A chaos-to-clean video is a B2B product video whose script states a messy before-state of work, such as manual work, spreadsheets, silos, overload, errors or busywork, and then offers the product as the fix. 118 qualified, from 108 companies including Atlassian, Plaid, SAP and Zoho, uploaded 2 June 2023 to 14 July 2026. 99 of 118 state the mess before naming the product, at a median 0:07 against 0:17 for the name, and 62 name silos or disconnected tools. On the 31 with a full visual record, 4 frameworks clear the floor: kinetic type, diagrams and illustration before the name, and the product interface after it, in 23 of 31. Whether any of it demonstrates productivity is the reader's test: the files record what the script says and what is on screen, not what a viewer concluded.
ADVIDS Research · 118 B2B product videos coded · 108 companies · 31 with a full visual record · 279 sampled frames · published 14 September 2026 · 20 min read
Top takeaways
Chaos-to-clean scripts state the mess first, the product second
99 of 118 chaos-to-clean B2B product videos uploaded 2 June 2023 to 14 July 2026 state the mess before naming the product, at a median 0:07 against 0:17; 19 name the product first, among them Zendesk, Atlassian and Worldpay.
Two in three scripts give the mess a tenth of their words
Chaos-to-clean B2B scripts that open on the mess give a median 19 per cent of their words to it before the product arrives; 78 of 118 spend a tenth or more and 31 a quarter or more, Swimlane and Exabeam among them.
Silos lead the mess; two in three scripts name more than one
Silos or disconnected tools lead the mess in chaos-to-clean B2B videos from 2 June 2023 to 14 July 2026: 62 of 118 name them, against 44 each for overload and busywork; 81 name two or more kinds, as Odoo, Procurify and Tungsten do.
Before the name, type, diagrams and illustration carry the mess
On the 31 chaos-to-clean B2B videos with a full visual record, 19 put kinetic type, a flowchart or network diagram, or illustration, abstract or b-roll imagery on screen before the product is named; kinetic type leads at 12, as in Orange Business and Plaid.
After the name, the product interface takes the screen
After the product is named, 23 of the 31 visual-record chaos-to-clean videos put a product interface on screen and 18 a dashboard, as SAP, Celigo and Rillion do; an interface appears before the name in 13.
Every video is narrated, runs a median 1:25, rarely goes dark
All 118 chaos-to-clean B2B videos uploaded 2 June 2023 to 14 July 2026 carry a voiceover, the median runs 1:25, and 92 of 118 sit on light or mixed frames against 26 on dark frames, Exabeam and Swimlane among them.
The mess comes back after the product in 64 of 118 scripts
A messy-state term comes back after the product is named in 64 of 118 chaos-to-clean B2B scripts: 17 of the 19 that name the product first, against 47 of the 99 that open on the mess, with Atlassian and Zendesk among them.
Questions this report answers
What is a chaos-to-clean B2B video?
A B2B product video whose script states a messy before-state of work, such as manual work, spreadsheets, silos or overload, and then offers the product as the fix. ADVIDS Research found 118 among 945 screened product videos, from 108 companies including Atlassian, Plaid, SAP and Zoho, uploaded 2 June 2023 to 14 July 2026.
Should a B2B product video show the problem before the product?
99 of 118 chaos-to-clean B2B videos do, and 19, including Zendesk and Worldpay, name the product first and bring the mess in at a median 0:20. The files record the order, not which order works better.
How much of an explainer script should cover the problem?
In chaos-to-clean scripts that open on the mess, the stretch before the product's name takes a median 19 per cent of the words, about 33 words or 0:14 of runtime. Paycom's personnel-forms spot spends 80 per cent.
What problems do B2B software videos show?
Silos and disconnected tools lead, named in 62 of 118 chaos-to-clean scripts, ahead of overload and busywork at 44 each, manual work at 42, spreadsheets, inboxes or sticky notes at 19 and errors at 17.
How do B2B videos show a messy workflow on screen?
On the 31 chaos-to-clean videos with a full visual record, 19 show kinetic type, a diagram or illustration before the product is named, while stock office imagery appears there in 6. After the name, 23 show the product interface.
Do chaos-to-clean videos use a voiceover?
All 118 do, because the rule reads a spoken script; the median video runs 1:25 and carries 164 words.
How was this analysis done?
ADVIDS Research screened 945 product videos against a rule and a lexicon written before intake, read all 212 candidate transcripts and qualified 118 from 108 companies. Timing came from captions and speech segments, frameworks were coded on the 31 with a full visual record, and the coded dataset ships with the report.
How to read this report
This report catalogues 118 B2B product videos from 108 companies, uploaded to YouTube 2 June 2023 to 14 July 2026, whose scripts state a messy before-state of work and offer the product as the fix. A figure such as 99 of 118 counts videos against a base. Script, sound and frame findings stand on all 118; the visual frameworks stand on the 31 with a full visual record, and each figure names its base. Mess first and product first give the order of two hand-picked sentences, the order known in sales as problem-solution selling. There are no periods, so nothing here describes change over time.
