An examination of weekend audits
In the current state of store audit research, the audit technique receiving the least use is the weekend audit. However, the weekend audit, which measures store sales and shares from Friday to Monday, promises to meet some research needs which may not be met by more traditional techniques. Among these needs are
- Accuracy — overcoming the invoice (delivery slip) gathering problem particularly present in categories such as soft drinks, beer, cigarettes, and snack items.
- Flexibility — to get necessary detail at particular points during the course of a study while maximizing research dollar efficiency over those time periods when detail is not needed
- Speed — providing for rapid data turnaround of detailed sales information over all classes of trade and in virtually any market.
In showing how weekend audits might meet these needs, traditional auditing must first be examined.
The Traditional Store Audit:
The purpose of traditional store auditing is to measure sales of particular brands and their competition while varying one or more factors in the environment. Tests commonly conducted include: new product, new package, promotional activity, media weight and pricing level.
In the traditional audit, sales of items not controlled by the testing company are determined by taking an opening audit of the store, recording deliveries / transfers of items into the stores during the audit period, and taking an ending inventory:
Sales = Opening inventory + (Deliveries - Transfers) - Ending inventory, or [(Amount sold) = (Amount for sale) - (Amount remaining)]
The accuracy of the sales data depends on the accuracy of the two inventory counts and the accuracy of the delivery information. It is relatively easy to get a correct count of items on a store's shelf, on display, and in the back room. But even if perfect inventory counts are obtained, missing or inaccurate delivery information will result in incorrectly calculated sales.
A factor of major importance in the traditional audit is that missing delivery slips will result in fewer sales being recorded across the category. Of course, if the test product is being delivered by the testing company, test product delivery slips are less likely to be lost. The danger is that if delivery slips are missed for other than the test brand, category sales will be lower than actual sales and test product brand share will be overstated.
Obtaining accurate delivery information turns out to be the most difficult part of the traditional audit procedure. The testing company becomes dependent on the cooperation of stores, chains, and suppliers, and on the accuracy and availability of their delivery invoices. Not only must invoices be available, but they must correctly delineate the items delivered by description, brand, flavor, size, and date of delivery. If flavors are not itemized, for instance, it becomes impossible to accurately apportion deliveries to the correct flavors. In the case of deliveries made near an audit day, it is important to include the delivery in the correct audit period. This becomes basically a question of whether the delivery was in the store and counted by the auditor at the time of the audit. Deliveries put in the wrong period can produce artificial fluctuations in sales from period to period.
The testing company is dependent on cooperation from all sectors in the market in regard to delivery information. Only those firms which are able to establish long term cooperative relationships are likely to produce accurate sales data. Information on store delivery patterns, invoice handling, volume, promotional patterns, and stocking practices is only acquired over time. Part time or inexperienced personnel may be sufficient for shelf counts, but greater capabilities are required to deal with delivery slips. In addition to the importance of good cooperation, Marketest, the store audit service of Market Facts, has found it essential to read invoices on a weekly or more frequent basis in order to help ensure accuracy.
Length of Audit Period:
An important consideration in store audit research is the length of the audit period. Theoretically, this length could vary from minutes to years. The most commonly utilized length is a four-week audit, in which a period of 28 days separates the opening and ending inventories. Shorter audit periods, such as a week, are less common. One of the reasons for this is increased cost, since it is more expensive to use shorter periods. For instance, to cover a test of six months duration requires, with the necessary opening audit. 25 [1 + (4 X 6)] weekly audits but only 7 (1 + 6) four-week audits. Another consideration involves the expectations of the marketer. Shorter audit periods will have greater variation in sales from period to period. However, these short term variations may be of little interest to the researcher who may, for instance, be interested solely in how his new product performs after the initial "novelty" effects wear off.
