by Mike Shatzkin

I don't have to tell a roomful of people from the Canadian publishing industry that Canada is the most difficult market for publishers in the English-speaking world. You know it, although it may add to my credibility with you to assure you that I know it too.

From an outside perspective, what makes your market so difficult are these factors:

1. You get deluged with editions from all over the English-speaking world, particularly the US and the UK, and can readily have more than one edition of a title available in your market.

2. Your trade can easily source directly from the US, no matter what Bill C32 says about parallel importation.

3. Although publishers in the US and UK complain about consolidation of the retail sector, they have a virtual "free market" compared to Canada, where one chain controls a portion of the market which is the work of two or more chains in any other major English-speaking territory.

4. Your core market is so small that it obliges a Canadian publisher or distributor to have an exceedingly long list of titles to survive, compounding the universal problems of managing copious amounts of detail that bedevil publishers everywhere.

5. Your market is so dispersed and "dumbbell-shaped" -- weighted heavily at the ends with not much in the middle -- that shipping and fulfillment costs and lead times are inevitably high on a per-unit basis.

Accurate title metadata and fulsome inventory and sales-tracking data can ameliorate some of these issues. By our observation, the benefits of both grow in proportion to the size of the list a publisher has to manage.

Bottom Line First: Cutting to the Chase

It is because your market is both so small and so challenging that it has required a government initiative to get BookNet Canada started, rather than relying on the profit motive to create the investment by private companies, as happened in the US and the UK. The fact that this initiative begins by thinking about what is best for the market constitutes an opportunity that did not exist in the larger markets. Keeping the following considerations in mind as you move forward will help you to a more efficient and productive book trade.

• Carefully follow and implement the metadata schemes, specifically ONIX, which have been developed elsewhere. BookNet Canada has apparently taken the initiative to stay in touch with BISG in the US and BIC in the UK on these developments; it is critical that there be rigorous industry-wide effort to know and implement those standards. It is worth noting that the major accounts in the States, particularly Barnes & Noble, are important drivers to encourage compliance with the evolving standards. In your market, Indigo would have the coercive power necessary to create widespread compliance. In a world where computers increasingly talk to computers in order to get books from a publisher's warehouse to a consumer's hands, it is critical that standards that describe the book be universal.

• Remember that POS data, particularly aggregated nationwide POS data, has real value with limitations. It definitely helps identify large market trends. It definitely helps an editor gauge the true sales success of an author's last book (although only when it has been in place for a while and has compiled a historical record.) But it is not really enough on its own to enable publishers to maximize the two most important benefits available from data: 1) getting the reprints right and 2) matching the inventory in each part of the supply chain to the real sales expectations for a book.

• Do not confuse the understandable reluctance of booksellers to broadcast their overall inventory positions with their general willingness to share inventory information on a particular publisher's books with that publisher. Editors may need to know about other publishers' books performance to make a signing decision, but sales management only needs to know about the performance of its own books to help accounts manage inventory. It is actually a little bit nuts that the primary industry-wide vehicle for collecting POS data in the US and UK does not routinely employ those same "feed" relationships to enable publishers to see unsold inventory and performance as measured by stock turn.

• Remember that daily or weekly or monthly feeds constitute a "snapshot", not a balanced picture of reality. In order to gain real value from sales and inventory information, it has to be viewed over time. As a practical matter in the US and the UK, that now means that publishers must capture and archive data in order to organize it for real analysis. Because publishers often want to integrate some of their own information into analysis, creating publisher-specific data warehouses is inevitable and perhaps is just as well that it be done by each publisher.

• When analyzing inventory in chains, both granular store-level and aggregated chain-level data is of value. Obviously, books don't sell uniformly in different stores, even different stores in the same chain. But chains generally try to differentiate their offerings, not just to accommodate geographical differences, but also store-size differences. That means that in everychain there are books stocked and selling in some places and not available in other places where they could sell. It is the aggregated picture that points to those opportunities.

