Optimizing inventory with kits or bundles

Merchants are frequently selling kits (or bundles), where several items are sold together, while the possibility remains to buy the items separately. The existence of kits further complicates inventory optimization because it introduces dependencies between items as far availability is concern. In this post, we try to shed some lights about optimizing inventory in presence of kits.

There are two opposed approaches to deal with kits:

• Do not store any kits, only separate items. Assemble the kits at the last moment assuming that all items are available.
• Store all kits pre-assembled as a separate SKUs. Kits are assembled in advance. If no kit is readily available, the kit is considered as out-of-stock.

In practice, most inventory policies toward kits tend to be a mix of those two approaches.

Let’s start the review with the first approach. The primary benefit of keeping everything disassembled is that it maximizes the availability of the separate items; however, this comes at the expense of the kit availability.

Indeed, assuming the availability levels of items are independent and refered with L1, L2, … Lk (for a kit with k items), then the availability of the kit LK = L1 x L2 x … x Lk. 

Let’s assume that we have a kit with 5 items, all items having the same service level. The graph above illustrates the correspondence between the service level of the kit w/o of the service levels of the separate items.

For example, with 5 items at 90% service level, the kit ends up with a service level slightly below 60%. This behavior illustrates weakest link behavior of kits: it only takes one item to be out-of-stock to put the whole kit out-of-stock. Even if all items have fairly high availability, the kit availability can be much lower; and the bigger the kit, the worse it gets. If instead of 5 items, we consider a kit with 10 items at 90% service level, then the kit service availability is reduced to 35%; which is typically unacceptable for most businesses.

The second approach consists of storing pre-assembled kits. This approach maximizes the availability of kits. In this case, kits are treated as like any other item: the demand for kits is forecast, with quantiles forecasts, and a reorder point is computed for the SKU representing the kits. This inventory policy preserves a strict decoupling of the kit and its items.

With this approach, the service level of the kit is driven by the quantile calculation. As such, the kit is not negatively impacted by the separate availability of the items. Each item also gets its separate reorder point.

The primary drawback of this second approach is that, in the worst case, the amount of inventory can be doubled for limited or no extra availability. In practice however, assuming that about half of the item consumption comes for kit’s sales, the stock is typically increased by roughly 50% when applying this second approach instead of the first one; the extra inventory is used to ensure the high level of availability of the kit itself.

The optimal inventory strategy, the one that maximizes the ROI (Return On Inventory), is usually a mix of those two approaches.

The exact inventory optimization of kits is a relatively intricate problem, however the problem could be rephrased as: at which point should the merchant start refusing to sell separately one of the kit’s items because she would risk losing more advantageous orders on kits instead?

Indeed, all long as kits are available, there is typically no incentive for the merchant to refuse selling a kit in order to preserve the availability of the separate items. (There might be an incentive if items have much higher gross margin than the kit, but for the sake of simplicity, this case is beyond the scope of the present discussion).

In order to determine how many items should be preserved for kits (assembled or not), one can use an alternative quantile forecasts, where the service level is not set as a desired availability target, but on a much lower probability that reflects a probable sales volume that should be preserved.

For example, let's assume that a 30% service level on a kit gives a quantile forecast at 5. This value can be interpreted as “there are 70% chances that 5 or more units of the kits will be sold over the duration of the lead time”. If a 70% confidence in selling 5 kits outweighs the benefits of selling the next item now (assuming only 5 items remain), then the item should be considered as reserved for kitting purposes.

We are still only scratching the surface as far kits are concerned. Don’t hesitate to post your question in comments.

Hiring a Big Data developer

Many software companies advertise themselves as Big Data companies, but few have access to datasets as rich and exploitable as the ones that Lokad is processing on a daily basis. We are hiring a Big Data developer. As a software developer at Lokad, you will help us to design, implement and run our Big Data apps.

Hiring a Web Marketing Expert

The business of Lokad is booming, and we are now looking for a Web Marketing Expert. We want to keep developing our inbound marketing strategy, and this person will be reporting directly to the main shareholder of the company (that is, well, myself).

Quantum Forecaster unlocked

The Lokad facility where the first quantum forecaster prototype is hosted.Delivering more accurate forecasts has been the mission of Lokad since the very beginning. Today, after years of R&D efforts, we are extremely proud to announce the immediate availability of new quantum forecasting technology that delivers unprecedented forecast accuracy well beyond the performance routinely achieve by statistical forecasting technologies.

This technology leverages several key aspects of quantum mechanics, and in particular the Heisenberg reverse uncertain principle that states that there a fundamental limit to the imprecision of measurements made on the physical properties of particles such as position and momentum.

From a demand forecasting viewpoint, it means that by shaping the course of particles to follow the patterns of observed in say, a time-series representing sales over time, the laws of physics guarantee a minimal amount of precision on the measurements to be made when the particles are observed continuing their trajectory beyond the point representing the “present”. Those measurements represent the physical projection of the time-series, that is, in layman terms, the actual demand forecast. Instead of relying on statistics, the quantum forecast leverages the law of physics themselves to produce the forecast.

