Why Your Integration Quote Is Wrong

The quote priced the connection. The work was always in the mapping.

You were told four to six weeks for a standard integration. It is month five. The connection technically works, but the SKUs coming across don’t match the ones on the rack, three of your customers send order files that break the parser, and somebody on your team now spends Tuesday mornings reconciling counts by hand, which was not in anyone’s job description in March.

Nobody lied to you here; the number priced the wrong scope from the outset.

What does an integration quote actually price?

Every integration quote you’ve ever seen was built on the same four assumptions: one direction of data flow, the current version of both systems, the happy path, and a sample file that somebody cleaned up before sending it over. Reasonable assumptions, all of them, and not one survives contact with a live warehouse.

Getting two systems to exchange a message is close to solved at this point, so the connection was never the expensive part. What costs you five months is getting your ERP, your customer, and the label on the pallet to agree on what the message actually means when one calls it a unit and the other calls it a case.

What costs are usually missing from an integration quote?

Your data isn’t shaped like the sample. Duplicate SKU codes, location codes that stopped matching the rack two reslots ago, unit of measure tables with conversions nobody remembers writing and everybody’s afraid to touch. None of that shows up while you’re still looking at a cleaned-up sample file, and by the time it does, you’re mid-project instead of mid-scoping. One firm running mid-market deployments puts roughly 40 percent of total project effort on the customer’s own side, most of it data cleanup that’s easy to miss until someone actually goes looking.

The systems you don’t control. If you’re a 3PL, half your integration surface belongs to your customers, and they may not be able to change much on their end no matter how the request is framed. Deposco tells the story of a 3PL that eventually built its own integration layer rather than keep working around this. That’s not a vendor missing something. That’s just what running integrations for other people’s customers looks like, and it’s solvable once you plan for it instead of hoping around it.

Drift and ownership. Connections break when either side upgrades, and somebody always upgrades, usually without asking you first. The quote covers building the connection. It almost never says who’s fixing it at 2am during peak, or what that costs you every year after go-live.

Testing and the freeze window. You can’t test a peak season integration during peak season, and you can’t go live in October either, no matter how badly the project plan wants you to. Slip past September and you haven’t lost six weeks, you’ve lost until January.

Add it up and the published ranges stop sounding pessimistic. Deposco puts hidden custom development at 40 to 60 percent beyond initial estimates. One implementation guide puts total implementation cost at 50 to 200 percent of license fees, another puts hidden costs at 25 to 50 percent above the initial budget. Vendor sources, all of them, so weigh that accordingly, but they’re also the ones who’ve actually run these projects, and none of them are landing anywhere close to the original quote.

Does a pre-built connector solve this?

“We have a pre-built connector for that.”

Usually true, and usually beside the point. A pre-built connector handles transport: authentication, endpoints, retries, keeping the pipe open without falling over. Worth having, but it was never the hard part.

Mapping is where your five months went, and a connector doesn’t touch it. Two facilities running the same WMS from the same vendor can still have different field usage, different naming conventions, different ideas about what belongs in a free text note. The connector is a road. Somebody still has to draw your map, facility by facility.

A logo wall tells you a vendor has connected to your system before. It won’t tell you what happens the day your mapping turns out weirder than their sample file ever showed, and that’s the question actually worth asking.

Eight questions to ask before you sign

1. What specifically is in scope, field by field, and what is explicitly out?
2. Which direction does data flow, and what happens on a write failure?
3. Who owns the mapping when our data doesn’t match your assumptions: you, us, or a change order?
4. Show me the last three deployments on this same system. How long did each actually take, measured from kickoff to the integration being signed off, not to go live?
5. What is the annual cost of keeping this connection running, including version upgrades on either side?
6. What is your escalation path during peak, with response times in writing?
7. What do you need from my team, in hours per week, and from whom?
8. If we part ways in two years, what do I keep and in what format?

Question four is the one that separates real answers from optimistic ones. Ask for it in writing.

What does a good integration quote look like?

A good answer isn’t a smaller number, it’s a number with its assumptions attached, a named owner for the mapping, and a range instead of a single point estimate. Anyone giving you a confident single figure before they’ve seen your actual data is quoting a template.

The honest version of this from our side: we assume the fragmentation is permanent. There is no dominant WMS or ERP in this industry, our customers’ customers run whatever they run, and nobody is standardizing. That assumption is why we connect to the systems and cameras already in the building rather than asking you to change them.

You’re still entitled to ask us all eight questions. Especially number four, and especially if the answer takes a while to come back.

FAQ

Why do integration projects usually take longer than the quote said? Most integration quotes price the connection between two systems, not the work of mapping each side’s data to match. Data cleanup, systems outside your control, ongoing maintenance, and seasonal freeze windows are rarely included in the original estimate, and they’re usually where the extra time goes.

What percentage of integration costs are typically hidden from the initial quote? Published estimates from implementation vendors put hidden costs anywhere from 25 to 60 percent above the original quote, with one source putting total implementation cost as high as 50 to 200 percent of license fees. These figures come from vendor sources, so they should be weighed accordingly, but they consistently point in the same direction.

Does a pre-built connector guarantee a fast integration? No. A pre-built connector handles transport, meaning authentication, endpoints, and keeping the connection open. It doesn’t handle mapping, which is where most of the time on an integration project actually goes, especially across facilities with different naming conventions and data practices.

What should you ask a vendor before signing an integration contract? Ask what’s explicitly in and out of scope, who owns data mapping when it doesn’t match assumptions, how long the last three similar deployments actually took from kickoff to sign-off, what ongoing maintenance costs annually, and what you keep if the relationship ends.