I have personally been involved with the analysis of bookstore inventory and sales data for more than three decades; my father literally invented this field of study in the book business more than 50 years ago. We have been applying that knowledge for the past 2-1/2 years to help publishers interpret their Barnes & Noble data. So far, we've worked with more than a dozen publishers. Some of the bigger ones have taken what we do and incorporated it into their own data management efforts, but we do the data warehousing and report delivery for a growing group of publishers.

Of course, our experience only matters to you to the extent that you can apply it in your own market. Sifting through chain data to get inventories in synch with actual sales appeal is certainly an idea that travels; our techniques should work well for Canadian publishers with Indigo, as well as for smaller chains like McNally-Robinson and Book City, if you get the right data and apply it with a sensitive understanding of each company's own supply chain realities. And if you can "see" inventory nationwide, you will also have the tool you need to make the right reprint decisions quickly.

Data in the US and UK Markets

Of course, Bookscan is the primary data source in both the US and UK. In the US, Bookscan is the "distributor" for data offerings from Barnes & Noble and Borders. Both chains offer only aggregate, chain-level, information to the publishers about their books.

As we observed earlier, Bookscan itself only tallies sales, not inventory. In the US, publishers gather inventory information from the feeds of the two major chains and from the data offerings of the country's two biggest wholesalers, Ingram and Baker & Taylor. There are six principal general trade accounts in the UK market -- four retailers and two wholesalers. One of our clients is now attempting to get feeds from them. There is apparently no standard distribution of information from accounts to publishers, such as there is in the US.

The biggest publishers in the US have created "supply chain departments" whose primary purpose is to guide reprint decisions. To accomplish this task, the concentration of business in few accounts is actually an advantage. Publishers long ago learned to "check stock" at their customers' major warehouses before any big reprint; automation promises to enable reprint decisions informed by knowledge of unsold stock "out there" for all books before long. We intend to offer a "Supply Chain Department" capability for all publishers by Q205.

Where we have placed most of our efforts over the past 2-1/2 years is in analysis of the sales and inventory data at Barnes & Noble. They are not only the biggest bookstore chain in the world, they also have a comprehensive and integrated supply chain and a data feed that really enables a publisher to pinpoint problems and opportunities.

Getting started: warehousing the feeds

As we have observed, publishers must create their own data warehouse, collecting and archiving the feeds from several places both across time and across sources. No external feed, whether from a single account or from an aggregated supplier comes "complete". All of them require enhancement from information available only in the publisher’s own computer system. And all of the feeds are weekly snapshots and it is imperative to view data over a longer period than a week to analyze inventory efficiency. This is more true for books than it is for other items retailers stock. Most books do not sell in most stores in most weeks. Or even most months! Unless one views performance over a longer period of time, books that are actually performing quite acceptably can appear not to be selling at all.

The key metrics

Of the many metrics and groups of metrics we analyze at Barnes & Noble, there are five headings I want to detail here today. Some are “universal” and you will see immediately that they would apply in much the same way to any account, including Indigo or the smaller chains you sell to; some must be interpreted within the B&N context and would require a different interpretation in another chain; and some apply only within B&N and would not be used in other accounts.

For “short-term” inventory performance, we look at percentage of on-hand sold in the superstores in a week. We call that our “Flash metric”. Using that percentage is a very useful tool to find books that are “flying below the radar” -- performing well in relation to the inventory B&N has, even if that quantity, and therefore the absolute sales number, is low.

For any period longer than a month, we gauge performance by using stock turn, which we present as an annualized number dividing sales by average inventory. We look at stock turn for three different periods: a rolling 4-weeks, 13-weeks, and 52-weeks. Looking at the stock turn performance this way allows us to see both trends and seasonality.

We also do very specific aggregations of data, looking at groups of books together. The most obviously helpful such grouping is by “store category”, which at Barnes & Noble usually equates to a specific buyer. Looking at books by imprint, by series, and -- for those publishers who distribute others -- by originating publisher, also yields useful, actionable, analysis.

