What 78 Mid-Market WMS Buying Conversations Reveal

Most published research about warehouse and inventory software is written by the companies that sell it, and it usually says the same thing: buyers want efficiency, visibility, and scale. That is true and useless.

We had access to something more specific. Between February 2025 and August 2026 we recorded 83 usable sales conversations with mid-market eCommerce merchants who were actively evaluating new inventory, order, and warehouse management software. These were not surveys. They were unscripted, hour-long conversations in which an owner or operations lead explains, in their own words, what is broken, what they run today, and what they can afford.

Five of the 83 are not buyer conversations at all: an internal call, a platform-partnership call, a systems-integrator call, a scoping call with an existing engagement, and a consultancy introduction. Excluding those leaves 78 buyer-eligible conversations, and 78 is the denominator behind every number on this page. Those 78 transcripts carry 438,801 spoken words, an average of 5,626 per conversation. The raw files come to 466,535 tokens, but 27,734 of those are the timestamp links and speaker labels attached to 10,214 turns, and those are not speech.

The headline counts, buyer speech only: buyers raised integrations or APIs in 58 of 78 operations (74%), more than any other family we counted. 51 of 78 named a specific eCommerce platform, yet only 17 of 78 named any dedicated inventory or warehouse management vendor. 21 of 78 described a spreadsheet as live infrastructure, and 37 of 78 called part of their own process manual.

What changed on 2026-09-07: the corpus was repaired, the denominator moved from the old 76-file count to the 78 buyer-eligible conversations, every figure was recounted from its published pattern, and figures that could not be recounted were removed rather than restated.

Methodology, up front

The sample. 83 usable recorded sales-demo conversations spanning 2025-02-17 to 2026-08-03, of which 78 are with prospective buyers of inventory, order, and warehouse management software. The other five are named above and are excluded from every count.

The unit of measurement. Every figure is conversation-level: "X of 78" means the term appeared in X distinct conversations. An operation that says "barcode" forty times counts once. Both columns, buyer and either-speaker, run over the same 78 conversations, so they are directly comparable.

Buyer counts versus either-speaker counts, always labeled. Each transcript is split into speaker turns, and turns belonging to SkuNexus staff are excluded from any buyer count, so a term only counts when a buyer said it, not when a salesperson pitched it. Figures that include vendor speech are labeled "either speaker" or "came up." The columns disagree in instructive ways: cycle counting comes up in 39 of 78 conversations, but buyers themselves raised it in only 14.

Literal counts only. Software can reliably count whether a buyer said "barcode." It cannot reliably count whether a buyer "expressed frustration with vendor lock-in": interpretations drifted by 25 to 30 percent when spot-checked. Every figure on this page is a literal count of concrete language, produced by the regular expression published beside it in the appendix, and every literal figure is a lower bound by construction. The correct reading is always "at least N of 78 said so," never "only N of 78 do this."

What this sample is not. These are merchants who had already decided to evaluate new software, so they are self-selected toward dissatisfaction with their current setup. Each conversation was with a single vendor. This is a benchmark of mid-market buyers in an active buying cycle, not a census of all merchants and not a neutral product comparison.

Part one: what mid-market operations are actually running

Integration is the most raised subject: 58 of 78 buyers

Buyers in 58 of 78 (74%) raised integrations or APIs, making it the single most frequently raised concrete subject in the corpus. Nothing else came close. The question rarely arrived as "do you integrate?"; it was almost always "do you integrate with this named thing that I already run and am not replacing." Buyers in 22 of 78 (28%) used the word API directly.

The storefront is the fixed point: 51 of 78 named their platform

Buyers in 51 of 78 (65%) named a specific eCommerce platform, and 41 of 78 (53%) named Shopify. It is named more often than any warehouse system, accounting system, or inventory vendor: operators describe their stack starting from the platform that takes the order, and work backwards.

The ledger is QuickBooks more often than it is an ERP

Buyers in 25 of 78 (32%) named a specific accounting or finance system, and 20 of 78 (26%) named QuickBooks. Counting either speaker, QuickBooks appears in 26 of 78 conversations, usually as the system of record any new platform must integrate with, sometimes as the only inventory tool in the business. As one merchant put it about QuickBooks for inventory: "it's not very good." Buyers in 18 of 78 (23%) named an ERP of any kind, the whole NetSuite, SAP, Sage, Dynamics, Epicor, Acumatica, Infor and Odoo family combined, and the enterprise names appear mostly as the rejected "too expensive, too legacy" pole of the decision. The ledger these operations reconcile against is more often a small business accounting package than an enterprise suite.