Finding 01 · Script · Order · 118 chaos-to-clean B2B videos, 2 June 2023 to 14 July 2026
1. Chaos-to-clean scripts state the mess first, the product second
99 of 118 chaos-to-clean B2B product videos uploaded between 2 June 2023 and 14 July 2026 state the messy before-state before they name the product. The first mess lands at a median 0:07, 7 per cent of the way into the words, and the product at 0:17, 24 per cent in. In the mess-first scripts the mess arrives at 0:05 and the product at 0:21; 72 of 118 state the mess inside the first 10 seconds and 71 hold the name past 0:15. The 19 product-first scripts state the mess at a median 0:20. The product is sold as an answer, so the question goes first.
Mess stated before the product is named (99 of 118): companies in this group include Genesys, HPE, IBM, Plaid, etc. Without Genesys, one of the largest companies in the pattern, the share moves 0.3 points.
Zendesk goes the other way. Its June 2025 AI agents video names the product in its first line and holds the mess until 0:55, when other vendors' agents "leave you tangled in endless configurations". Where the mess belongs to a rival, the product can open the script and the contrast can wait.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
In the mess-first scripts, the product's name follows the first mess by a median 0:14.
Finding 02 · Script · Problem share · 118 chaos-to-clean B2B videos, 2 June 2023 to 14 July 2026
2. Two in three scripts give the mess a tenth of their words
In the 99 chaos-to-clean B2B scripts that open on the mess, the stretch from the first mess to the product's name takes a median 19 per cent of the words, about 33 words or 0:14 of runtime. Across all 118, 78 spend a tenth of their words or more there and 31 a quarter or more; 6 spend half or more, and 21 mess-first scripts spend under a tenth. Sentences that name the mess take a median 13 per cent of all words, and a tenth or more in 79 of 118. The mess gets a paragraph, not a slogan, and the product takes the rest.
A tenth of the words or more between the mess and the product (78 of 118): companies in this group include HPE, Moveworks, Paycom, FundApps, etc. Without HPE, the share moves 0.6 points.
Paycom goes the other way. Its April 2026 personnel-forms spot runs 0:55 and spends 80 per cent of its words on a missing raise, emailed requests and a botched transfer before one line names the product at 0:41. Where the mess is the joke, the product needs only the punchline.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
The middle half of the mess-first scripts spends 11 to 29 per cent of its words before the product.
Finding 03 · Script · Kinds of mess · 118 chaos-to-clean B2B videos, 2 June 2023 to 14 July 2026
3. Silos lead the mess; two in three scripts name more than one
62 of 118 chaos-to-clean B2B scripts uploaded 2 June 2023 to 14 July 2026 name silos or disconnected tools: data trapped across systems, too many tools, teams switching between apps. Overload and busywork follow at 44 scripts each, manual work at 42, spreadsheets, inboxes or sticky notes at 19 and errors at 17. 81 of 118 name two or more kinds and 25 name three or more; 13 name silos alone. The mess these scripts sell against is disconnection before drudgery.
Names silos or disconnected tools (62 of 118): companies in this group include HPE, Salesforce, Adverity, Celigo, etc. Without HPE, the silos share moves 0.8 points.
Octopus Deploy goes the other way. Its December 2025 spot names one kind of mess, errors, and no silos: before the product, deployments meant human error and late-night rollbacks, in a 89-word script that names Octopus in its first line. Where the product guards a single risky step, one kind of mess carries the contrast.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
Silos lead overload and busywork by 18 scripts; spreadsheets, the picture a brief reaches for first, appear in 19.
Finding 04 · Sound and frame · Frameworks before the name · 31 of the 118 with a full visual record
4. Before the name, type, diagrams and illustration carry the mess
On the 31 chaos-to-clean B2B videos that carry a full visual record, 19 put kinetic type, a flowchart or network diagram, or illustration, abstract or b-roll imagery on screen before the product is named. Kinetic type, lines of kinetic typography set on screen, appears there in 12; flowcharts or network diagrams in 10; illustration, abstract backgrounds or b-roll in 10. Title cards sit in the same window in 14 and a product interface in 13, and 25 of the 31 state the mess first. Stock office imagery appears before the name in 6. Before the product appears, the mess is written and drawn, not staged in an office.
Kinetic type, a diagram, or illustration, abstract or b-roll imagery before the name (19 of 31): companies in this group include IBM, Salesforce, Amplitude, Orange Business, etc. Without IBM, the share moves 2.7 points.