All of these metrics are universal. You can use them for any store or chain. But aggregating up to the buyer level is an account-specific exercise. Some accounts assign buyers to particular publishers while others buy by subject. And we have found that the subject categorization provided by accounts is not always as useful as it is at B&N. B&N devotes an uncommon effort to getting the metadata right; other accounts are less meticulous and one large one we know puts a “general” label on books far too many books which require a subject category. This makes the data harder to analyze, but it is also a problem that it costs sales to ignore.

As we said, some metrics are very specific to B&N’s own supply chain. These include models, indexing to dot com and mall sales, deciding what books need inventory increases or decreases, and the stock on-hand and on-order at the DCs (distribution centers.)

Models are a B&N device by which buyers set a stock level for a title in an outlet that is maintained by automated computer reordering. Adjusting models up or down is one of the most important components of having the right stock levels on backlist books.

When viewing B&N's chain-level data, remembering that there are approximately 650 superstores is critical. Knowing that identifies books that can't possibly be fully distributed or modeled in the chain because the on-hand position or total chainwide model is lower than 650. This magic number is different at Borders and would be different at Indigo or any other chain.

The relative performance of books through B&N’s dot com and mall store channels, as compared to their superstores, can also point to opportunities to adjust inventory profitably.

And although the ordering for B&N’s distribution centers is highly automated and very effective, it is not perfect and systematically watching the on-hand and on-order levels can also locate opportunities to improve sales. More often than not, problems with DC inventories trace back either to metadata errors or to a publisher’s own stock shortage. In either case, knowing about it can either correct a problem or avoid misinterpreting an apparently poor sales performance that has nothing to do with a book’s appeal.

The “ACE” Reports concept

Now I'd like to show you what these concepts mean in practice, analyzing things from a publisher’s perspective at Barnes & Noble. We call the reports we create “ACE Reports” because the acroynm explains what makes them distinctive.

A is for “archiving the data", putting those weekly feeds into a database so that one can view the impact of sales and inventory over time.

C is for "calculating meaningful metrics", such as those we have discussed earlier: percentage sellthru, stock turn, and others.

E is for "editing the reports", so that what is seen is what is needed for analysis, either by selecting out particular groups of titles, or sorting by various metrics, or moving unhelpful data off the screen.

Demonstrating how they work

We have three standard report formats to analyze Barnes & Noble inventory for publishers.

The Flash report looks primarily at sellthru in the superstores for the past two weeks. Here is an example of a Flash report from a major publisher from last November 26 - December 6. Explanations of the headers are as follows:

• CATEGORY: Category as shown in BookScan.
  
• DEPT CODE: 1-Trade paperback; 2-Trade hardcover; 3-Music; 4-Newsstand; 5-Gift; 6-Juvenile; 7-Bargain Books; 8-Mass market; 9-Non-Book; 10-DVD's; 11-Used books; 12-Retail Food; 13-Video; 14-Calendars; 15-Maps/Misc.
  
• ISBN: The ISBN for the title.
  
• TITLE: Title of book as shown in BookScan.
  
• AUTHOR: Author of book as shown in BookScan.
  
• SHIP DATE: The date the book first shipped, provided by the publisher.
  
• SS SALES: Superstore sales for one week.
  
• YTD SS SALES: Total Year-to-Date sales in Superstores.
  
• SS MODEL: Superstore model for this title. Numbers lower than 10 often are new store orders rather than a model.
  
• SS OH: Total Superstore inventory on hand for one week.
  
• SS OO: Total Superstore units on order. We don’t know if the “on order” is aimed at the publisher or at B&N’s own DC.
  
• PY SS SALES: Total Superstore sales for the prior calendar year.
  
• DOT COM YTD: Year-to-Date Dot Com Sales.
  
• DC OH: Total Distribution Center inventory on hand. The DC wants at least 2 copies of every book on hand.
  
• DC OO: Total number of units on order for the Distribution Center.
  
• % SS OH SOLD THIS WEEK: The percentage of Superstore on hand inventory sold in one week. Any unpromoted and undiscounted title selling more than 10% is considered a fast moving title.
  
