Want to save 15 to 20 percent of revenue? Then make sure you have the right ad server. Here are two key factors.

Despite the amazing technology that goes into modern meteorology, the strange weather conditions we have been experiencing in the past few years have given climate forecasting a reputation as a field of more art than science. Obviously, nothing could be further from the truth. Happily, in the same time period, technology enabling publishers to more accurately predict their inventory "climate" has made forecasting in our business regarded purely as science.

Obviously, inaccurate publisher forecasting will result in lost revenue. When publishers underforecast, they leave money on the table because they are left with unsold inventory. When publishers overforecast, they underdeliver, leaving the publisher in the position of making up for the impressions they could not deliver.

Failures in forecasting cost publishers money; some publishers we've worked with found that they had been losing as much as 15 percent to 20 percent of total revenue. Forward-looking mistakes also aggravate and frustrate agencies and advertisers. For both these reasons, publishers increasingly are learning that they need as much insight into their inventory as possible.

If you are a publisher, ask yourself if you are using the best forecasting technology available to you, and whether this technology makes your forecasting exercise as simple as you think it should be. Many of the publisher toolsets and popular ad-serving technologies available today are not keeping pace with publisher opportunities and challenges. There are better technology choices for many of the default solutions on the market. Below are two features to look for in forecasting tools.

Your ad-serving solution should offer you actual data and not sample extrapolations
It used to be true that the size of server log files would necessitate the employment of small tracking samples by most ad servers. Just a few years ago, data could be analyzed only in small portions, and then extrapolated to predict future inventory availability. If your current ad-serving solution is still sampling certain impressions or only tracking certain portions of your website, you are not getting a complete picture of your inventory.

In fact, many publishers do not utilize the tools their ad server does offer because they realize analytics for most solutions have a considerable margin of error. Many work with third parties that have no relationship to their ad-serving company but which have built businesses on top of what these technologies lack.

The problem with this technique is that sampling provides just that, a sample, not an accurate reflection of actual inventory or a report of real data. The latest ad-serving technology provides accurate real-time inventory forecasting that is based on an exponentially larger data set instead of on a small sample.

Increasing historical periods when generating forecasts increases accuracy
Historical sampling limitations pose another problem that can hinder effective forecasting by old ad-serving technologies. When ad servers can only collect seven to 10 days' worth of data at a time, historical analysis is anything but exact. Obviously, data collected in such short time frames -- less than 10 days -- cannot reflect seasonal or holiday changes in impression rates. Nor can such condensed time frames help in determining whether an ad's success or failure is related to low impression rates or simply lack of traffic on the site.

New ad-serving technologies provide flexible historical reporting. Publishers can now request historical data for any time period, providing greater accuracy and more control. New toolsets enable publishers to select the start and end dates that most accurately predict future goals, allowing publishers to account for seasonality and trends.

In addition, publishers can compare each day of the week against that same day of the previous week, which becomes particularly important in determining the ending date of a campaign. For example, if a campaign ends on a weekend day and historical reporting data shows the highest volume occurs on Tuesdays, publishers unable to compare specific days of the week may underdeliver.

Accurate forecasting closes the inventory loop, optimizes publishers' revenue
Whether you are using content, demographic or behavioral criteria to analyze forecast results, no amount of analysis will substitute for real-time calculation of inventory numbers, historical data and impressions -- the backdrop against which a publisher can make accurate forecasts. Without this insight into the big picture, no publishers can model and package their inventory for maximum value. Insight and accuracy in reporting make a significant difference. No matter which tools you are relying on today, if they are not keeping pace with today's forecasting opportunities and challenges, it is time for a system upgrade.

Jeff Wood is vice president of sales, Microsoft Advertiser and Publisher Solutions, Microsoft Corp. Read full bio.