Deal economics
Supplier data export cost decision: cheaper routes before a full pull
By SourceX Editorial · Reviewed by Noah Loul ·
9 sections
Short answer
Before deciding an export costs too much, verify four things: the cheapest route that keeps what a buyer's test needs, the narrowest scope a buyer has actually named, which hours you would keep even without a deal, and the conditions and payment triggers in the terms you would sign. Most full-export work can wait until an agreement is signed, so a high first estimate is rarely the real decision.
Value drivers
Supplier data export cost
Key takeaways
- A first export estimate usually prices the whole archive through the slowest route, before any buyer has named a scope.
- The same records can leave through a native export, an API, an existing warehouse copy or vendor help, and each route keeps and drops different history.
- Sort every effort line into no-regret, reusable or deal-specific hours; only the first two belong before a signed agreement.
- Eligibility, an indicative offer and a signed deal are different stages, and only signed terms set conditions and payment triggers.
- Decide with a one-page memo: proceed, narrow, defer or decline, with the reason and the fact that would change it.
What should a company verify before calling the export cost too high?
Verify the route, the scope, how much of the work you would keep without a deal, and the signed terms before treating a high export estimate as a reason to stop. In a supplier data export cost decision, the first estimate tends to err in the expensive direction: it assumes the whole archive, every field, full history and attachments, pulled through whichever route the estimator knows best, before any buyer has said what it needs.
The decision itself is narrow. A company believes some of its records could be licensed for AI training or evaluation, someone has estimated the work to get them out, and the number looks too large for an uncertain deal. The useful output is a decision backed by evidence: proceed on a cheaper route, narrow the scope, defer the bulk work until terms are signed, or decline because even the cheapest credible route costs more than the deal could carry.
Why the first export estimate is usually the most expensive one
Because it is built before a scope exists. A buyer's request names scope, volume, format and timing, and SourceX's published guidance sizes what one system usually has to supply: tens of thousands of records or more, across a consistent stretch of history that typically runs two years or longer, with each deal assessed on its own. A request framed that way rarely needs every system, every year and every attachment.
The estimate also front-loads work that a deal sequence spreads out. In SourceX's process the first step is a short company assessment with no data uploads, and nothing is delivered until the company has approved the scope and terms and signed. Work priced as if it all happens before an offer is priced against the most uncertain moment in the process.
Rows are also not the main thing a buyer pays for. An audit led by MIT researchers of more than 1,800 text datasets found license information missing for over two-thirds of datasets on widely used hosting sites and error rates above 50 percent, with corrected licenses often more restrictive than the ones shown. A smaller export with clear documentation of where the records came from and who may license them is worth more than a larger one without it.
Four export routes and what each one costs you
The same records can usually leave a system by four routes, and the cheapest credible one is rarely the route in the first estimate. Each route has its own cost driver and its own blind spots, and both come from the vendor's documentation rather than from assumptions.
Salesforce's Data Export service, for example, writes CSV files into zip archives of up to about 512 MB each, can run every 7 days on Enterprise, Performance and Unlimited editions or every 29 days on Professional and Developer editions, and deletes the files 48 hours after the notification email, weekends excluded. Zendesk limits its incremental export endpoints to 10 requests per minute, with a paid High Volume API add-on raising the limit, so a full ticket history through the API costs calendar time as well as engineering time. Slack lets workspace owners and admins on every plan export public channels, while exports of private channels and direct messages need an approved application, and the Free plan exports only links to files from the last 90 days.
Compare the routes on the same four columns before drawing any conclusion about cost.
| Route | Main cost driver | What it tends to keep or drop | Verify before choosing |
|---|---|---|---|
| Native scheduled export | Admin time to schedule, download and store files before they expire | Keeps object records in flat files; history and links depend on what the export covers | Edition cadence, file expiry window, objects included |
| API extraction | Engineering time plus calendar time under rate limits | Can reach archived and historical records the user interface hides | Published rate limits, incremental endpoints, add-on terms |
| Existing warehouse or backup copy | Mapping and checking a copy the company already pays for | Keeps whatever the sync captured, often without attachments or deleted records | Sync start date, field coverage, gaps after migrations |
| Vendor help or a plan change | Fees, plan upgrades or professional services | Can unlock content other routes cannot reach, such as private conversations | Contract export clauses, approval steps, consent requirements |
History, links and attachments: what the cheap route can lose
A cheaper route is only cheaper if it keeps what the buyer's acceptance test will check, and history is where routes differ most. With Field Audit Trail turned off, Salesforce keeps field history for up to 18 months, or up to 24 months through the API, and history between 18 and 24 months old is retrieved with Data Loader or the queryAll API call. Waiting has a cost of its own: field history past the retention window may not be retrievable by any route chosen later.
