Learn about the “standard” option for time periods that many companies get, plus available upgrades (for more money) and downgrades (to save money).
Standard Option
The standard option for time periods from IRI and Nielsen is 2 years of rolling history of weeks and quad-weeks, updated monthly. So what do these things mean?
The basic building block of all time periods is the individual week. A week is always 7 days, but a week in IRI starts on Monday and ends on Sunday, while a week in Nielsen runs from Sunday to Saturday. IRI/Nielsen will also provide other aggregates of weeks (like 12, 13, 24, 26, 52 and year-to-date). For measures like sales and pricing these are fine, but for non-additive facts these longer periods are just averages of the individual weeks.
The other key time period is a quad-week, or aggregate of 4 weeks. You can think of these like months, except that there are 13 of them in a year instead of 12. Although the use of quad-weeks may seem kind of strange, there are two good reasons why it is standard. First, companies use different definitions of “month” in their accounting calendars so there is no option that will please everyone. Second, with quad-weeks, there are always the same number of days (28) in each period, which is good when you are looking at several periods trended over time. Relative to calendar months, quad-weeks line up pretty well early in the calendar year but by the time you get closer to the end of the year, you may have 2 quad-weeks that end in October or November!
“Rolling years” means that when the newest periods are added, the oldest periods are deleted. So a rolling 2 years is 104 weeks, or 26 quad-weeks. You want to get at least 2 years of data so you can look at growth. Even if your products may be new, it is still useful to look at growth over time for your category and competition.
Most companies have their databases updated every four weeks. You would get the most recent individual weeks and quad-week available (and lose the oldest 4 individual weeks and the oldest quad-week). Expect to get new data about 2-3 weeks after the end of each quad-week when going with the standard option.
Deluxe Option
Once you have weeks and quad-weeks, all other time periods can be created so the deluxe option here is spending some money to have IRI/Nielsen create custom months that align with your accounting calendar. These can be purchased along with or instead of quad-weeks. This is something to consider especially if you will be using IRI/Nielsen data to better forecast shipments. (Look for a future post on common period aggregates and which ones to use when.)
You may want to upgrade and have an additional year of history on your database. Fortunately the cost is not linear, so getting 3 years of data is not 50% more expensive than getting 2 years of data, it is less than that. It is usually a good investment to go for the 3 years of data, as this gives you a bit longer term perspective on things.
You can also choose to spend extra money to get weekly updates (instead of every 4 weeks). This is only worthwhile, though, if a. your products are seasonal (like looking at weekly data for beer or hot dogs from May/Memorial Day through September/Labor Day) and/or b. you can act on the information and influence sales that quickly. If neither of those are true, then monthly updates are usually sufficient.
Another upgrade is to have the data delivered faster, as fast as 8-10 days after a period ends. Only do this if your company can actually react to having the data sooner!
Money-Saving Tips
One way to save some money on your database is to only have quad-weeks and forego the individual weeks. If your company does little to no trade promotion, this is something to consider. The downside is not being able to create many of the other period aggregates that are very useful.
In theory, you could choose to have only 1 year of history, but we have never seen this happen in real life! Most analyses are all about comparisons over time or vs. year ago, so one year is not that useful.
Some companies with limited budgets choose to get quarterly updates (which include quad-weeks and weeks) instead of monthly updates. This can save lots of money and should be considered when there are limited resources to analyze the data on a monthly basis.
As Robin points-out, one use of scanner data is to forecast shipment. I believe three years of weekly data is helpful if you plan on building a statistically based volume forecasting model. For example, if you are forecasting a seasonal product and the season in the current year is 10% greater than the previous year, you do not know if this is a trend or if the previous year was weak and you have just returned to “normal”. While two years of weekly data is minimally sufficient to answer this question, three years of data is likely to provide a more definitive answer.