Pricing

Repricing Experiment – Part 2: The Results

Today’s blog post is long overdue.  Many of you have been asking (pestering and nagging may be more accurate terms, but I finally listened!) for an update on my repricing experiment, and I’m excited to share those results with you.  So… let’s bring on the data!

Setting the Stage:  If you’d like to read the full back story about the repricing experiment, you can check it out here.

For those who would prefer the Cliffs Notes, here they are:

  • Sample size – 2 batches of 100 books each.
  • Matched pairs – I assembled similar batches of books to give an accurate comparison. If I found two books ranked 1.2 MM that I would price at $14.99, I split them apart and put one in each batch.
  • Batch metrics:
  • Batch 1 – average rank: 678k, average list price: $16.50
  • Batch 2 – average rank: 650k, average list price: $16.46
  • Blind pricing – I priced both batches as I normally would, without knowing which batch would be the control group and which one would be repriced automatically.
  • Coin flip – A 1985 US quarter was used to select which batch would be repriced, and Batch 2 won the toss and elected to receive. Tails never fails!
  • Repricer settings – RepriceIt was the software I selected, based on feedback from dozens of other booksellers I trust. I employed basic settings as follows:
  • Inventory was repriced three times a day (morning, noon, and night)
  • Software analyzed FBA offers only, MF prices were ignored
  • Condition was ignored, to give a better chance for the software to actually “see” competitive offers (plus I don’t believe condition matters all that much)
  • Floor price was set at $7.99

Hypothesis:  A quote from my original post: “My hunch is that I will sell more books out of Batch 2 but I will sacrifice price to do so.  The question at the end of this experiment is whether or not the loss of price is worth the additional unit sales.”

My hypothesis factored in Amazon’s API limitations, which mean that if there are no FBA offers in the lowest 20 offers, then the API won’t return those values.  Thus, because I wanted to compare my prices to FBA offers only, in many cases I figured the repricing software wouldn’t be able to “see” any competitive offers and wouldn’t be able to reprice many of my items.

I do believe a repricer plays a crucial role in moving older inventory.  If a book hasn’t sold in six months, you can compare it to MF offers in an effort to flip it quickly, even at a lower price.  Now that Amazon has removed the single-unit exemption for long-term storage fees, it’s critical to sell older inventory instead of waiting for the right Prime buyer to come along.  Although the long-term benefits of using a repricer are clear to me, I believed there would be minimal, if any, short-term benefits.

Experiment Results:

Units sold over time – I thought Batch 2 would sell more items than Batch 1, but perhaps only marginally so.  Here’s the chart of units sold over the first six months:

screen-shot-2016-12-02-at-12-45-56-pm

As you can see, Batch 2 beat out Batch 1 early and often.  By the end of week four, Batch 2 had nearly double the sales of Batch 1 – 28 compared to 15.  Even after three months, Batch 2 had 25% more sales than the non-repriced batch.  It’s hard to argue with those metrics.  So far, the repricer seems to be earning its monthly fee.

But surely the repricer would have to drop prices significantly in an effort to earn those extra unit sales, right?  Let’s take a look.

Price change over time – If the repricer gave up too much price, it could potentially wipe out any gains experienced from the increased unit sales.  After the first week, I thought my hypothesis was going to be spot on, as the repriced batch had sold 6 more books than the control group, but had dropped prices by 5% on average to generate those sales.  Go me!  But as I was in the middle of patting myself on the back, the week two results came in.  That week, the repricer actually earned me an extra 9% compared to my original list prices, which brought the cumulative price change up to a positive 1%.  Wowzer.

By the end of three months, the repricer had actually earned me an additional 1% compared to my initial pricing.  Here’s how the chart shook out.  You’ll notice that in months five and six the prices had begun to drop a bit, which is to be expected with older inventory:

screen-shot-2016-12-02-at-12-46-15-pm

How many items were actually repriced?  After 3 months, Batch 2 had sold 53 items.  Only 32 of those had different prices than the initial list price.  This means that 21 items – or 40% – were NOT repriced, either because the software agreed with my price or because it couldn’t “see” enough FBA data to reprice them.

Out of the 32 items that were repriced, here are some interesting insights:

  • 11 items were sold at higher prices, for a total gain of $52.95 ($4.81 per book increase)
  • 21 items were sold at lower prices, for a total loss of $43.82 ($2.09 per book decrease)
  • Net price INCREASE of $9.13 overall

So not only did I sell MORE items, but I actually sold them at an overall HIGHER price than my original list price!  There goes my hypothesis…

Total sales over time – Let’s look at how the sales stacked up over time between the two batches:

screen-shot-2016-12-02-at-12-46-24-pm

Batch 2 had $120 in extra sales compared to Batch 1 after the first four weeks.  After six weeks, Batch 2 had an astonishing $206 in additional sales.  Those increased sales mean you can turn a profit faster and pull more money out to buy more books, pay your mortgage, etc.  And honestly, apart from Notorious B.I.G., who doesn’t want more money?

Recommendations:  This data really blew my mind.  I had been holding firm to my beliefs that the API limitations would render repricing software virtually useless.  And even once I saw the results, I was still trying to find reasons to disprove them.  It’s a small sample size, with only 100 items in each batch, but it’s hard to argue with the results from my experiment.

I still believe it’s important to price your items carefully at the time of listing, since 40% of the items sold were not repriced at all.  But I intend to challenge that assumption as well by running another test where I price a batch at $200 and allow the repricer to kick in from there.  Stay tuned…

What are YOUR experiences with repricing software?  Does this data make you reconsider your strategies?  For me, it does.  Remember to trust the data.  If not, your mind can easily lead you astray and cause you to mistakenly hold on to your misguided ideas.

Disclaimer:  I am not affiliated with any repricing software, and make no commissions or referral fees if you choose to sign up for any of them.  I’m writing this article based solely on my experience with the software from RepriceIt.com, and I desire to share those results with a wider audience.  If the founder of RepriceIt wants to reward me with a branded t-shirt, or a lifetime supply of golf tees, I wouldn’t turn either offer down!

Related Posts