Another aspect of audit length is the effect on accuracy. Weekly audits should be more accurate than four-week audits, even though invoices/delivery slips may be read weekly in either case. This is because any one missed delivery or miscounted inventory has a more visible effect on sales over a one-week period than over a four-week period. Mistakes on a weekly audit are more obvious and can be caught and substantially corrected by backreading invoices and recounting stores within a few days of the end of the period. By contrast, backreading over a four-week audit period may mean searching for invoices that are four weeks old and are therefore less likely to be found.
Although weekly audits tend to be more accurate than four-week audits, all traditional audits will be inaccurate if correct delivery information can not be procured. Indeed, as mentioned above, many categories have proven resistant to attempts to obtain accurate delivery information.
Weekend Audits:
The weekend audit, a non-traditional method for obtaining share and sales data, is well suited for categories in which invoice reading is a problem. A weekend time period (Friday - Monday) eliminates most invoice reading problems due to the very low frequency of store deliveries occurring over a weekend. In all cases where there are no deliveries, sales over a weekend are determined simply as the difference between the Friday inventory count and the Monday inventory count. In the few cases where there are deliveries, calculated sales (Friday inventory - Monday inventory) are likely to be negative or obviously depressed. Because of this the auditor is in a superior position in regard to being alert to problems involving deliveries. Delivery slips are searched for before the auditor leaves the store after the Monday audit.
Those deliveries which do occur on the weekend cannot be ignored. Personnel who are familiar with a store are more likely to be aware of deliveries and more likely to be able to track down any invoices.
The weekend audit also allows for a more detailed examination of the effects of promotional activity. The longer the audit period the more any short term promotional or special pricing effects will be averaged out, since sales data for an item on promotion will be combined with sales data for that item while it is not on promotion. There will also be the effect of sporadic promotional activities on competitive items during the period. These considerations make any analysis of promotional effects rather difficult with the longer audit periods.
Weekend audits, as opposed to weekly and four week audits, will tend to correspond to the promotional activity cycle that generally runs from Wednesday through Sunday. With weekend audits the true selling price of items can be recorded, while with longer periods an average or "regular retail" price must be used. Additionally, major brands within a category attempt to avoid conflicting promotions. As a consequence of the relative absence of extraneous effects, weekend audits should be very useful in the analysis of promotional activity.
Another useful aspect of weekend audits is that they allow faster data turnaround. With weekend audits there is no need for the one week wait for delivery information following the closing audit) that is necessary for the longer audit periods.
It should be noted that it is not necessary that audits be contiguous. The usual practice for weekly and longer audit periods is for the ending audit inventory of one period to become the opening inventory for the following period. Weekend audits, of course, are noncontiguous, each requiring an opening and closing audit. But to suit particular research purposes, any type of audit could be conducted at intervals in a non-contiguous fashion. In the same context, weekend audits need not be conducted every weekend but could be concentrated on those weekends when promotions were run. The selection, timing, and intermixing of audit types can be adjusted to most efficiently utilize research dollars while providing necessary information to the researcher.
It should be noted that it is not necessary that audits be contiguous. The usual practice for weekly and longer audit periods is for the ending audit inventory of one period to become the opening inventory for the following period. Weekend audits, of course, are noncontiguous, each requiring an opening and closing audit. But to suit particular research purposes, any type of audit could be conducted at intervals in a noncontiguous fashion. In the same context, weekend audits need not be conducted every weekend but could be concentrated on those weekends when promotions were run. The selection, timing, and intermixing of audit types can be adjusted to most efficiently utilize research dollars while providing necessary information to the researcher.
To summarize:
- Weekend audits tend to avoid delivery tracking problems, allow for picking up actual prices, provide for fast data turnaround, and allow for accurate measurement of promotional effects;
- Weekly audits appear to be more accurate than longer period s but are more expensive; and
- Four-week or longer periods are less expensive and tend to average out short-term promotional effects.