Rackspace Technology goes the other way. Its March 2026 financial-spreading video runs 1:19 and shows none of the three before the name: stock imagery of an analyst reviewing documents, then a close-up of spreadsheet numbers, while the voiceover says analysts spend hours extracting figures. Where the mess is a real document, the video shows the document.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
Kinetic type leads the three at 12; a product interface is already on screen before the name in 13.
Finding 05 · Screen · Framework after the name · 31 of the 118 with a full visual record
5. After the name, the product interface takes the screen
After the product is named, 23 of the 31 chaos-to-clean B2B videos with a full visual record put a product interface on screen: a dashboard in 18, a chat panel in 12. Before the name, an interface appears in 13. Among the 25 that state the mess first, 18 show the interface after the name. Data charts follow the name in 7, under the floor, and 2 of the 31 carry none of the four frameworks. The clean state is the product's own screen.
Product interface on screen after the name (23 of 31): companies in this group include HPE, IBM, Amplience, SAP, etc. Without HPE, the share moves 1.8 points.
Swimlane goes the other way. Its February 2026 spot runs 0:48 and puts no product interface on screen after the name: kinetic type and a logo card carry the close, as type carried the problem, "Thousands of alerts and dozens of disconnected tools". Where the product is agents working behind a security team, the clean state is stated rather than shown.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
The interface is mostly an after-state: 13 of 31 show one before the name.
Finding 06 · Sound and frame · Voice, length and frames · 118 chaos-to-clean B2B videos, 2 June 2023 to 14 July 2026
6. Every video is narrated, runs a median 1:25, rarely goes dark
All 118 chaos-to-clean B2B videos uploaded 2 June 2023 to 14 July 2026 carry a voiceover, which a rule that reads the spoken script makes close to certain. The median video runs 1:25 and the middle half 1:04 to 1:57; 60 run 0:46 to 1:30, 51 run past 1:30 and 7 run 0:45 or less. The median script carries 164 words at 126 words a minute. 92 of 118 sit on light or mixed frames and 26 on dark frames, with a median brightness of 139 on a 0 to 255 scale. Narration is the default in this set, not a choice.
Voiceover (118 of 118): companies in this group include Zoho, Genesys, HPE, ScienceLogic, etc. Without Zoho, the light-or-mixed share moves 0.6 points.
Exabeam goes the other way. Its March 2026 security operations video runs 3:14 on dark frames and spends 41 per cent of its 420 words before the product, opening on "Security teams are overwhelmed with data." A security product makes its case in dark frames and more than three minutes of narration.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
Dark frames sit with the mess-first scripts: 24 of 99, against 2 of the 19 that name the product first.
Finding 07 · Script · The mess after the name · 118 chaos-to-clean B2B videos, 2 June 2023 to 14 July 2026
7. The mess comes back after the product in 64 of 118 scripts
In 64 of 118 chaos-to-clean B2B scripts uploaded 2 June 2023 to 14 July 2026, a sentence after the product's name carries a messy-state term again, as a restated problem or as a contrast such as no more manual work. The split follows the order: 17 of the 19 scripts that name the product first return to the mess, against 47 of the 99 that open on it, a gap of 42.0 points. An explicit contrast cue, such as no more, instead of or say goodbye, follows the name in 22 of 118. The before-state keeps working after the product arrives.
Messy-state term after the product is named (64 of 118): companies in this group include Zoho, Salesforce, Adverity, Atlassian, etc. Without Zoho, the share moves 1.2 points.
KEBA goes the other way. Its May 2026 asset-management video runs 0:38 and 57 words: one sentence says assets are still handled manually, the next names KEBA, and the rest describes what the product does without naming the mess again. Where the script is four sentences, the before-state gets one.
Source: ADVIDS Research coded set, 118 videos from 108 companies, uploaded 2 June 2023 to 14 July 2026
Product-first scripts bring the mess back in 17 of 19; the mess-first scripts, which have already spent it, in 47 of 99.
Benchmarks by script order
Descriptive medians for placing your own script against each order. The product-first group holds 19 videos, so its medians describe a small group.