Amazon is an operating constraint, not a channel

Buyers in 31 of 78 (40%) named a marketplace, and 29 of 78 (37%) named Amazon specifically, more than every dedicated inventory software vendor combined. For an operations team a marketplace is not a sales channel; it is a set of packaging rules, label rules, and cutoff times the warehouse has to satisfy.

The spreadsheet is still load-bearing: 21 of 78

Buyers in 21 of 78 (27%) referred to a spreadsheet, Excel, or Google Sheets while describing how their operation currently runs. Widen the pattern to include manual-work language ("manually," "by hand," "typing it in") and the literal buyer union reaches 44 of 78 (56%). Many merchants never say "spreadsheet"; they say "we do everything by memory," and the manual reality sits underneath. That union figure is also the evidence behind the buyer section of our guide to custom inventory management software.

The detail is more instructive than the count. A UK operation running two warehouses kept everything in Google Sheets, deducting stock by hand. A B2B parts supplier types returns onto a spreadsheet one at a time, never scanned. A twelve-year Amazon seller answered the question of what software manages its warehouse, in full: nothing, Google Sheets. None of the three is a small business.

"We're using Google Spreadsheet. So it's very cumbersome, it's very manual. Sometimes inventory goes missing, even though it's physically available... And sometimes inventory is gone and it's still represented on the sheet."
- perishable-meat food company

The category incumbent is missing from the buyer's own story: 17 of 78

Buyers in 17 of 78 (22%) named any dedicated inventory or warehouse management vendor, against 51 who named an eCommerce platform. It is the most structurally interesting number in the study, and it is not a market-share statement: the platform is named, the ledger is named, the marketplace is named, and the inventory system, more often than not, is a set of manual habits stitched between them.

Counting either speaker against a broader tracked list that adds shipping tools, a third-party tool is named in 42 of 78 conversations. The most-referenced point solution is ShipStation, named by buyers in 15 of 78, and in several conversations it appears as an incumbent pushed past its intended purpose, used as an order-management hub rather than a shipping tool.

Part two: what the work looks like on the floor

Scanning is the most discussed floor practice: 51 of 78

Buyers in 51 of 78 (65%) used barcode or scanning language, with 41 of 78 using the noun barcode itself and 19 of 78 referring to scanner hardware directly; either speaker, the topic comes up in 65 of 78. It runs in two registers: operations asking whether the software can generate barcodes for items that never had them, and operations that already scan asking about harder cases, like barcoding a location rather than a product. Our guide to barcode scanner inventory management software starts there.

Nearly half describe their own process as manual: 37 of 78

Buyers in 37 of 78 (47%) described part of their process as manual, done by hand, or typed in. That this coexists with the platform finding is the core tension of mid-market fulfillment: the front end is current, the back end is a person retyping.

Inbound is not a footnote: purchase orders 28, receiving 27

Buyers in 28 of 78 (36%) discussed purchase orders and 27 of 78 (35%) discussed receiving or putaway, frequently wanting the receiving scan and the purchase order to be the same event rather than two records reconciled later.

Bins, kits, and the shape of the exceptions

Buyers in 24 of 78 (31%) discussed bins, bin locations, aisles, or slotting, and the recurring request underneath is permission to break the assumption that one SKU lives in one place.

Buyers in 20 of 78 (26%) discussed kitting, kits, bundles, or bundling, across unrelated verticals. One protein producer described virtual kitting: the sale can be composed any number of ways, but the pick is against the true underlying items. That distinction, between how something is sold and how it is picked, is the practical definition of the problem.

Multi-warehouse: 19 of 78 buyers describe it, in a corpus where 68 say "warehouse"

Buyers in 19 of 78 (24%) described operating more than one warehouse or location, in a corpus where 68 of 78 buyers used the word warehouse at all; either speaker, multi-warehouse phrasing comes up in 50 of 78. The descriptions were rarely tidy: one contractor runs four warehouses plus a corporate office; another splits inventory across two buildings by channel and wanted routing to respect the split. Raised by vendors far more often than volunteered by buyers, multi-warehouse reads as a near-future requirement more often than a present fire.