The above conclusions say little about how the results of different audit types compare. Given the availability of invoices in a category, weekly audits should produce more accurate results than audits of any longer period. Therefore, the most stringent test we can devise for weekend audits is to compare their results with the results of weekly audits. The major hypotheses for this study were that across stores:
- Brand shares from weekend audits would closely agree with brand shares from weekly audits; and
- Brand sales from weekend audits would be closely related to weekly brand sales
Study Design:
In order to evaluate differences and similarities between the audit types, Marketest, the store audit division of Market Facts, conducted audits in 5 cities, 165 stores, in 4 categories, over a 12-week time frame. One four-week audit, 4 weekly audits, and 4 weekend audits were taken on a rotation basis. Each store had one month of 4 weekly audits. one month of 4 weekend audits, and one month of a four-week audit. All data normally picked up by Marketest, including promotional activity, were gathered over the course of the test. The categories tested were spaghetti sauce, 46 ounce canned fruit drink, toothpaste, and cold relief products, the last two being audited in both food and drug stores. Marketest had no control over the stocking or deliveries of the items audited. The audits were conducted in markets where Marketest has permanent field offices.
Empirical Results:
One of the major questions of the study is the comparability of brand shares across audit types. Brand share is a measure that allows comparisons to be made from audit periods of different length. However, there are some difficulties in using brand share. First of all, brand shares are mathematically dependent on one another, since the sum of all shares must equal 100%. A change in one share will affect the shares of at least one, and more likely several other brands. This dependence becomes more obvious the smaller the number of major brands in a category. An increase in sales for a brand due to a promotion will come from two sources:
- Switchers from competitive brands, and
- New or infrequent purchasers entering the category.
Both of these effects will lower the shares of competitive brands while only the switching effect will lower their sales. Promotional activity causes a change not only in the share of the promoted brand, but also in the shares of other brands. But even with this uncontrolled promotional activity, there was found to be a high degree of correspondence between weekly and weekend shares.
Week and Weekend Share Comparisons:
The following analysis of share differences was conducted by comparing weekend shares to weekly shares for each city, in the same manner that shares might be evaluated by researchers using typical audit data. Shares were calculated by summing city brand totals over the four weeks or weekends and dividing by the appropriate category total.
The weekend and weekly shares for representative brands are presented in Tables 1 and 2. As can be seen from Table 1, in the fruit drink category all comparisons were within 5 share points, 15 of 20 being within 2 share points. For spaghetti sauce, 11 of 18 comparisons (Chef Boyardee not being distributed in two cities) were less than 1 share point apart, only 2 were more than 2 points apart and one of those occurred in a city where the brand was in limited distribution .
For the toothpaste category, 18 of 20 comparisons were within 3 share points in food stores. In the cold relief category, 16 of 20 were within 5 points in the food stores. Toothpaste and cold relief products in drug stores (Table 2) exhibit greater share differences, which may be due to our smaller sample of drug stores — 9 per city as compared to 24 food stores. For toothpaste, 14 of 20 comparisons were within 5 share points, while the cold relief category had 10 of 20 with in 5 points.
The plots of weekly brand share by weekend share (Figures 1-4) for all brands graphically demonstrate these results. In these figures, each plotted point represents a brand share in one city for four weekends on one axis and for four weeks on the other. Figures 3 and 4 have two points per city, one for food stores and one for drug. If weekend shares for all brands in a city exactly matched weekly shares the data points would fall on a straight diagonal line with a slope of 1.0. In Figures 1 and 2, the data are very close to this ideal line. As can be seen in Figures 3 and 4, toothpaste and cold relief show more variance around this line.
Share Correlations:
As expected, the correlations between brand shares for the two audit types revealed a pattern of close agreement. Correlations between weekend and weekly shares for the two food categories, associated with the plots in Figures 1 and 2, were over .99. (A coefficient of 1.0 indicates perfect agreement. while a coefficient of 0 indicates no relationship.) Since there was some concern that this pattern of agreement might be different for individual cities, audit periods, or store types, correlations were also run within city, audit period, and store type as follows. Within the cities, the correlations for fruit drink and spaghetti sauce remained larger than .99. When broken out by individual periods and cities, the average correlation coefficient (of 16 combinations of 4 weekends and 4 weeks) ranged from .80 to over .99 for fruit drink and from .92 to .99 for spaghetti sauce. The lowest coefficient, .80, for fruit drink resulted from a major promotion in one chain. Matching week and weekend audits by time period in that city resulted in an average correlation of .98, showing that both weekly and weekend audits picked up the effect of that promotion.