Timing, length and frames
| Measure | Mess first (99) | Product first (19) | All (118) | What it means |
|---|---|---|---|---|
| Runtime, median | 1:24 | 1:34 | 1:25 | Half the videos in the group run shorter than this |
| Words, median | 163 | 179 | 164 | Words in the transcript |
| Words a minute, median | 127 | 125 | 126 | Words over runtime |
| Mess stated at, median | 0:05 | 0:20 | 0:07 | Time into the video |
| Product named at, median | 0:21 | 0:09 | 0:17 | Time into the video |
| Mess position, median per cent of words | 4 | 21 | 7 | Share of the script before the mess sentence |
| Product position, median per cent of words | 28 | 8 | 24 | Share of the script before the product sentence |
| Problem share, median per cent | 19 | 0 | 16 | Words between the mess and the product; 0 when the product comes first |
| Frame brightness, median | 143 | 126 | 139 | Mean luminance over nine sampled frames, 0 to 255 |
What the set rarely uses
| Framework | Videos | Companies |
|---|---|---|
| Data chart after the product is named | 7 of 31 | Foyr, HPE, IBM, Plaid, etc. |
| Stock office imagery before the name | 6 of 31 | IBM, HPE, Rackspace Technology, SAP, etc. |
| Interface labelled with a messy term before the name | 5 of 31 | Otio, Rackspace Technology, Relativity, Rillion, etc. |
| Presenter on camera before the name | 4 of 31 | Deel, Nooks, SAP and WebWork |
| Split-screen or before-and-after label | 3 of 31 | Atlassian, Otio and Workday |
These are the literal pictures of chaos, each under the floor of 8 of 31; a spreadsheet, inbox or paper named on screen before the name is rarer still, at 1. The split-screen code reads segment text, so it can also catch a label such as side by side that is not a before-and-after at all.
What to do with this
- Finding 01 If 99 of 118 chaos-to-clean scripts state the mess first, at a median 0:07, open the draft on the customer's mess and name the product about 0:14 later.
- Finding 02 If mess-first scripts give a median 19 per cent of their words to the mess, cap the problem stretch near a fifth of the word count.
- Finding 03 If silos lead the mess in 62 of 118 scripts, test whether your buyer's mess is disconnection before defaulting to spreadsheets, named in 19.
- Finding 04 If 19 of the 31 visual-record videos write or draw the mess, brief the problem section as type and diagrams rather than stock office imagery, used in 6.
- Finding 05 If 23 of the 31 put the product interface on screen after the name, have a presentable interface ready for the second half of the video.
- Finding 06 If all 118 videos carry a voiceover, budget for a narrator from the first estimate.
- Finding 07 If 64 of 118 scripts name the mess again after the product, write one after-state line as a contrast with the opening.
Methodology
| Vertical (visual record only) | Videos |
|---|---|
| not recorded | 87 |
| fintech | 7 |
| itsm | 2 |
| retail | 2 |
| devtools | 2 |
| enterprise | 2 |
| cx | 2 |
| hr_tech | 1 |
| other verticals | 13 |
| Coding scheme | Values | Definition |
|---|---|---|
| Messy before-state | manual · spreadsheets · silos · overload · errors · busywork | Coded by reading the transcript: the kinds of disorder the script states as a condition a team, role, viewer or customer is in. A lexicon found candidates; a recorded call decided each. |
| Mess sentence and product sentence | sentence index · word position · seconds | The first sentence stating the mess and the first sentence naming the product or company as the fix, picked by reading. Position is the first word's index over all words; seconds come from word-timed captions where they exist and from interpolation inside timed speech segments otherwise. |
| Order | mess first · product first | Mess first when the mess sentence starts before the product sentence; within one sentence, an anchor phrase decides. |
| Problem share | 0 to 1 | Words from the first mess sentence up to the product sentence, over all words; 0 when the product comes first. |
| Mess-language share | 0 to 1 | Words in sentences carrying a messy-state term from the lexicon, over all words. |
| Mess after the product | yes · no | A sentence after the product sentence carries a messy-state term from the lexicon, as a restated problem or as a contrast such as no more manual work. |
| Visual framework | kinetic type · flowchart or network diagram · illustration, abstract or b-roll · product interface after · data chart after · stock office · presenter · messy-text interface · split or before-and-after label · spreadsheet, inbox or paper label | Coded from timed segment motifs and on-screen text overlapping the window before the product sentence, or around the mess sentence when the product comes first, and the window after the product sentence. Only videos with a full visual record carry these codes. |
| Sound and frame | voiceover · runtime · brightness 0 to 255 · light, mixed or dark frames | Voiceover from audio analysis; runtime from metadata; brightness and light, mixed or dark computed over nine sampled frames, not the interface alone. |
| Not coded | cast · altitude · surface · build | Who appears and what the video argues are not recorded in the library. Surface and build exist only for the visual-record videos, so they are not coded for the set. |
Cite this report
Jai Ghosh, Advids. 118 Chaos to Clean B2B Videos Decoded: The 4 Core Visual Frameworks Used to Demonstrate Productivity. 14 September 2026. https://advids.co/blog/chaos-to-clean-b2b-video-frameworks
Author & editor bio
About Advids
Advids produces video for AI, SaaS and DeepTech companies. ADVIDS Research publishes coded studies of what those companies ship, with the dataset attached so every count can be checked.
Image credits and licensing
Video frames on this page are reproduced for analysis and commentary. Copyright in each frame remains with the company credited in its caption; to license a frame, contact that rights holder. Advids grants no rights over them.
Charts and images credited ADVIDS Research, and the coded dataset behind them, are © 2026 Advids and licensed CC BY 4.0: reuse them with attribution to Advids and a link to this page.