• % SS OH SOLD PRV WEEK: The percentage of Superstore on hand inventory sold in the previous week. Any unpromoted and undiscounted title selling more than 10% is considered a fast moving title.
  
• % SS OH/SS MODEL, END OF WK: Percentage of Superstore model that is in stock at the end of the week. The model indicates buyers' intent of a stock level. If this number falls below 85% consistently and there has been no stock shortage at the publisher’s warehouse or the DC, it strongly suggests there should be a model increase.

Of course all identifying data in this particular report has been anonymized. To do so, we sorted the report by that week’s sales (the column headed “SS Sales, Wk of 12/06/03”) and labeled the top-selling title Title 1, the second-highest Title 2, and so on. It is interesting to note that an alternate sort - sorting instead by inventory (Column “SS OH, Wk of 12/06/03”) - yields a nearly-identical result, showing that high sales most often indicate a lot of inventory.

A more useful sort would be by what we term the Flash metric - the percent of superstore inventory sold in the last week. This way, the books are sorted according to performance - not simply sales, which we just saw was directly related to inventory - and it is easier to spot titles with smaller on-hands that are still doing well. We call these titles ‘flying below the radar.’ You’ll see that, with this sort, the top ‘action title’ here is the title number 89 in sales among the TOP PERFORMERS, much lower than that if we were looking at all their titles!

The Category Summary “rolls up” groups of titles for examination, usually by store category. Here’s an example of such a report, for a different publishers in a different period, ending last September 22nd. The column headings for this report use abbreviations similar to those in the Flash report. In this case, we have sorted the report in descending order of on-hand quantity, to see them in order of importance. As we’ve discussed, this organization of information is also useful to look at books by imprint, by distributed line, or by series.

The Stock Turn report calculates the key metric of “stock turn” for every single title for three periods - 4-week, 13-week, and 52-week - that we generally employ. Here’s an example of such a report for the same publisher we just looked at, and for the same period. (Note that the numbering here is by on-hand quantity.)

With the Stock Turn report, first we sort by 52-week stock turn in descending order. This way, we can see books which are performing well, and see if increased inventory would likely result in increased sales. We can also locate problem titles by sorting in the reverse order - 52-week stock turn in ascending order. It is then easy to note the books that have high inventories, but which are not turning over at an acceptable rate. Lastly, we sort by Model in descending order. This way, we can see titles which might be potential candidates for a model increases or reductions.

Using our Stock Turn Reports, we believe a sales rep can do a very thorough job of reviewing all the titles in any section in less than an hour and be confident that no opportunities to change inventory to improve profits will be missed.

Making use of the analysis

One thing our analysis accomplishes -- we believe for the first time in this history of publishing -- is that we standardize a publisher's view of inventory data by the metrics that affect their customer's profitability, namely the velocity of sellthru, or stockturn. Publishers' customers are acutely aware of this metric. Buyers are often bonused on the basis of it. Accounts doing annual reviews of publisher performance often calculate and refer to it. But until we developed ACE Reports, no publisher that we are aware of ever systematically calculated it. We believe, at the very least, that it is helpful to a sales-and-service relationship to look at the profitability of inventory the way your customer does and we're proud of starting that process for the industry.

Precisely how these reports are used varies with the publisher and the B&N buyer. We envisioned the Flash Report to be the tool to catch things in a "right now" fashion -- books that are spiking or suddenly tanking -- and the Stock Turn Report as the tool for periodic backlist reviews and "section balancing." However, we are aware of at least two mega-publishers who are using Flash Reports in monthly backlist reviews. One of our clients has commenced a regular backlist review since we started delivering these reports with a very positive impact on their sales at B&N.

Our intention is to deliver analysis which is actionable, but also "reportable." In a perfect world, we'd hope that the same Excel row which told a rep that a title should have its inventory expanded or its model reduced, for example, could just be sent to the buyer and provide the information the buyer needs to consider the suggestion.

Making use of store-level data

We have been made aware that, at least until the new SAP installation, Indigo has made sales and inventory data at store-level available at least to some publishers. There are some obvious additional opportunities for analysis enabled here but also some serious data management challenges. Multiplying the number of data points by a factor of 100 or more to see each store separately vastly complicates the warehousing and report generation process. But it also greatly expands the opportunties.