Archived records are the second trap. Zendesk generally archives tickets 120 days after they close; archived tickets disappear from views, but most ticket API endpoints and the incremental export still return them. A report built from views can look complete while silently omitting years of closed tickets, which the slower API route would have kept.
Permissions are the third. Under Slack's rules a public-channel export is available on any plan, but private channels need an approved application, and Slack rejects applications from Free and Pro workspaces unless the owner shows valid legal process, member consent or a legal right to the data. If the buyer's task does not need private conversations, leaving them out is cheaper and simpler on every count.
The effort worksheet: no-regret, reusable and deal-specific hours
Sort every hour of export work by what survives if no deal closes. No-regret hours pay off anyway, through retention, migration or audit work the company needs regardless. Reusable hours build something a second buyer could use, such as extraction scripts or a field map. Deal-specific hours serve one buyer's scope and are lost if that deal falls through.
Illustrative example: a 300-person field services company holds six years of help desk tickets and CRM accounts. A full pull of everything before any offer comes to 160 staff hours, 80 of them deal-specific. The same records scoped to the systems and years a buyer names, with the bulk work after signing, come to 80 hours, of which 20 fall before signing. Every figure is an assumption to replace, not a measurement.
Copy the worksheet, swap in your own estimates, keep the category column and total the hours that would fall before a signed agreement.
| Effort line | Category | Illustrative hours, full pull up front | Illustrative hours, scoped and gated | When, on the gated plan |
|---|---|---|---|---|
| System inventory and field map | No-regret | 12 | 12 | Before signing |
| Extraction scripts or export jobs | Reusable | 40 | 24 | After signing |
| Extraction runs, retries and monitoring | Deal-specific | 36 | 10 | After signing |
| Attachments and archived records | Deal-specific | 30 | 6 | After signing, only if the test needs them |
| Linking tickets to accounts and outcomes | Reusable | 20 | 12 | After signing |
| Sample extraction and review | Reusable | 8 | 8 | Before signing |
| Packaging and delivery logistics | Deal-specific | 14 | 8 | After signing |
Eligibility, an indicative offer and a signed deal carry different economics
Only signed terms attach payment to export work, so reconcile the effort with the terms you would actually sign rather than with eligibility or an indicative number. Eligibility means the company and dataset fit what buyers look for; an indicative offer is a starting position. Neither commits anyone to selection, closing or a particular amount.
Four items in the signed terms decide whether the export cost is justified: the exact dataset scope in system terms, the conditions that must be met before payment, the payment triggers themselves, and which party carries each preparation step. Read each one in the draft agreement, not in a summary email, and price the export against that scope alone.
Some steps may not belong on the supplier's worksheet at all. Under SourceX's standing terms, removing identifiers falls to SourceX, which does it itself or through its processing partner, so those hours need not sit on the supplier's sheet. Exclusivity and term are settled in the signed agreement for each deal, which makes exclusivity a clause to read rather than an effort to estimate.
Is the export cost high, or just early? Matching spend to each gate
Many export costs that look too high are early rather than high: they are priced before the gate that would justify them. Match each category of hours to the earliest gate at which losing it would be acceptable.
At a glance
Is the export cost high, or just early? Matching spend to each gate
The carry test is simple to state: before signing, commit only the hours you would accept losing. In the illustrative worksheet that puts 20 hours at risk on the gated plan against 160 on the full pull.
- Before any assessment: nothing beyond knowing which systems hold the records and roughly how many years they cover.
- After a fit check: no-regret work only, the inventory and field map, which help retention and migration planning even if no deal follows.
- After a buyer names a scope: a sample extraction and review on the cheapest route that keeps the fields the buyer's test checks.
- After the agreement is signed: the reusable and deal-specific hours, run against the signed scope.
- At any gate: stop if the cheapest credible route at the signed scope would cost more than the company would accept losing should the deal not close.