Cycle counting: the clearest vendor-led topic in the corpus

Buyers in 14 of 78 (18%) raised cycle counting or physical inventory. The same language appears in 39 of 78 conversations once SkuNexus staff turns are included. That gap is why this page counts buyer speech separately: it is the difference between what vendors like to discuss and what buyers arrive wanting to discuss.

Every topic, buyer column versus room column

Where the room column towers over the buyer column (customization, dropshipping, purchase orders, cycle counting), the topic is substantially vendor-led. Where the columns sit close (spreadsheets, ShipStation, direct-to-consumer language), the buyer owns it. Overselling was raised by buyers in only 9 of 78, but when it is a pain, it is visceral:

"Main challenges are overselling... We think that we have three items of a specific SKU... when we actually go to that unit to pick up the stock... we find only two, we find only one."
- furniture retailer

One more pattern worth isolating: vendor neglect of a critical integration is a churn signal, not just a pain. As one hardware merchant put it about an incumbent: "their integration exists, but when things break, they break hard."

Part three: what buyers ask the software to do

Buyers in 30 of 78 (38%) used customization language, asking whether something could be customized, configured to their process, or built to their specification. Either speaker, the language appears in 72 of 78, which mostly measures how often vendors say it.

Buyers in 19 of 78 (24%) described wholesale, B2B, or distribution activity, and 11 of 78 (14%) used direct-to-consumer language. The mixed-model operation is normal here: the same building ships a pallet to a retail account and a single unit to a consumer, and the two flows want different paperwork, packing, and carriers.

Buyers in 21 of 78 (27%) discussed dropshipping or supplier direct shipment, and 14 of 78 (18%) referred to their own developers, engineers, or programmers, in at least one case because the person on the call had personally written the warehouse software the company was running. That last group is the one that asks what happens at the code layer, which is the subject of our post on source code access in a WMS.

How buyers talk about budget

This page publishes no price figures, ours or anyone else's, and no count of price talk. What it can report is the shape of the conversation, in buyers' own words.

The reference point is almost always what the buyer already pays for a starter tool or a stack of subscriptions, not a category benchmark. A fast-growing Magento retailer, weighing a platform against the tools it had assembled:

"The question is, just biting it off right at the bat when, you know, it hasn't, quote unquote, been in the realm of what we've been paying right now is you got to just factor everything in."
- Magento retailer, multi-warehouse, 2025-10-20

The same instinct shows up as arithmetic rather than objection. An operations lead at a water treatment equipment retailer, adding up the tools a single platform would replace:

"They're coming in close to what we're currently paying. when you compile everything on one another versus it, this is one shot go."
- water treatment equipment retailer, 2025-05-21

The second pattern is structural rather than numeric. Buyers want to know whether the meter moves with their order count, and they ask it directly, unprompted, before any figure has been discussed:

"So you guys are more or less just one flat fee? It's not like, you know, up or down based off orders is kind of what you're getting at there?"
- junkyard operator, 2025-06-04

Worth being precise about that last one, because our own materials have been sloppy here: the memorable line about being penalized for growing was said by a SkuNexus salesperson, not by a buyer. What the buyer said is the question above. The question is the finding; the slogan is not. To put your own numbers against your current process, use the WMS ROI calculator, which shows every formula.

Volume

Buyers in 32 of 78 (41%) stated a concrete order or shipment volume using the strict noun grammar published in the appendix. Looser grammar reaches 49 either-speaker, but it also matches money talk, so the strict pattern is the one to cite. We previously published a median and a range for those stated volumes; neither could be recomputed from the transcripts, so both are gone rather than restated.

Takeaways

If you are a merchant evaluating software

1. Your integration list is your requirements document. Integration was raised by buyers in 58 of 78 operations, more than any feature. The storefront and the accounting package are the systems nobody plans to replace; everything else has to fit around them.
2. The spreadsheet is not a beginner tool. It is usually there because it is the only layer flexible enough to hold the part of the process the real systems refused to model.
3. Scanning is the highest-leverage physical change available. It converts a manual step into a recorded event, and it is the prerequisite for reliable counts and location accuracy.
4. Buy for the exceptions. Kitting, multi-location, dropshipping, and mixed B2B and DTC all describe the same situation: the exceptions are the business. Ask how the software behaves at three to four times your current volume before you sign.