Among the things I would do if I had access to store-level sales and inventory data:

1. Calculate the stock turn by section by store for your titles. This will tell you, and Indigo, where your odds are best for expanding your offerings for any category of your titles.

2. Calculate the stock turn by title by store, but only over a period of 3 months or longer and only for books that are stocked pretty much continuously. This is actually a tricky exercise because the "average inventory" component of the stock turn calculation is hard to figure for books stocked in 1s and 2s and selling pretty quickly. It is hard with a weekly feed to correctly calculate how any days of "zero" inventory there were.

3. For all books that sell more than 15% of on-hand two weeks in a row, check to be sure you have full distribution in the chain. Sometimes it won't be warranted, of course, because books often do have some kind of regional appeal, but this is something we attempt with chain-level data that is much easier to do with store-level data.

4. If possible, track turns on promotions. This can be hard to do, because you have to flag the weeks when promotions applied to particular titles. But it can tell you whether promotional dollars are well spent.

Applying aggregated national data

Since there is a good chance that Canadian publishers will have the opportunity to apply aggregated national data on sales in the forseeable future, I want to offer a few thoughts on how best to use it.

First of all, remember that rankings by sales are almost always closely correlated to rankings by inventory. That is, if you see a competitor's book is outselling a similar one of yours, don't leap to the conclusion that their book has more appeal to the consumer. It is almost certainly more broadly distributed if it is selling significantly better and that may reflect on the sales or marketing execution, not just on the book's appeal.

Second, try hard to understand the relationship between what you know you've shipped to the POS world (those stores or kinds of stores that are like what is captured in your POS reporting) and what is reported as sold. Even the most perfect POS-reporting system will not capture every sale, and there is a decent chance that the percentage it captures will vary somewhat by the book. You're going to try to subtract POS sales reported from copies shipped into the marketplace to understand your supply chain exposure. It is obvious that this exercise will be more accurate if you are able to back out the shipments and POS reports from a few key accounts.

The best use of aggregated national POS data is for marketing decisions: watching the sales of the competition and seeing the impact of publicity and promotion. Both of these exercises are more about relative performance rather than absolute numbers and the data can tell you a lot.

How hard is all this?

Although we offer a lot of "value-adds" in knowledge, consulting, and a constant improvement of our reports, there is inherently nothing about the ACE Reports concept that can't be done by any publisher with strong Excel and Access skills in-house. If you want to "try this yourself at home", we would offer only a few words of caution.

First, the set-up is not trivial. Going from the feeds you get to the spreadsheets you need requires building an Access database and painstaking creation of queries. This is not something even a skilled database manager will accomplish quickly or easily. It will certainly take days, perhaps weeks.

Second, this process requires ironclad discipline. You must grab and archive each relevant feed every single week without fail.

Third, we strongly discourage having a rep do this work him- or herself. This is a process which takes time and care. Reps have very important jobs to do with the care and feeding or their customers. It is too easy for really important priorities to shove the data work aside. So even if a rep is fully capable in the data programs and, frankly, most aren't, it would be wiser to provide the reports as a service to a rep, not ask them to maintain the data warehouse.

And fourth, even with the best of intentions and processes, we have learned that databases sometimes "break" and have to be rebuilt. This is a pain when it happens and can change the work in a week when it does from something requiring minutes to a task requiring hours.

Conclusion

The new era of ubiquitous data offers enormous opportunities for publishers to cut costs out of their supply chain. The publishers that do this can improve their unit costs the easy way, by not manufacturing books they'll never sell. We have also demonstrated that books do "fly under the radar"; overburdened reps and buyers require so much time to manage new title acquisition that such books have too often been allowed to die out, even though customers bought them when they were available.

So with reduced costs and increased sales as the incentives, there is opportunity for every publisher, every wholesaler, and every retailer to benefit from sharing inventory information and analyzing it by the measures of profitability.