The supplier decision memo: proceed, narrow, defer or decline
Write the decision on one page, with the evidence behind it and the single fact that would reverse it. The same fields work for one system or a whole archive.
- Dataset scope in system terms: systems, objects, date range and fields, as the buyer's request or the draft terms state them.
- Route chosen and route rejected, with the documentation fact that decided it, such as an export cadence, a rate limit or an approval requirement.
- Effort by owner, split into no-regret, reusable and deal-specific hours, with the total that falls before signing.
- Terms check: conditions, payment triggers and preparation responsibilities as written in the terms you would sign.
- Decision: proceed on the chosen route, narrow to the named scope, defer bulk work until signing, or decline because the cheapest credible route exceeds what the company would accept losing.
- Reversal trigger: the one fact that would change the decision, such as a cheaper route confirmed in a sample run or a narrower scope in the draft terms.
- Limits: estimates stay assumptions until a sample run measures them, and nothing in an assessment or indicative offer commits a buyer to close.
When export cost is the sticking point with SourceX
SourceX helps companies turn selected business records into licensed datasets for AI, and the work it takes on spans sourcing and qualification, rights review, contracting, preparing the data, delivering it and handling payment. Its first step is a short company assessment with no data uploads, so the inventory question can be answered before any export work starts.
Nothing reaches SourceX before the company has approved both the scope and the terms and signed the agreement, so the full pull can be timed to that signature. A large delivery can remain in the seller's own storage or go out on encrypted drives, because SourceX does not host multi-terabyte datasets; that can remove a cost a first estimate may include, building somewhere to send the data.
An assessment is a fit check, not an offer, and it does not promise selection, closing or any particular amount. Counsel should review the specific contracts and draft terms before any export or license.
Frequently asked questions
How much export work should happen before a data licensing deal is signed?
Only work you would keep if the deal fell through: a system inventory, a field map and, once a buyer names a scope, a sample run on the cheapest route. Bulk extraction, attachments and packaging belong after signing, against the signed scope.
Can a company export its entire Slack workspace for licensing?
Not always on its current plan. Slack lets owners and admins on every plan export public channels, but exports of private channels and direct messages need an approved application, and Free and Pro workspaces must show valid legal process, member consent or a legal right. Check whether the buyer's task needs private conversations before paying for a plan change or applying.
Why can an API export be worth more than a native export?
An API pull costs more engineering time, but it can reach records that a view-based or report-based export misses, such as Zendesk tickets archived after closing, which drop out of views but remain available through most ticket API endpoints. If the buyer's test needs those records, the cheaper-looking route may have to be run twice.
Who performs de-identification when a company licenses data through SourceX?
SourceX's standing terms give the de-identification work to SourceX, carried out directly or by its processing partner. Scope remains the supplying company's decision, and no data is delivered until the company has approved and signed the terms.
Is an indicative data offer worth paying for a full export?
Not on its own. An indicative offer is a starting position that commits no one to selection, closing or a particular amount, and only signed terms set conditions and payment triggers. Spend no more before signing than you would accept losing, and use the indicative stage to settle scope and route.
Sources
- Salesforce Data Export: CSV files in zip archives of up to about 512 MB; weekly exports every 7 days on Enterprise, Performance and Unlimited editions, every 29 days on Professional and Developer; files deleted 48 hours after the email, excluding weekends. Source
- With Field Audit Trail off, Salesforce retains field history for up to 18 months, and up to 24 months via the API; 18 to 24 month history is retrieved with Data Loader or queryAll(). Source
- Zendesk's incremental export endpoints are rate limited to 10 requests per minute, with higher limits under the High Volume API add-on. Source
- Zendesk generally archives tickets 120 days after they are closed; archived tickets are excluded from views but returned by most ticket API endpoints and the Incremental Exports API. Source
- Slack: owners and admins on all plans can export public channels; private channels and DMs need an approved application; Free and Pro applications need legal process, member consent or a legal right; Free plan exports file links from the last 90 days only. Source
- An audit of 1,800+ text datasets (Nature Machine Intelligence, 2024) found license omission above two-thirds on widely used hosting sites and error rates above 50 percent, with corrected licenses often more restrictive. Source
See if your company qualifies
A short company assessment. No data uploads are needed.