If you are a vendor or an analyst

• The buyer's baseline competitor is a spreadsheet, not another platform. Only 17 of 78 buyers named a dedicated inventory or WMS vendor at all. The incumbent is manual process, and the real switching cost is behavioral.
• Buyers judge a quote against their current spend, not against other platforms. Three of them do it out loud in the section above.
• Count buyer speech, not room speech. Cycle counting looks like a 39-conversation topic until vendor turns are excluded, and then it becomes 14. Sales-call mining that does not separate speakers will mistake its own pitch for demand.

Limits, stated plainly

This is a sample of operations that agreed to a vendor conversation, not a random market sample, and it skews toward operations where something was already broken. Literal counts undercount every behavior they measure. Percentages are for readability only; with n=78, a difference of two or three conversations is not meaningful, and no claim here depends on one. The five excluded conversations are excluded from both columns, not just the buyer column.

Cite this study

If you reference these figures, please use this form, so the numbers stay attached to their method:

SkuNexus Mid-Market WMS Buying Benchmark. A review of 78 buyer-eligible conversations, drawn from 83 usable recordings with mid-market eCommerce and distribution operations, February 2025 to August 2026. Figures are literal buyer-language counts at the conversation level and are lower bounds, except where labeled either-speaker. https://www.skunexus.com/blog/mid-market-wms-buying-benchmark

Methodology appendix: every number and the pattern that produced it

All patterns are case-insensitive regular expressions applied at the conversation level, n=78; the buyer column counts buyer turns only, the either-speaker column counts all turns in the same 78 conversations. Word boundaries are enforced, so "Excel" does not match "excellent." Reproduce any row with node scripts/corpus-verify.mjs "<pattern>". Every row below was recomputed against the 78 buyer-eligible transcripts on 2026-09-07; no figure on this page comes from anywhere else.

Families deliberately not published as counts. Vendor abandonment, inability to see inventory, and fear of implementation risk are real themes in these transcripts that cannot be counted from literal language. They are not published, in any form. Nor is any count of price or budget language, under a standing rule that this site publishes no pricing signal, ours or a competitor's.

Corrections log

1. Denominator. Every figure on this page previously ran against a 76-file corpus. Seven recordings that had never been captured were added in August 2026, and five non-buyer conversations were identified and excluded. The corpus is now 83 usable recordings and 78 buyer-eligible conversations, and every number above was recomputed from its own published pattern against those 78.
2. Spreadsheet count. This page's own spreadsheet figure previously read 26 and was never reproducible from its pattern. The literal buyer count is 21 of 78 (27%).
3. Extraction figures removed. A section of this page reported figures produced by model extraction rather than string matching, including several that other pages cite. None of them can be recomputed from the transcripts by any published method, so all of them are deleted rather than relabeled. If you arrived here from a page quoting an extraction figure, that page is wrong until it is corrected.
4. Barcode and scanning. Our barcode software guide once cited barcode and scanning at a figure matching neither column. Buyer-only is 51 of 78, either-speaker 65; the guide now carries both.
5. Cycle counting. Our inventory management KPIs post cited the either-speaker cycle counting figure as though it were a buyer count. The buyer-only count is 14 of 78. Corrected: that post no longer publishes a cycle counting figure.
6. Corpus size in words. This page previously said the 78 transcripts run to 466,535 words at an average of 5,981. That number counts the timestamp link and speaker label on every turn as words. Spoken words are 438,801, an average of 5,626, and that is what the sentence above now reports.
7. Budget attribution. The line about being penalized for growing was quoted in internal materials as a buyer quote. It is a SkuNexus salesperson's line. Corrected above.

A note on method and motive

SkuNexus recorded these conversations in the ordinary course of its own sales process, which is both the reason this dataset exists and its clearest limitation. We counter that bias by counting rather than cherry-picking, by counting buyer speech separately from our own, by publishing the pattern behind every number, and by deleting figures we cannot reproduce instead of softening them. If this page helps a buyer ask a sharper question, or an analyst check a lazy assumption, it has done its job.

Yitz Lieblich is the Founder and CEO of SkuNexus. He has worked in eCommerce since 2007, founded Web Solutions NYC in 2007, and founded SkuNexus in 2018.