools, and database systems. Companies are scrambling to make light work of analytics and quickly disseminate their findings to key stakeholders and project personnel. There is an assumption that the technology will enable the end-user to collect and quantify project progress without long periods promptly spent pouring over reports, interpreting results, and double handling data. Yes, data platforms have indeed been proven to reduce reporting times substantially. However, if the data that underpins them contains outliers, once in a lifetime event and is erratic overall, how can we be sure that the results produced by our new tech are even accurate?

The impact of Brexit is only really being felt this year, made worse by the Pandemic. Supply chain issues spurred on by the lockdown have driven up the costs due to materials like bricks, timber and steel being held up on their way to UK construction sites. Coupled with a reduction in the amount of Labour available due to Brexit, prices have soared by 23% in 2021, according to the BEIS. Due to the inability to deliver materials to site, material pricing data for this year would be abnormally high in contrast to the last five years since the decision to leave the EU was revealed in 2016. This leads us to the next factor affecting the composition of the data.

Overall, the Pandemic has had the most significant impact by far. Normality is only just setting in, but reminders of the previous year still linger. The Pandemic wrought a blow to the economy, which has made it difficult to pick up on underlying trends given the volatility still present in the market. This doesn’t mean the data isn’t available. It means that the economic fallout post-COVID, financial stimulus for UK businesses and declining GDP have created conditions that skew the data. As we can see in the chart below, this makes the data vary significantly compared to previous years.

Putting it all together.

Q: So, what has this got to do with the usability of the previous year’s data?

A: It concerns the potential impact the data may have upon platforms and ML models which are designed to optimise the decision-making process.

In one case, the impact upon ML models trained on material costing data for the last financial year may create a model that returns predictions that could swing between overly conservative or risky. This depends on the assumptions built into the ML model as most are built to recognise historic patterns and predict their likelihood of occurrence in the future.

In another case, this data could impact platforms that rely on internal information to analyse a business’s short-term and long-term financial health. A sudden shift on a macro or microeconomic scale could affect the flow of data and any pre-COVID data models used in benchmarking business performance, internal resources, and forecasting. It could be that the analysis could lead to some poor decision making or reduce a business’s agility under turbulent market conditions.