A few years ago, one pharma chief executive, presented with a forecast for the company's planned push into a new therapy area, said, "Double the development costs, halve the sales forecasts and then tell me if you still have a profitable business." Luckily, they did, and the business is now beginning to ramp up nicely.

But for every forecast that faces up to that kind of reality check, ten or twenty must pass unscrutinised into the corporate bloodstream, leading to hasty revisions later.

Small wonder then, that management is sceptical about forecasting. What can planners do to allay this scepticism?

There are four main strands of forecasting tools: using historical data (in marketing mix and promotional response models and econometric ROI models), using analogue data (in marketing mix and promotional response models and econometric ROI models), relying on the 'wisdom of crowds' and the principle of averaging and using current market data in predictive models.

The historical approach is perhaps best summarised by IMS, which says on its website it 'provides a solid benchmarking framework for every drug launch. By using past successful product launches as a reference - and applying our analytic expertise to this historical data to produce customised forecasts - we can set the right strategic and tactical parameters for customers.'

Due to this and the limitations of traditional ROI models, in particular the difficulty of relating individual inputs to individual outcomes, many companies have started using econometric ROI models, promotional response models and resource allocation tools using historical data.

These models measure past relationships between variables (usually marketing activity spend and sales or market share) and then forecast how changes in some variables will affect the future course of others.

On the surface they appear to be all that the marketer is looking for - sophisticated models providing insight into customer decision processes, media channels, or spend.

Says Dr Andrée Bates, managing director of pharma marketing return measurement specialists Eularis, "Econometric models using historical data were embraced initially by the finance sector many years ago but they were found to be insufficient to make predictions with any accuracy in a changing environment.

Part of the problem is that reference to past trends largely ignores the casual relationships between events.

"For instance, if a previously fast-growing product starts to show a slow down in sales growth, it might be assumed that the graph of sales reflects an S-shaped curve, and the market is reaching its saturation point.

"However, closer analysis might reveal that manufacturing and supply issues may be the problem, or that the sales message is no longer reaching the target audience, or any number of factors that once addressed would lead to a period of renewed sales growth.

Without the ability to examine and critique the data to pick out underlying causes, and a method to test which of the causes might be to blame, the value of forward predictions based on past data is only going to be reactive."

The difficulty with the historical approach is that it carries an inherent risk. The data is by definition out of date, which in a fast-changing environment runs the risk of being misleading.

The Wisdom of Crowds

Another approach is based on the 'wisdom of crowds'. This centres on the observation that a surprisingly accurate estimate of, say, the number of jelly beans in a jar can be achieved by taking the average of a large number of individual guesses.

The forecasting strategy asks a 'Crowd' or panel of experts to estimate future commercial activity. A final forecast is then constructed by a co-ordinator plotting a middle course between individual estimates.

A variation is the Delphi panel, where the experts agree amongst themselves on the final forecast. This removes a lot of effort, and any coordinator bias, but the predictions are only as good as the panel's expertise. Experts are paid for what they already know, not for what they are prepared to find out, so key information may be unused or unavailable.

Delphi panels are also inflexible. If conditions change, a new competitor arrives on the scene or the economy declines unexpectedly, the panel must be reconvened to produce a new forecast.

It can also be difficult to challenge or critique the findings of the panel because the data in the forecast is based on is mostly in the experts' heads.

It can also be difficult to challenge or critique the findings of the panel because the data in the forecast is based on is mostly in the experts' heads.

A third approach, the principle of averaging, asks sales reps to produce forecasts and a middle course is steered between them.

Although averaging is common practice, it is not without its problems. Flaws may be baked in, driven by individual preferences or even personalities. Individual predictions might be over-inflated by optimistic sales reps, or under-inflated if the rep wants an easy target.

Additionally, estimates from people in the field would also tend not to factor in new product launches or other changes to the product portfolio, which are increasingly not being pre-signalled to sales staff to prevent new initiatives leaking to competitors.

Says Andree Bates, "The issue worsens as forecasts move up the chain of command. When figures and metrics are altered at each passing level, inaccuracies compound and, by the time the final predictions reach executive level for review, the numbers may be so divorced from reality that they render the final forecast useless."

So not only has inaccuracy been stirred into the mix, but averaged forecasts cannot be challenged or tested because their underlying assumptions are not clear. The executive board might either accept the forecast as the best of a bad job, or demand to see the justification of the figures.

Those demands would trigger a domino effect of demand for explanations and clarifications stretching back down the chain. This expensive process would be taking place while the forecast period ticks ever closer - or even ticks quietly by. So now what?

One way to remove bias might be to leave it to a team with more access to company strategic information, with a better handle on product life cycles or wider understanding of macro-economic/market conditions but, if the sales information is inherently flawed in the first place, the final prediction will not be improved.

Clearly, companies and brands that are accountable in their marketing investments, that can maximise the value return of these investments and deliver on them with financial results, are the ones that will prosper.

However, these factors are highly interlinked. What happens in the market impacts your results and accountability, so solving the strategy/performance gap involves untangling all the inter-relationships (separating the ones you control from the ones you don't) which is the key towards developing lasting solutions.

Given this interlinking, it is not so surprising that marketers pull the wrong levers to turn financial results around. They may be trying to execute more successfully when, in fact, the messaging is under-performing. They may try to change focus when, in fact, it is execution that lets them down. The results can be lost time, wasted energy, wasted budget and ongoing under-performance.

So, if averaging builds large amounts of gut feel and subjectivity into the forecasting process, the wisdom of crowds is suspect, and econometrics is sometimes fatally dependent on flawed source data, what is the way forward?