Schema of the Lokad quantum forecaster facility in Ufa, RussiaWhile the implications and benefits of quantum mechanics in terms of forecasting have been known to specialists for decade, the physical design of a quantum forecaster had remained an incredible challenge. A couple of years ago, Lokad opened a large scale facility in Ufa (Russia) to build such a device.

It took us a long time to get it built, and an ever longer time to get it properly tuned. In particular, the launch was delayed because quantum interferences caused by the Large Hadron Collider operated by the CERN which proved to be troublesome. The addition of several extra Higgs boson detectors did finally solve the problem. At present time, Lokad is actively working on the miniaturization of the quantum forecaster.

We don’t have a public pricing yet, but don’t hesitate to contact us.

Upload TSV files from the web

A few days ago,  we were announcing a file hosting service now properly named BigFiles. Today, we have just released a web interface that let you upload and download your files directly from the web.

FTP remains a preferable transfer mechanism for larger files or when there are many files involved.

FTP hosting, push your files to Salescast

Since December 2012, Salescast supports importing TSV files. However, until now, Salescast was expecting you to plug your own FTP server to retrieve those files. We felt this was an unnessary complication.

Indeed, while there are a myriad of file hosting services available on the web, we have found that most of them are simply not good at supporting business data transfers: annoying limits are encountered with:

• the maximal number of concurrent connections, 
• the maximal file size,  
• the maximal bandwidth, 
• ... 

Thus, we decided to roll our own.

We are proud to announce the immediate availaility of our FTP hosting service. Upload and download files from Lokad. The Express Plan comes with 1GB of free storage and 1GB of free bandwidth (per month). This service is compatible with Salescast and the other apps of Lokad.

Technical nugget: In order to deliver maximum scalability and reliability, this service is built on top of Windows Azure - like all the other technologies developed at Lokad. The architecture schema below illustrates how we scale out the workload on multiple virtual machines.

Subtleties in lead time calculations

The lead time is one of the two variables, along with the service level, that are essential to optimize inventory. Yet, calculating the correct lead time can be somewhat tricky because of start-of-day vs end-of-day variations, business days vs calendar days, reordering delays, etc 

The illustration shows a store reordering every 3 days with a supplier lead time of 4 days. The applicable lead time in this case is 7 days.

Check out our new tutorial for lead time calculations.

Stay tuned for more.

Machine Learning and Big Data talk at TechDays 2013

Last week, we had the chance to speak to an audience of roughly 3000 people attending the Machine Learning and Big Data keynote at the Microsoft TechDays 2013 in Paris. A special thanks to Bernard Ourghanlian for making this possible.

Our client, Pierre-Noel Luiggi (Founder and CEO of Oscaro) was also present - and a formidable support.

For those who could not attend the event, check the video of the session (15min, in French).

Phantomscan, get rid of phantom inventory

Delivering more accurate demand forecasts has been the goal of Lokad since its creation. However, no matter how good are the forecasts, if underlying inventory records are inaccurate, the whole inventory optimization, as delivered by Salescast, is off.

A bit more than one year ago, we were launching Shelfcheck, a tool to detect out-of-shelf (OOS) problems based on the analysis of improbable sales drops. However, Shelfcheck is only palliative care as OOS alerts are produced after the beginning of the stockout problem.

Today, we announce the launch of Phantomscan, a new webapp that help retailers getting rid of their phantom inventory. In short, by analyzing patterns observed in past inventory corrections, Phantomscan predicts which SKUs are most likely to have inaccurate records. Instead of having employees performing a classic cycle count, they focus directly where counting is needed the most.

There are not that many empirical studies that have been conducted about inventory accuracy in retail, however the few that exist are stunning: in all studies, inventory records have been found to wildly inaccurate at the store level.

Although computerized tracking of inventory at the stock keeping unit (SKU) level is commonly assumed to be accurate, we found discrepancies in 65% of the nearly 370,000 inventory records we gathered from multiple stores of a leading supply chain. DeHoratius and Raman (2004)

The problem is serious because phantom inventory acts as an invisible hand lowering service levels on all SKUs impacted. Yet, it seems that the only option remains cycle counting which happens to be a very expensive way to improve inventory accuracy.

Phantomscan, in contrast, is designed to make the most of each minute spent on counting inventory. It's a curative care against OOS, helping the retailer to remove inaccuracies before OOS problems emerge. With an aggressive per-store pricing, we believe that Phantomscan will be suitable for retail companies of any size.

At this point, we are looking for volunteers to take part in Phantomscan beta. Early adopters will benefit from Phantomscan free of charge for the duration of the beta, plus 6 months after the end of the beta. Furthermore, beta users will also get the chance to influence the development of the webapp toward the features that serve them the most. 

To take part in the beta, just drop a line to contact@lokad.com.

