Frequently Asked Questions (FAQs)

What is technical forecasting?

Other names for technical forecasting are "data-led" and "mechanical" forecasting. All these are methods of projecting the future values of a series of data points, based exclusively on data related to the target series, i.e. not subject to opinion or other speculative input. You might be interested in reading more detailed technical answers on our forecasting methods.

Why should I be interested in technical forecasting?

If your organisation has extensive data sets with multiple variables and/or data of a non-linear nature, use of TFL's services will optimize the value of your information to provide the best-informed decision-making environment.

Why is technical forecasting such a good idea?

Technical forecasting, when properly executed, gives the very best interpretation of the available data. This is particularly evident when applied to complex data, with many variables. The result is the most probable movement of the data over a specific horizon. A high degree of consistency is maintained between forecasts.

Should TFL forecasts replace my current system?

Definitely not. TFL strongly supports the use and application of best practices as advocated by the property profession. TFL products should be used to augment the current decision-making process in your organization.

What experience do you have in the forecasting industry?

Our directors and key personnel are eminently qualified in their respective fields of study. TFL has combined dozens of years experience in data forecasting and directly in the property research industry. TFL technical personnel are recognized leaders in the management, analysis and forecasting of sophisticated data systems.

What is the difference between your forecasts and those of an expert?

An expert can include ‘soft’ information - personal impressions and public opinion - when making his or her judgements. We only use ‘hard’ (real) data, on the basis that the quantitative influence of any soft information has already been discounted by the markets and is reflected in one or more of the time series our models are using. Our objective forecasts are superior when analysing hard data, although an expert can compensate to a certain extent by their use of soft, subjective data.

An expert can include localized, specific knowledge, such as impressions gleaned from consideration of an impending planning decision when making his or her judgements. Contrarily, TFL only uses 'hard' or quantitative data on the basis that the quantitative influence of any soft information is already discounted by the markets and is reflected in one or more of the time series used in the forecast. TFL believes that these methods can do a better job at extracting relevant information from data than an expert, but the expert can compensate to a certain extent by making better use of soft data.

More importantly, since TFL has the capability to mass-produce forecasts, TFL can offer a much finer grain of forecast than can possibly be managed by an expert. For example, TFL can offer month by month forecasts of residential property prices for up to 85% of all postcodes in England & Wales, or similarly of 90% of monthly commercial property price indices from sources such as Investment Property Databank (IPD). This feat is physically impossible for a single expert to produce in a similar time frame.

Why should people buy your forecasts?

Many professionals in the property and financial industries use TFL forecasts to provide a unique perspective on future values of specific market indicators. This is important because TFL forecasting methods provide a completely independent, third party evaluation of expected market movements which are used to verify their positions or internally generated forecasts.

Most importantly, TFL clients benefit from an 'early warning system' of value changes so that actions relating to acquisitions and disposals can be planned further in advance of expected market changes. This is particularly true for forecasts longer than 1 year where TFL methods have been shown to continue to provide more accurate results than other available methods.

A significant benefit to the verification process is that it saves valuable research time particularly when forecasts are found to be congruent between methods. Clients can direct their research efforts more efficiently by focusing on the instances where disparate results between forecasting methods are observed or where turning points are anticipated.

The results: are they really any better?

This method is the best way to get a forecast from non-linear data. This can be proven statistically.

Are you always right?

No, and we would never claim to be. For the right time series, we believe that over a long period of time we are more likely to be right than wrong.

Nobody can guarantee their forecasts, but using ours offers people the knowledge that the forecast is based on the rational use of past data. If an individual has extra knowledge beyond this, which makes the future environment different from anything reflected in previous data, then our forecast can provide a base from which he or she can make a judgement of the effects of that environment change.

TFL’s forecasts are designed for professional users who have knowledge of the markets that are being forecast. In this sense, the professional can decide for him or herself what level of confidence to have in the forecasts, given their own specific market knowledge.

How often are you right?

There are two significant issues in this question. The first is the definition of often, and the other surrounds the use of the word right.

The property profession recognizes that a forecast would be considered to be right if the direction and magnitude of a trend were correctly predicted. In many cases, a reasonable margin of error depends highly upon the data being forecast, and the market sector for which the forecast is produced.

TFL is currently reviewing the frequency issue. In general, it is our belief that our forecasts fall well within acceptable error ranges a very high percentage of the time. More data is available on our downloads page in our accuracy document.

How often is a TFL forecast 'dead wrong'?

This would be considered a very unlikely occurrence, but would not be discounted as impossible. There are a number of reasons why this situation is unlikely, but in general, this type of occurrence would be the result of an unforeseeable event (shock) or some gross error inherent to the data series being forecast.

Another reason this is an unlikely occurrence is that the technology TFL employs has the intrinsic capability to indicate when a forecast is impossible given the specific target and parallel series. This means TFL will never produce a forecast of random noise intrinsic to all data. If a true signal cannot be identified and subsequently forecast, TFL is aware of the situation.

What conditions must exist for TFL to provide accurate forecasts?

• A sufficient amount of historical data must be available.
• The data must be relatively smooth rather than represented by step functions.
• Data frequency should be at least quarterly, preferably monthly
• The target series must share some mutual information with other associated, parallel series.

What is the right type of data?

Data that has reasonably stable relationships with its previous values and other time series. For example, the movement of property prices is often influenced by prior changes in interest rates and so property-related time series are good candidates for our methods.

What is the wrong type of data?

Data that is either unrelated or weakly related to its prior values. For example, forecasting electricity demand a century ago must have been very difficult because of opportunities presented by revolutionary technology and the immaturity of the market. Now the electricity market has matured, demand can be forecasted reasonably well.

How does TFL differentiate the right from wrong data?

The method used by TFL will intrinsically or automatically determine whether the data is forecastable or not. This is one of the significant differences between how TFL forms a forecast and those methods that use 'soft' or qualitative inputs. TFL can clearly tell what series cannot be forecast and more importantly, why not.

How soon are the forecasts available?

This depends highly upon several factors. If TFL has previously forecast the data, parallel series are available and relevant and the data format does not require significant manipulation, the forecast can be ready in one day. In extreme cases, up to a week may be necessary.