The correlation coefficients for toothpaste and cold relief shares, corresponding to Figures 3 and 4, were both .91. When broken out by store type and city, toothpaste correlations ranged from .72 to .98 for drug stores, and from .96 to .99 for food stores. Cold relief products ranged from .77 to .98 for drug stores and from .94 to .97 for food stores. When broken out by individual periods, cities, and store types, average correlations for toothpaste ranged from .47 to .81 in drug stores and from .89 to .97 in food stores. Average correlations for cold relief ranged from .76 to .89 in drug stores and from .88 to .95 in food stores.
The correlations tended to be lower in those situations where brands could have tittle or no sales over a period. For instance, lower correlations were found in drug stores, of which there were fewer in the sample, some of which had low retail volume, and also had slower moving categories. Thus, weekend audits require categories and stores with a turnover rate high enough that at least some sales occur over a weekend across the brands being audited.
Week and Weekend Sales Comparisons and Correlations:
A separate question is the relationship of week and weekend sales. Plots of weekend sales by week sales are presented in Figures 5 to 8. It is clear that there is a good linear relationship exhibited in these plots, again with the food categories showing a tighter pattern. The food categories also appear to be dominated by one major brand. This dominance causes the corresponding correlation coefficients to be quite high. But it is also clear from the plots that there is still a good linear relationship even when the major brand is ignored.
The correlations corresponding to Figures 5 to 8 were over .99 for both fruit drink and spaghetti sauce, and were .95 for both toothpaste and cold relief categories. All additional breakdowns of correlations are identical to those reported above for shares because the denominator for the share calculation is the same for all sales figures in a city, making the shares a linear transformation of the sales figures.
In the categories examined, the ratio of week to weekend sales was approximately 2 to 1, with half of all sales occurring Friday to Monday. While there is a very strong relationship between weekend and week sales, some knowledge of the week to weekend ratio for the individual brand or category will be necessary in order to predict week sales from weekend sales, since this ratio will vary by category.
Conclusions:
It is apparent from these results that weekend brand shares will be very close to weekly brand shares given:
- A sufficiently large sample of stores, and
- A product or category that shows some movement over a weekend.
The few cases where lower correlations and greater differences in shares were found involved a smaller number of stores (drug stores) and/or a slow moving category. It can also be concluded that given the same conditions, weekend sales are closely related to week sales.
The main conclusion of this analysis is that for certain categories and circumstances weekend audits are a meaningful and efficient method of gathering data on sales and shares. Indeed, in cases where delivery information is lacking or where detailed examination of promotional effects is desired, weekend audits offer unique advantages. Weekend audits also offer rapid data turnaround which can be provided in all classes of trade (supermarket, drug , liquor, etc.) and in any market.
Greater confidence can be placed in weekend audits for faster moving categories and for brands likely to show some sales over a weekend than for slow moving categories and brands with very low sales figures. In designing a weekend test, care should be taken to get a sufficient number of stores in each panel or city to ensure stability of sales and shares.
The qualities of the different audit types, the research objectives, and the nature of the audit category need to be taken into account in selecting audit types and designing research. Assuming that delivery invoices can be obtained, perhaps the best research strategy would be a combination of audit types timed to coincide with critical management decision points. Such a strategy would allow for detailed information on promotions and pricing effects from weekend audits to be available for early evaluation, while long term trends could be investigated using less costly longer term audits. In this manner, timely information can be provided while maximizing research dollar efficiency.
As has been seen, there are advantages and disadvantages to each audit type. It is suggested that the advantages of weekend audits make them both an important additional research tool when used with other audits and a valid alternative when other audit types are impractical.
¹not in distribution
²limited distribution