Getting the raw data right

One key to successful forecasts is getting correct data into models in the first place.

Seeking sources and "crunching numbers" is tedious, under-appreciated and often left to junior people. A good way to promote data integrity is to make sure someone with sufficient clout makes it their job to ensure the required information is accurate and a fair representation of reality.

CRM systems might be expected to ride to the rescue, providing current information about the business. But here again, forecasts are at the mercy of data quality. Recent surveys suggest less than half of sales reps were using their CRM systems properly, about a third were using them incorrectly and a fifth were not using CRM at all.

As difficult as it is to include unforeseen events and changes into planning, some companies are tackling this issue using plans grounded solidly in the economics of their current marketplace, as well as a strong idea of the relationships between activity and current - not historical - results.

A fresh set of challenges appear when market research and business intelligence need to consider first-in-class molecules.

The introduction of national health technology assessment bodies in Europe, including the UK's NICE and Germany's German Institute for Quality and Efficiency in Healthcare (IQWiG) has elevated the position of evidence-based medicine and cost-effectiveness on the healthcare agenda. It has also made it extremely difficult for first-in-class molecules to break into the market place.

Despite marketing authorisations and clinical support, a negative NICE can put a stop to prescribing, even with storing cost-benefit arguments.

This changed environment creates a range of new challenges for market research, says Danuta Holtz, Business Intelligence Manager at Merck-Serono: "Researchers now need to contact new groups of influencers such as payers, formulary members, and members of NICE to understand their needs. You also need to consult with people who can offer opinions on likely NICE reaction.

"They might have served on NICE committees or they could be key opinion leaders and influential health economists who have advised NICE."

Other influencers include primary care commissioners, who may fund exceptional cases early in the post-launch period, senior private health insurance executives who will be gatekeepers to much private sector use, and reviewing bodies like patient groups or charities.

Earlier research with key opinion leaders is needed to ascertain the clinical trials and comparators that are likely to be accepted for cost-benefit arguments.

"The new environment also produces recruitment challenges," adds Charlotte Tatton Brown of Research Quorum. "Target groups will come from small universes, so sample sizes will be small. For instance, you might be looking for established consultants with significant private patient volume."

Another challenge is setting up new performance monitors among small but significant prescriber groups. As NHS doctors are reluctant to fight for exceptional cases, greater potential lies with clinicians prescribing for private patients, but this data is not captured in currently available audits and has to be developed independently.

Research can't be nationwide when a drug may be approved in some areas and not others e.g. Scotland but not England, so there is a need to research in different areas of the country.

New research challenges, says John Peirce of Research Quorum, mean that attitudinal topics have to be researched to cover factors other than simply how good the new molecule product is.

However desirable the profile, it is not enough to make doctors prescribe it, which means the researcher needs to understand the prescribers attitude to the new product.

His or her willingness to 'fight the cause' for funding, often requiring a strong emotional buy-in from the prescriber accurately identifying local regulations and restrictions which might inhibit prescribing.

Testing proposed health economic models and arguments with each decision making group (doctors, pharmacists), and correctly weighting their differing reactions to predict success, where cost-benefit is likely to be a vital issue.

Ensuring, during the medical education process, that stimulus materials avoid over-positive biasing and present 'cons' as well as 'pros' to get a balanced and informed judgement.

Box: Analytics in action: a case study from pharma

A core brand was launched in the 90's and was a star performer for its company in a high revenue therapy area. As the years went by, it became a large brand for the company and was well-respected in the market.

However, over the last few years, performance has declined. Market share fell month after month. The marketers didn't know why. The product was good, run by a strong team with a good plan, and focussed on accountability and performance. The business unit submitted detailed financials with key performance indicators and updated forecasts.

The team tried increasing spend, but the brand consistently dropped in market share, despite spending close to double any of the competitors.

A new head of marketing came in. She felt that they were missing something obvious and needed to dig deeper and understand the problem before throwing more money at it.

She decided to use marketing analytics - basically a mathematical way to identify relationships between variables to increase effectiveness of marketing activities.

The marketing head employed the Eularis' analytics approach, designed especially for this industry and situation, which involved a five-step process:

* evaluate promotional elements & market environment

* validate actual influencers for a therapy category

* Use Predictive Algorithm Analytics and Dynamic Modelling

* Analysis of findings implications

* Implement the recommendations throughout the sales and marketing processes

This process highlighted where the problem lay. The brand was well liked by physicians - hence the high market share - the messages were strong, the sales force were skilled, the promotional activities were also strong, and yet smaller competitors were stealing market share nibble by nibble.

It would be easy to say 'That's just how it is - we will have to live with it' but the analytics uncovered three interesting areas.

Firstly, they showed that the reps were de-motivated because they had nothing new to say.

The second finding was that cost was an issue, which was really the brand's only weakness. To deal with this, the company could take pharmaco-economic data and add it to their rep calls to give the reps something new to discuss. They could incentivise reps to compete region by region, as the analytics also showed performance by region, and use the pharmaco-economic data to show the cost was actually a saving in the long term.

The third thing the analytics showed was that the market had three market share points in play, and it was clear that one of the stronger brands could take three vulnerable market share points, if they were positioned correctly.

Six months later, the analytics were re-employed. The recommendations had been followed and the brand had gained market share for the first time in a few years. They did not take the full three market share points, but they did take a sizeable chunk of it.

John Hosken is a principal consultant at Information Advisers and a qualified business coach. E-mail: jhosken@hotmail.com