Roadmap for 2013

As we did before (see roadmaps of 2010, 2011 and 2012), let's share some insights about future developments of Lokad for 2013.

Forecasting Technology

Improving the accuracy of our forecasts remains a key priority. In 2012, our progresses took an unexpected twist, as we did stumble upon quantile forecasts. Forecasting benchmarks that we performed over 2012 with quantile vs classic have systematically shown that quantiles were vastly ahead accuracy-wise. When it comes to inventory optimization, quantile forecasts vastly outperform classic forecasts.

At this point, we believe that quantile forecasts (or their descendant) will be the de facto standard in 10 years from now for all retailers and manufacturers interested in inventory optimization. In particular, at the store level, quantile forecasts make the difference between a forecast that just work vs a forecast that doesn't.

In 2013, we intend to keep improving our forecasts, but from a more financial angle: it's no more about reducing percents of forecast error, but rather about reducing Euros or Dollars of forecast error. This is a change of paradigm: it means that we will bring our statistical models closer to the business, as opposed of chasing further decade's old classic forecasting accuracy metrics.

Salescast

Our flagship webapp has undergone many improvements in 2012 with dozens of quiet upgrades. Overall, Salescast has evolved from a pure demand forecasting app toward an inventory optimization app. The integration of quantile forecasts, which are key to compute both optimized reorder points and optimized reorder quantities, have been the primary evolution of Salescast in this area.

Then, the user interface has been extensively revisited (again), if you are not a regular user, don't hesitate to look at the latest screenshots. In particular, Salescast now provides a lot more feedback to support troubleshooting data integration problems.

Also, flat TSV files exposed over FTP can now be imported into Salescast. Our classic SQL-based data format remains available; however, for a growing number of companies, flat text files seem to be easier and more scalable than SQL. However, data in flat text files can lead to all sort of data corruptions; one goal for 2013 will be to make Salescast very resilient against this type of problems.

As far pricing is concerned, in 2012, we had the first major evolution in years with the introduction of both Express and Enterprise subscription plans. Salescast has now a freemium business model, with a free version (Express, no support) and a paying version (Enterprise, dedicated support). In 2013, we remain committed in preserving and extending further this approach. The Express plan will remain free of charge; while we extend the levels of support we deliver for our paying subscription plans.

New: FTP as a Service

The File Transfer Protocol (FTP), while dating from the early 70's, remains one of the most simple and powerful way to move data over the Internet. Furthermore, in 2012, we have noticed a strong resurgence of the use of flat text files, as opposed to relational systems which tend to fail to deliver the desired performance levels. This observation lead us to support for TSV data over FTP in Salescast.

Then, in order to relieve our clients from the burden of managing their own FTP servers, we have started to develop an app that will deliver FTP as a Service, leveraging Windows Azure for reliability and scalability. This service will made available for all Lokad account holders, much like Salescast.

As far pricing is concerned, we plan to follow a freemium approach: FTP hosting will be made available for free (within limits) to Express users, and Enterprise subscribers will benefits from more storage and bandwidth.

New: Stockwatch

While Salescast delivers inventory optimization, we realize that most our SMB clients are lacking of inventory intelligence tools; that is, tools that let them survey, monitor, investigate stocks and stock-outs. Classic business intelligence software are not satisfying because they require too much setup work to start delivering anything useful for advance inventory management.

Thus, we have started to develop Stockwatch, a new webapp that delivers inventory intelligence, i.e. extensive reports outlining inventory performance. Stockwatch is intended as fully complementary to Salescast, but it will be strictly independent. Hence, it will be possible to use either Salescast or Stockwatch, or both.

A freemium version of Stockwatch is also anticipated, with a free version for Express users at Lokad.

New: Phantomscan (update: 2013-01-21)

This webapp is has been announced separately.

New: (Big) Data Platform

Big Data has been a major buzzword in 2012, and for Lokad, it's a been a stream of Big Data consulting missions. We will keep leveraging our experience as a software company operating Big Data analytics software to help other companies ramping up their own capabilities.

In particular, we have started to develop a new open source product named Lokad Data Platform (see also the infographic). In short, our (Big) Data Platform gathers patterns, practices and software bricks that come very handy to service large amount of data for advance analytics purposes.

With less than 4000 lines of code, one of the strengths of the Lokad Data Platform is its minimalism. Hence, we plan to keep the codebase small and clean to keep the product attractive to fellow developers. However, we intend to keep adding more samples, use cases and documentation.

Shelfcheck

Our out-of-shelf monitoring webapp has been steadily developed during 2012. Since it addresses a relatively narrow segment of large food retailers, we have no plan, to date, to make the webapp self-serving (unlike Salescast and most of the other webapps of Lokad). However, the data format expected by Shelfcheck is now publicly documented. Shelfcheck favors flat files exposed through FTP, hence it will benefit too from the FTP service that will soon be released.

In 2013, we intend to push our OOS analytics technology further to refine again the OOS detection trilemma.

This roadmap isn't carved in stone. Don't hesitate to voice your opinion.